USENIX Security '22 has three submission deadlines. Prepublication versions of the accepted papers from the fall submission deadline are available below. The full program will be available soon.
Security at the End of the Tunnel: The Anatomy of VPN Mental Models Among Experts and Non-Experts in a Corporate Context
Veroniek Binkhorst, Technical University of Delft; Tobias Fiebig, Max-Planck-Institut für Informatik and Technical University of Delft; Katharina Krombholz, CISPA Helmholtz Center for Information Security; Wolter Pieters, Radboud University; Katsiaryna Labunets, Utrecht University
With the worldwide COVID-19 pandemic in 2020 and 2021 necessitating working from home, corporate Virtual Private Networks (VPNs) have become an important item securing the continued operation of companies around the globe. However, due to their different use case, corporate VPNs and how users interact with them differ from public VPNs, which are now commonly used by end-users.
In this paper, we present a first explorative study of eleven experts' and seven non-experts' mental models in the context of corporate VPNs. We find a partial alignment of these models in the high-level technical understanding while diverging in important parameters of how, when, and why VPNs are being used. While, in general, experts have a deeper technical understanding of VPN technology, we also observe that even they sometimes hold false beliefs on security aspects of VPNs. In summary, we show that the mental models of corporate VPNs differ from those for related security technology, e.g., HTTPS.
Our findings allow us to draft recommendations for practitioners to encourage a secure use of VPN technology (through training interventions, better communication, and system design changes in terms of device management). Furthermore, we identify avenues for future research, e.g., into experts' knowledge and balancing privacy and security between system operators and users.
GAROTA: Generalized Active Root-Of-Trust Architecture (for Tiny Embedded Devices)
Esmerald Aliaj, University of California, Irvine; Ivan De Oliveira Nunes, Rochester Institute of Technology; Gene Tsudik, University of California, Irvine
Embedded (aka smart or IoT) devices are increasingly popular and becoming ubiquitous. Unsurprisingly, they are also attractive attack targets for exploits and malware. Low-end embedded devices, designed with strict cost, size, and energy limitations, are especially challenging to secure, given their lack of resources to implement sophisticated security services, available on higher-end computing devices. To this end, several tiny Roots-of-Trust (RoTs) were proposed to enable services, such as remote verification of device's software state and run-time integrity. Such RoTs operate reactively: they can prove whether a desired action (e.g., software update or program execution) was performed on a specific device. However, they can not guarantee that a desired action will be performed, since malware controlling the device can trivially block access to the RoT by ignoring/discarding received commands and other trigger events. This is an important problem because it allows malware to effectively "brick" or incapacitate a potentially huge number of (possibly mission-critical) devices.
Though recent work made progress in terms of incorporating more active behavior atop existing RoTs, much of it relies on extensive hardware support in the form of Trusted Execution Environments (TEEs), which are generally too costly for low-end devices. In this paper, we set out to systematically design a minimal active RoT for low-end MCU-s. We begin with three questions: (1) What functionality is required to guarantee actions in the presence of malware? (2) How to implement this efficiently? and (3) What are the security benefits of such an active RoT architecture? We then design, implement, formally verify, and evaluate GAROTA : Generalized Active Root-Of-Trust Architecture. We believe that GAROTA is the first clean-slate design of an active RoT for low-end MCU-s. We show how GAROTA guarantees that even a fully software-compromised low-end MCU performs a desired action. We demonstrate its practicality by implementing GAROTA in the context of three types of applications where actions are triggered by: sensing hardware, network events and timers. We also formally specify and verify GAROTA functionality and properties.
A Large-scale and Longitudinal Measurement Study of DKIM Deployment
Chuhan Wang, Kaiwen Shen, and Minglei Guo, Tsinghua University; Yuxuan Zhao, North China Institute of Computing Technology; Mingming Zhang, Jianjun Chen, and Baojun Liu, Tsinghua University; Xiaofeng Zheng and Haixin Duan, Tsinghua University and Qi An Xin Technology Research Institute; Yanzhong Lin and Qingfeng Pan, Coremail Technology Co. Ltd
DomainKeys Identified Mail (DKIM) is an email authentication protocol to protect the integrity of email contents. It has been proposed and standardized for over a decade and adopted by Yahoo!, Google, and other leading email service providers. However, little has been done to understand the adoption rate and potential security issues of DKIM due to the challenges of measuring DKIM deployment at scale.
In this paper, we provide a large-scale and longitudinal measurement study on how well DKIM is deployed and managed. Our study was made possible by a broad collection of datasets, including 9.5 million DKIM records from passive DNS datasets over five years and 460 million DKIM signatures from real-world email headers. Moreover, we conduct an active measurement on Alexa Top 1 million domains. Our measurement results show that 28.1% of Alexa Top 1 million domains have enabled DKIM, of which 2.9% are misconfigured. We demonstrate that the issues of DKIM key management and DKIM signatures are prevalent in the real world, even for well-known email providers (e.g., Gmail and Mail.ru). We recommend the security community should pay more attention to the systemic problems of DKIM deployment and mitigate these issues from the perspective of protocol design.
Neither Access nor Control: A Longitudinal Investigation of the Efficacy of User Access-Control Solutions on Smartphones
Masoud Mehrabi Koushki, Yue Huang, Julia Rubin, and Konstantin Beznosov, University of British Columbia
The incumbent all-or-nothing model of access control on smartphones has been known to dissatisfy users, due to high overhead (both cognitive and physical) and lack of device-sharing support. Several alternative models have been proposed. However, their efficacy has not been evaluated and compared empirically, due to a lack of detailed quantitative data on users' authorization needs. This paper bridges this gap with a 30-day diary study. We probed a near-representative sample (N = 55) of US smartphone users to gather a comprehensive list of tasks they perform on their phones and their authorization needs for each task. Using this data, we quantify, for the first time, the efficacy of the all-or-nothing model, demonstrating frequent unnecessary or missed interventions (false positive rate (FPR) = 90%, false negative rate (FNR) = 21%). In comparison, we show that app- or task-level models can improve the FPR up to 88% and the FNR up to 20%, albeit with a modest (up to 15%) increase in required upfront configuration. We also demonstrate that the context in which phone sharing happens is consistent up to 75% of the time, showing promise for context-based solutions.
Cheetah: Lean and Fast Secure Two-Party Deep Neural Network Inference
Zhicong Huang, Wen-jie Lu, Cheng Hong, and Jiansheng Ding, Alibaba Group
Secure two-party neural network inference (2PC-NN) can offer privacy protection for both the client and the server and is a promising technique in the machine-learning-as-a-service setting. However, the large overhead of the current 2PC-NN inference systems is still being a headache, especially when applied to deep neural networks such as ResNet50. In this work, we present Cheetah, a new 2PC-NN inference system that is faster and more communication-efficient than state-of-the-arts. The main contributions of Cheetah are two-fold: the first part includes carefully designed homomorphic encryption-based protocols that can evaluate the linear layers (namely convolution, batch normalization, and fully-connection) without any expensive rotation operation. The second part includes several lean and communication-efficient primitives for the non-linear functions (e.g., ReLU and truncation). Using Cheetah, we present intensive benchmarks over several large-scale deep neural networks. Take ResNet50 for an example, an end-to-end execution of Cheetah under a WAN setting costs less than 2.5 minutes and 2.3 gigabytes of communication, which outperforms CrypTFlow2 (ACM CCS 2020) by about 5.6× and 12.9×, respectively.
Inferring Phishing Intention via Webpage Appearance and Dynamics: A Deep Vision Based Approach
Ruofan Liu, Yun Lin, Xianglin Yang, and Siang Hwee Ng, National University of Singapore; Dinil Mon Divakaran, Trustwave; Jin Song Dong, National University of Singapore
Explainable phishing detection approaches are usually based on references, i.e., they compare a suspicious webpage against a reference list of commonly targeted legitimate brands' webpages. If a webpage is detected as similar to any referenced website but their domains are not aligned, a phishing alert is raised with an explanation comprising its targeted brand. In comparison to other techniques, such explainable reference-based solutions are more robust to ever-changing phishing webpages. However, the webpage similarity is still measured by representations conveying only partial intentions (e.g., screenshot and logo), which (i) incurs considerable false positives and (ii) gives an adversary opportunities to compromise user confidence in the approaches.
In this work, we propose, PhishIntention, to extract precise phishing intention of a webpage by visually (i) extracting its brand intention and credential-taking intention, and (ii) interacting with the webpage to confirm the credential-taking intention. We design PhishIntention as a heterogeneous system of deep learning vision models, overcoming various technical challenges. The models "look at" and "interact with" the webpage for its intention, which are robust to potential HTML obfuscation. We compare PhishIntention with four state-of-the-art reference-based approaches on the largest phishing identification dataset consisting of 50K phishing and benign webpages. For similar level of recall, PhishIntention achieves significantly higher precision than the baselines. Moreover, we conduct a continuous field study on the Internet for two months to discover emerging phishing webpages. PhishIntention detects 1,942 new phishing webpages (1,368 not reported by VirusTotal). Comparing to the best baseline, PhishIntention generates 86.5% less false alerts (139 vs. 1,033 false positives) while detecting similar number of real phishing webpages.
Electronic Monitoring Smartphone Apps: An Analysis of Risks from Technical, Human-Centered, and Legal Perspectives
Kentrell Owens, University of Washington; Anita Alem, Harvard Law School; Franziska Roesner and Tadayoshi Kohno, University of Washington
Electronic monitoring is the use of technology to track individuals accused or convicted of a crime (or civil violation) as an "alternative to incarceration." Traditionally, this technology has been in the form of ankle monitors, but recently federal, state, and local entities around the U.S. are shifting to using smartphone applications for electronic monitoring. These applications (apps) purport to make the monitoring simpler and more convenient for both the community supervisor and the person being monitored. However, due to the multipurpose nature of smartphones in people's lives and the amount of sensitive information (e.g., sensor data) smartphones make available, this introduces new risks to people coerced to use these apps.
To understand what type of privacy-related and other risks might be introduced to people who use these applications, we conducted a privacy-oriented analysis of 16 Android apps used for electronic monitoring. We analyzed the apps first technically, with static and (limited) dynamic analysis techniques. We also analyzed user reviews in the Google Play Store to understand the experiences of the people using these apps, and also the privacy policies. We found that apps contain numerous trackers, the permissions requested by them vary widely (with the most common one being location), and the reviews indicate that people find the apps invasive and frequently dysfunctional. We end the paper by encouraging mobile app marketplaces to reconsider their role in the future of electronic monitoring apps, and computer security and privacy researchers to consider their potential role in auditing carceral technologies. We hope that this work will lead to more transparency in this obfuscated ecosystem.
ppSAT: Towards Two-Party Private SAT Solving
Ning Luo, Samuel Judson, Timos Antonopoulos, and Ruzica Piskac, Yale University; Xiao Wang, Northwestern University
We design and implement a privacy-preserving Boolean satisfiability (ppSAT) solver, which allows mutually distrustful parties to evaluate the conjunction of their input formulas while maintaining privacy. We first define a family of security guarantees reconcilable with the (known) exponential complexity of SAT solving, and then construct an oblivious variant of the classic DPLL algorithm which can be integrated with existing secure two-party computation (2PC) techniques. We further observe that most known SAT solving heuristics are unsuitable for 2PC, as they are highly data-dependent in order to minimize the number of exploration steps. Faced with how best to trade off between the number of steps and the cost of obliviously executing each one, we design three efficient oblivious heuristics, one deterministic and two randomized. As a result of this effort we are able to evaluate our ppSAT solver on small but practical instances arising from the haplotype inference problem in bioinformatics. We conclude by looking towards future directions for making ppSAT solving more practical, most especially the integration of conflict-driven clause learning (CDCL).
"Like Lesbians Walking the Perimeter": Experiences of U.S. LGBTQ+ Folks With Online Security, Safety, and Privacy Advice
Christine Geeng and Mike Harris, University of Washington; Elissa Redmiles, Max Planck Institute for Software Systems; Franziska Roesner, University of Washington
Given stigma and threats surrounding being gay or transgender, LGBTQ+ folks often seek support and information on navigating identity and personal (digital and physical) safety. While prior research on digital security advice focused on a general population and general advice, our work focuses on queer security, safety, and privacy advice-seeking to determine population-specific needs and takeaways for broader advice research. We conducted qualitative semi-structured interviews with 14 queer participants diverse across race, age, gender, sexuality, and socioeconomic status. We find that participants turn to their trusted queer support groups for advice, since they often experienced similar threats. We also document reasons that participants sometimes reject advice, including that it would interfere with their material livelihood and their potential to connect with others. Given our results, we recommend that queer-specific and general security and safety advice focus on specificity—why and how—over consistency, because advice cannot be one-size-fits-all. We also discuss the value of intersectionality as a framework for understanding vulnerability to harms in security research, since our participants' overlapping identities affected their threat models and advice perception.
CamShield: Securing Smart Cameras through Physical Replication and Isolation
Zhiwei Wang, Yihui Yan, and Yueli Yan, ShanghaiTech University; Huangxun Chen, Huawei Theory Lab; Zhice Yang, ShanghaiTech University
Smart home devices, such as security cameras, are equipped with visual sensors, either for monitoring or improving user experience. Due to the sensitivity of the home environment, their visual sensing capabilities cause privacy and security concerns. In this paper, we design and implement the CamShield, a companion device to guarantee the privacy of smart security cameras, even if the whole camera system is fully compromised. At a high level, the CamShield is a shielding case that works by attaching it to the front of the security camera to blind it. Then, it uses its own camera for visual recording. The videos are first protected according to user-specified policies, and then transmitted to the security camera and hence to the Internet through a Visible Light Communication (VLC) channel. It ensures that only the authorized entities have full access to the protected videos. Since the CamShield is physically isolated from the shielded security camera and the Internet, it naturally resists many known attacks and can operate as it is expected to.
PatchCleanser: Certifiably Robust Defense against Adversarial Patches for Any Image Classifier
Chong Xiang, Saeed Mahloujifar, and Prateek Mittal, Princeton University
The adversarial patch attack against image classification models aims to inject adversarially crafted pixels within a restricted image region (i.e., a patch) for inducing model misclassification. This attack can be realized in the physical world by printing and attaching the patch to the victim object; thus, it imposes a real-world threat to computer vision systems. To counter this threat, we design PatchCleanser as a certifiably robust defense against adversarial patches. In PatchCleanser, we perform two rounds of pixel masking on the input image to neutralize the effect of the adversarial patch. This image-space operation makes PatchCleanser compatible with any state-of-the-art image classifier for achieving high accuracy. Furthermore, we can prove that PatchCleanser will always predict the correct class labels on certain images against any adaptive white-box attacker within our threat model, achieving certified robustness. We extensively evaluate PatchCleanser on the ImageNet, ImageNette, and CIFAR-10 datasets and demonstrate that our defense achieves similar clean accuracy as state-of-the-art classification models and also significantly improves certified robustness from prior works. Remarkably, PatchCleanser achieves 83.9% top-1 clean accuracy and 62.1% top-1 certified robust accuracy against a 2%-pixel square patch anywhere on the image for the 1000-class ImageNet dataset.
Phish in Sheep's Clothing: Exploring the Authentication Pitfalls of Browser Fingerprinting
Xu Lin, Panagiotis Ilia, Saumya Solanki, and Jason Polakis, University of Illinois at Chicago
As users navigate the web they face a multitude of threats; among them, attacks that result in account compromise can be particularly devastating. In a world fraught with data breaches and sophisticated phishing attacks, web services strive to fortify user accounts by adopting new mechanisms that identify and prevent suspicious login attempts. More recently, browser fingerprinting techniques have been incorporated into the authentication workflow of major services as part of their decision-making process for triggering additional security mechanisms (e.g., two-factor authentication).
In this paper we present the first comprehensive and in-depth exploration of the security implications of real-world systems relying on browser fingerprints for authentication. Guided by our investigation, we develop a tool for automatically constructing fingerprinting vectors that replicate the process of target websites, enabling the extraction of fingerprints from users' devices that exactly match those generated by target websites. Subsequently, we demonstrate how phishing attackers can replicate users' fingerprints on different devices to deceive the risk-based authentication systems of high-value web services (e.g., cryptocurrency trading) to completely bypass two-factor authentication. To gain a better understanding of whether attackers can carry out such attacks, we study the evolution of browser fingerprinting practices in phishing websites over time. While attackers do not generally collect all the necessary fingerprinting attributes, unfortunately that is not the case for attackers targeting certain financial institutions where we observe an increasing number of phishing sites capable of pulling off our attacks. To address the significant threat posed by our attack, we have disclosed our findings to the vulnerable vendors.
FreeWill: Automatically Diagnosing Use-after-free Bugs via Reference Miscounting Detection on Binaries
Liang He, TCA, Institute of Software, Chinese Academy of Sciences; Hong Hu, Pennsylvania State University; Purui Su, TCA / SKLCS, Institute of Software, Chinese Academy of Sciences and School of Cyber Security, University of Chinese Academy of Sciences; Yan Cai, SKLCS, Institute of Software, Chinese Academy of Sciences; Zhenkai Liang, National University of Singapore
Memory-safety issues in operating systems and popular applications are still top security threats. As one widely exploited vulnerability, Use After Free (UAF) resulted in hundreds of new incidents every year. Existing bug diagnosis techniques report the locations that allocate or deallocate the vulnerable object, but cannot provide sufficient information for developers to reason about a bug or synthesize a correct patch.
In this work, we identified incorrect reference counting as one common root cause of UAF bugs: if the developer forgets to increase the counter for a newly created reference, the program may prematurely free the actively used object, rendering other references dangling pointers. We call this problem reference miscounting. By proposing an omission-aware counting model, we developed an automatic method, FREEWILL, to diagnose UAF bugs. FREEWILL first reproduces a UAF bug and collects related execution trace. Then, it identifies the UAF object and related references. Finally, FREEWILL compares reference operations with our model to detect reference miscounting. We evaluated FREEWILL on 76 real-world UAF bugs and it successfully confirmed reference miscounting as root causes for 48 bugs and dangling usage for 18 bugs. FREEWILL also identified five null-pointer dereference bugs and failed to analyze five bugs. FREEWILL can complete its analysis within 15 minutes on average, showing its practicality for diagnosing UAF bugs.
ReZone: Disarming TrustZone with TEE Privilege Reduction
David Cerdeira and José Martins, Centro ALGORITMI, Universidade do Minho; Nuno Santos, INESC-ID / Instituto Superior Técnico, Universidade de Lisboa; Sandro Pinto, Centro ALGORITMI, Universidade do Minho
In TrustZone-assisted TEEs, the trusted OS has unrestricted access to both secure and normal world memory. Unfortunately, this architectural limitation has opened an aisle of exploration for attackers, which have demonstrated how to leverage a chain of exploits to hijack the trusted OS and gain full control of the system, targeting (i) the rich execution environment (REE), (ii) all trusted applications (TAs), and (iii) the secure monitor. In this paper, we propose ReZone. The main novelty behind ReZone design relies on leveraging TrustZone-agnostic hardware primitives available on commercially off-the-shelf (COTS) platforms to restrict the privileges of the trusted OS. With ReZone, a monolithic TEE is restructured and partitioned into multiple sandboxed domains named zones, which have only access to private resources. We have fully implemented ReZone for the i.MX 8MQuad EVK and integrated it with Android OS and OP-TEE. We extensively evaluated ReZone using microbenchmarks and real-world applications. ReZone can sustain popular applications like DRM-protected video encoding with acceptable performance overheads. We have surveyed 80 CVE vulnerability reports and estimate that ReZone could mitigate 86.84% of them.
Double Trouble: Combined Heterogeneous Attacks on Non-Inclusive Cache Hierarchies
Antoon Purnal, Furkan Turan, and Ingrid Verbauwhede, imec-COSIC, KU Leuven
As the performance of general-purpose processors faces diminishing improvements, computing systems are increasingly equipped with domain-specific accelerators. Today's high-end servers tightly integrate such accelerators with the CPU, e.g., giving them direct access to the CPU's last-level cache (LLC).
Caches are an important source of information leakage across security domains. This work explores combined cache attacks, complementing traditional co-tenancy with control over one or more accelerators. The constraints imposed on these accelerators, originally perceived as limitations, turn out to be advantageous to an attacker. We develop a novel approach for accelerators to find eviction sets, and leverage precise double-sided control over cache lines to expose undocumented behavior in non-inclusive Intel cache hierarchies.
We develop a compact and extensible FPGA hardware accelerator to demonstrate our findings. It constructs eviction sets at unprecedented speeds (<200 µs), outperforming existing techniques with one to three orders of magnitude. It maintains excellent performance, even under high noise pressure. We also use the accelerator to set up a covert channel with fine spatial granularity, encoding more than 3 bits per cache set. Furthermore, it can efficiently evict shared targets with tiny eviction sets, refuting the common assumption that eviction sets must be as large as the cache associativity.
The Dangers of Human Touch: Fingerprinting Browser Extensions through User Actions
Konstantinos Solomos, Panagiotis Ilia, and Soroush Karami, University of Illinois at Chicago; Nick Nikiforakis, Stony Brook University; Jason Polakis, University of Illinois at Chicago
Browser extension fingerprinting has garnered considerable attention recently due to the twofold privacy loss that it incurs. Apart from facilitating tracking by augmenting browser fingerprints, the list of installed extensions can be directly used to infer sensitive user characteristics. However, prior research was performed in a vacuum, overlooking a core dimension of extensions' functionality: how they react to user actions. In this paper, we present the first exploration of user-triggered extension fingerprinting. Guided by our findings from a large-scale static analysis of browser extensions we devise a series of user action templates that enable dynamic extension-exercising frameworks to comprehensively uncover hidden extension functionality that can only be triggered through user interactions. Our experimental evaluation demonstrates the effectiveness of our proposed technique, as we are able to fingerprint 4,971 unique extensions, 36% of which are not detectable by state-of-the-art techniques. To make matters worse, we find that ≈67% of the extensions that require mouse or keyboard interactions lack appropriate safeguards, rendering them vulnerable to pages that simulate user actions through JavaScript. To assist extension developers in protecting users from this privacy threat, we build a tool that automatically includes origin checks for fortifying extensions against invasive sites.
MundoFuzz: Hypervisor Fuzzing with Statistical Coverage Testing and Grammar Inference
Cheolwoo Myung, Gwangmu Lee, and Byoungyoung Lee, Seoul National University
A hypervisor is system software, managing and running virtual machines. Since the hypervisor is placed at the lowestlevel in the typical systems software stack, it has critical security implications. Once compromised, the entire software components running on top of the hypervisor (including all guest virtual machines and applications running within each guest virtual machine) are compromised as well, as the hypervisor has all the privileges to access those.
This paper proposes MUNDOFUZZ, a hypervisor fuzzer to enable both coverage-guided and grammar-aware fuzzing. We find that the coverage measurement in hypervisors suffers from noises due to the hypervisor's asynchronous system event handling. In order to filter out such noises, MUNDOFUZZ develops a statistical differential coverage measurement methods, allowing MUNDOFUZZ to capture the clean coverage information for hypervisor inputs. Moreover, we observe that hypervisor inputs have complex input grammars because it supports many different devices and each device has its own input format. Thus, MUNDOFUZZ learns the input grammar through inspecting the coverage characteristics of the given hypervisor input, which is based on the idea that the hypervisor behaves in a different way if the grammatically correct (or incorrect) input is given. We evaluated MUNDOFUZZ with popular hypervisors, QEMU and Bhyve, and MUNDOFUZZ outperformed other state-of-the-art hypervisor fuzzers ranging from 4.91% to 6.60% in terms of coverage. More importantly, MUNDOFUZZ identified 40 previously unknown bugs (including 9 CVEs), demonstrating its strong practical effectiveness in finding real-world hypervisor vulnerabilities.
Exploring the Security Boundary of Data Reconstruction via Neuron Exclusivity Analysis
Xudong Pan, Mi Zhang, Yifan Yan, Jiaming Zhu, and Min Yang, Fudan University
Among existing privacy attacks on the gradient of neural networks, data reconstruction attack, which reverse engineers the training batch from the gradient, poses a severe threat on the private training data. Despite its empirical success on large architectures and small training batches, unstable reconstruction accuracy is also observed when a smaller architecture or a larger batch is under attack. Due to the weak interpretability of existing learning-based attacks, there is little known on why, when and how data reconstruction attack is feasible.
In our work, we perform the first analytic study on the security boundary of data reconstruction from gradient via a microcosmic view on neural networks with rectified linear units (ReLUs), the most popular activation function in practice. For the first time, we characterize the insecure/secure boundary of data reconstruction attack in terms of the neuron exclusivity state of a training batch, indexed by the number of Exclusively Activated Neurons (ExANs, i.e., a ReLU activated by only one sample in a batch). Intuitively, we show a training batch with more ExANs are more vulnerable to data reconstruction attack and vice versa. On the one hand, we construct a novel deterministic attack algorithm which substantially outperforms previous attacks for reconstructing training batches lying in the insecure boundary of a neural network. Meanwhile, for training batches lying in the secure boundary, we prove the impossibility of unique reconstruction, based on which an exclusivity reduction strategy is devised to enlarge the secure boundary for mitigation purposes.
SARA: Secure Android Remote Authorization
Abdullah Imran, Habiba Farrukh, Muhammad Ibrahim, Z. Berkay Celik, and Antonio Bianchi, Purdue University
Modern smartphones are equipped with Trusted Execution Environments (TEEs), offering security features resilient even against attackers able to fully compromise the normal operating system (e.g., Linux in Android devices). The academic community, as well as the smartphone manufacturers, have proposed to use TEEs to strengthen the security of authorization protocols. However, the usage of these protocols has been hampered by both practicality issues and lack of completeness in terms of security.
To address these issues, in this paper, we design, implement, and evaluate SARA (Secure Android Remote Authorization),an Android library that uses the existing TEE-powered Android APIs to implement secure, end-to-end remote authorization for Android apps. SARA is practical in its design, as it makes use of Android APIs and TEE features that are already present in modern Android devices to implement a novel secure authorization protocol. In fact, SARA does not require any modifications to the Android operating system nor to the code running in TrustZone (the TEE powering existing Android devices). For this reason, it can be readily used in existing apps running on existing smartphones. Moreover, SARA is designed to ensure that even developers that have no experience in implementing security protocols can make use of it within their apps. At the same time, SARA is secure, since it allows implementing authorization protocols that are resilient even against attackers able to achieve root privileges on a compromised Android device.
We first evaluate SARA by conducting a user study to ascertain its usability. Then, we prove SARA's security features by formally verifying its security protocol using ProVerif.
Trust Dies in Darkness: Shedding Light on Samsung's TrustZone Keymaster Design
Alon Shakevsky, Eyal Ronen, and Avishai Wool, Tel-Aviv University
ARM-based Android smartphones rely on the TrustZone hardware support for a Trusted Execution Environment (TEE) to implement security-sensitive functions. The TEE runs a separate, isolated, TrustZone Operating System (TZOS), in parallel to Android. The implementation of the cryptographic functions within the TZOS is left to the device vendors, who create proprietary undocumented designs.
In this work, we expose the cryptographic design and implementation of Android's Hardware-Backed Keystore in Samsung's Galaxy S8, S9, S10, S20, and S21 flagship devices. We reversed-engineered and provide a detailed description of the cryptographic design and code structure, and we unveil severe design flaws. We present an IV reuse attack on AES-GCM that allows an attacker to extract hardware-protected key material, and a downgrade attack that makes even the latest Samsung devices vulnerable to the IV reuse attack. We demonstrate working key extraction attacks on the latest devices. We also show the implications of our attacks on two higher-level cryptographic protocols between the TrustZone and a remote server: we demonstrate a working FIDO2 WebAuthn login bypass and a compromise of Google's Secure Key Import.
We discuss multiple flaws in the design flow of TrustZone based protocols. Although our specific attacks only apply to the ≈100 million devices made by Samsung, it raises the much more general requirement for open and proven standards for critical cryptographic and security designs.
Counting in Regexes Considered Harmful: Exposing ReDoS Vulnerability of Nonbacktracking Matchers
Lenka Turoňová, Lukáš Holík, Ivan Homoliak, and Ondřej Lengál, Faculty of Information Technology, Brno University of Technology; Margus Veanes, Microsoft Research Redmond; Tomáš Vojnar, Faculty of Information Technology, Brno University of Technology
In this paper, we study the performance characteristics of nonbacktracking regex matchers and their vulnerability against ReDoS (regular expression denial of service) attacks. We focus on their known Achilles heel, which are extended regexes that use bounded quantifiers (e.g., '(ab){100}'
). We propose a method for generating input texts that can cause ReDoS attacks on these matchers. The method exploits the bounded repetition and uses it to force expensive simulations of the deterministic automaton for the regex. We perform an extensive experimental evaluation of our and other state-of-the-art ReDoS generators on a large set of practical regexes with a comprehensive set of backtracking and nonbacktracking matchers, as well as experiments where we demonstrate ReDoS attacks on state-of-the-art real-world security applications containing SNORT with Hyperscan and the HW-accelerated regex matching engine on the NVIDIA BlueField-2 card. Our experiments show that bounded repetition is indeed a notable weakness of nonbacktracking matchers, with our generator being the only one capable of significantly increasing their running time.
SCRAPS: Scalable Collective Remote Attestation for Pub-Sub IoT Networks with Untrusted Proxy Verifier
Lukas Petzi, Ala Eddine Ben Yahya, and Alexandra Dmitrienko, University of Würzburg; Gene Tsudik, UC Irvine; Thomas Prantl and Samuel Kounev, University of Würzburg
Remote Attestation (RA) is a basic security mechanism that detects malicious presence on various types of computing components, e.g., IoT devices. In a typical IoT setting, RA involves a trusted Verifier that sends a challenge to an untrusted remote Prover, which must in turn reply with a fresh and authentic evidence of being in a trustworthy state. However, most current RA schemes assume a central Verifier, which represents a single point of failure. This feature is problematic when mutually suspicious stakeholders are involved. Furthermore, scalability issues arise as the number of IoT devices (Provers) grows.
Although some RA schemes allow peer Provers to act as Verifiers, they involve unrealistic (for IoT devices) requirements, such as time synchronization and synchronous communication. Moreover, they incur heavy memory, computation, and communication burdens, while not considering sleeping or otherwise disconnected devices. Motivated by the need to address these limitations, we construct Scalable Collective Remote Attestation for Pub-Sub (SCRAPS), a novel collective RA scheme. It achieves scalability by outsourcing Verifier duties to a smart contract and mitigates DoS attacks against both Provers and Verifiers. It also removes the need for synchronous communication. Furthermore,RA evidence in SCRAPS is publicly verifiable, which significantly reduces the number of attestation evidence computations, thus lowering Prover burden. We report on a prototype of SCRAPS over Hyperledger Sawtooth (a blockchain geared for IoT use-cases) and evaluate its performance, scalability, and security aspects.
Poisoning Attacks to Local Differential Privacy Protocols for Key-Value Data
Yongji Wu, Xiaoyu Cao, Jinyuan Jia, and Neil Zhenqiang Gong, Duke University
Local Differential Privacy (LDP) protocols enable an untrusted server to perform privacy-preserving, federated data analytics. Various LDP protocols have been developed for different types of data such as categorical data, numerical data, and key-value data. Due to their distributed settings, LDP protocols are fundamentally vulnerable to poisoning attacks, in which fake users manipulate the server's analytics results via sending carefully crafted data to the server. However, existing poisoning attacks focused on LDP protocols for simple data types such as categorical and numerical data, leaving the security of LDP protocols for more advanced data types such as key-value data unexplored.
In this work, we aim to bridge the gap by introducing novel poisoning attacks to LDP protocols for key-value data. In such a LDP protocol, a server aims to simultaneously estimate the frequency and mean value of each key among some users, each of whom possesses a set of key-value pairs. Our poisoning attacks aim to simultaneously maximize the frequencies and mean values of some attacker-chosen target keys via sending carefully crafted data from some fake users to the sever. Specifically, since our attacks have two objectives, we formulate them as a two-objective optimization problem. Moreover, we propose a method to approximately solve the two-objective optimization problem, from which we obtain the optimal crafted data the fake users should send to the server. We demonstrate the effectiveness of our attacks to three LDP protocols for key-value data both theoretically and empirically. We also explore two defenses against our attacks, which are effective in some scenarios but have limited effectiveness in other scenarios. Our results highlight the needs for new defenses against our poisoning attacks.
Arbiter: Bridging the Static and Dynamic Divide in Vulnerability Discovery on Binary Programs
Jayakrishna Vadayath, Arizona State University; Moritz Eckert, EURECOM; Kyle Zeng, Arizona State University; Nicolaas Weideman, University of Southern California; Gokulkrishna Praveen Menon, Arizona State University; Yanick Fratantonio, Cisco Systems Inc.; Davide Balzarotti, EURECOM; Adam Doupé, Tiffany Bao, and Ruoyu Wang, Arizona State University; Christophe Hauser, University of Southern California; Yan Shoshitaishvili, Arizona State University
In spite of their effectiveness in the context of vulnerability discovery, current state-of-the-art binary program analysis approaches are limited by inherent trade-offs between accuracy and scalability. In this paper, we identify a set of vulnerability properties that can aid both static and dynamic vulnerability detection techniques, improving the precision of the former and the scalability of the latter. By carefully integrating static and dynamic techniques, we detect vulnerabilities that exhibit these properties in real-world programs at a large scale.
We implemented our technique, making several advancements in the analysis of binary code, and created a prototype called ARBITER. We demonstrate the effectiveness of our approach with a large-scale evaluation on four common vulnerability classes: CWE-131 (Incorrect Calculation of Buffer Size), CWE-252 (Unchecked Return Value), CWE-134 (Uncontrolled Format String), and CWE-337 (Predictable Seed in Pseudo-Random Number Generator). We evaluated our approach on more than 76,516 x86-64 binaries in the Ubuntu repositories and discovered new vulnerabilities, including a flaw inserted into programs during compilation.
Breaking Bridgefy, again: Adopting libsignal is not enough
Martin R. Albrecht, Information Security Group, Royal Holloway, University of London; Raphael Eikenberg and Kenneth G. Paterson, Applied Cryptography Group, ETH Zurich
Bridgefy is a messaging application that uses Bluetooth-based mesh networking. Its developers and others have advertised it for use in areas witnessing large-scale protests involving confrontations between protesters and state agents. In August 2020, a security analysis reported severe vulnerabilities that invalidated Bridgefy's claims of confidentiality, authentication, and resilience. In response, the developers adopted the Signal protocol and then continued to advertise their application as being suitable for use by higher-risk users.
In this work, we analyse the security of the revised Bridgefy messenger and SDK and invalidate its security claims. One attack (targeting the messenger) enables an adversary to compromise the confidentiality of private messages by exploiting a time-of-check to time-of-use (TOCTOU) issue, side-stepping Signal's guarantees. The other attack (targeting the SDK) allows an adversary to recover broadcast messages without knowing the network-wide shared encryption key.
We also found that the changes deployed in response to the August 2020 analysis failed to remedy the previously reported vulnerabilities. In particular, we show that (i) the protocol persisted to be susceptible to an active attacker-in-the-middle, (ii) an adversary continued to be able to impersonate other users in the broadcast channel of the Bridgefy messenger, (iii) the DoS attack using a decompression bomb was still applicable, albeit in a limited form, and that (iv) the privacy issues of Bridgefy remained largely unresolved.
"The Same PIN, Just Longer": On the (In)Security of Upgrading PINs from 4 to 6 Digits
Collins W. Munyendo, The George Washington University; Philipp Markert, Ruhr University Bochum; Alexandra Nisenoff, University of Chicago; Miles Grant and Elena Korkes, The George Washington University; Blase Ur, University of Chicago; Adam J. Aviv, The George Washington University
With the goal of improving security, companies like Apple have moved from requiring 4-digit PINs to 6-digit PINs in contexts like smartphone unlocking. Users with a 4-digit PIN thus must "upgrade" to a 6-digit PIN for the same device or account. In an online user study (n=1010), we explore the security of such upgrades. Participants used their own smartphone to first select a 4-digit PIN. They were then directed to select a 6-digit PIN with one of five randomly assigned justifications. In an online attack that guesses a small number of common PINs (10–30), we observe that 6-digit PINs are, at best, marginally more secure than 4-digit PINs. To understand the relationship between 4- and 6-digit PINs, we then model targeted attacks for PIN upgrades. We find that attackers who know a user's previous 4-digit PIN perform significantly better than those who do not at guessing their 6-digit PIN in only a few guesses using basic heuristics (e.g., appending digits to the 4-digit PIN). Participants who selected a 6-digit PIN when given a "device upgrade" justification selected 6-digit PINs that were the easiest to guess in a targeted attack, with the attacker successfully guessing over 25% of the PINs in just 10 attempts, and more than 30% in 30 attempts. Our results indicate that forcing users to upgrade to 6-digit PINs offers limited security improvements despite adding usability burdens. System designers should thus carefully consider this tradeoff before requiring upgrades.
Networks of Care: Tech Abuse Advocates' Digital Security Practices
Julia Slupska, University of Oxford; Angelika Strohmayer, Northumbria University
As technology becomes an enabler of relationship abuse and coercive control, advocates who support survivors develop digital security practices to counter this. Existing research on technology-related abuse has primarily focused on describing the dynamics of abuse and developing solutions for this problem; we extend this literature by focusing on the security practices of advocates working "on the ground", i.e. in domestic violence shelters and other support services. We present findings from 26 semi-structured interviews and a data walkthrough workshop in which advocates described how they support survivors. We identified a variety of intertwined emotional and technical support practices, including establishing trust, safety planning, empowerment, demystification, supporting evidence collection and making referrals. By building relationships with other services and stakeholders, advocates also develop networks of care throughout society to create more supportive environments for survivors. Using critical and feminist theories, we see advocates as sources of crucial technical expertise to reduce this kind of violence in the future. Security and privacy researchers can build on and develop these networks of care by employing participatory methods and expanding threat modelling to account for interpersonal harms like coercive control and structural forms of discrimination such as misogyny and racism.
Khaleesi: Breaker of Advertising and Tracking Request Chains
Umar Iqbal, University of Washington; Charlie Wolfe, University of Iowa; Charles Nguyen, University of California, Davis; Steven Englehardt, DuckDuckGo; Zubair Shafiq, University of California, Davis
Request chains are being used by advertisers and trackers for information sharing and circumventing recently introduced privacy protections in web browsers. There is little prior work on mitigating the increasing exploitation of request chains by advertisers and trackers. The state-of-the-art ad and tracker blocking approaches lack the necessary context to effectively detect advertising and tracking request chains. We propose Khaleesi, a machine learning approach that captures the essential sequential context needed to effectively detect advertising and tracking request chains. We show that Khaleesi achieves high accuracy, that holds well over time, is generally robust against evasion attempts, and outperforms existing approaches. We also show that Khaleesi is suitable for online deployment and it improves page load performance.
DeepPhish: Understanding User Trust Towards Artificially Generated Profiles in Online Social Networks
Jaron Mink, Licheng Luo, and Natã M. Barbosa, University of Illinois at Urbana-Champaign; Olivia Figueira, Santa Clara University; Yang Wang and Gang Wang, University of Illinois at Urbana-Champaign
Fabricated media from deep learning models, or deepfakes, have been recently applied to facilitate social engineering efforts by constructing a trusted social persona. While existing works are primarily focused on deepfake detection, little is done to understand how users perceive and interact with deepfake persona (e.g., profiles) in a social engineering context. In this paper, we conduct a user study (n=286) to quantitatively evaluate how deepfake artifacts affect the perceived trustworthiness of a social media profile and the profile's likelihood to connect with users. Our study investigates artifacts isolated within a single media field (images or text) as well as mismatched relations between multiple fields. We also evaluate whether user prompting (or training) benefits users in this process. We find that artifacts and prompting significantly decrease the trustworthiness and request acceptance of deepfake profiles. Even so, users still appear vulnerable with 43% of them connecting to a deepfake profile under the best-case conditions. Through qualitative data, we find numerous reasons why this task is challenging for users, such as the difficulty of distinguishing text artifacts from honest mistakes and the social pressures entailed in the connection decisions. We conclude by discussing the implications of our results for content moderators, social media platforms, and future defenses.
TLB;DR: Enhancing TLB-based Attacks with TLB Desynchronized Reverse Engineering
Andrei Tatar, Vrije Universiteit, Amsterdam; Daniël Trujillo, Vrije Universiteit, Amsterdam, and ETH Zurich; Cristiano Giuffrida and Herbert Bos, Vrije Universiteit, Amsterdam
Translation Lookaside Buffers, or TLBs, play a vital role in recent microarchitectural attacks. However, unlike CPU caches, we know very little about the exact operation of these essential microarchitectural components. In this paper, we introduce TLB desynchronization as a novel technique for reverse engineering TLB behavior from software. Unlike previous efforts that rely on timing or performance counters, our technique relies on fundamental properties of TLBs, enabling precise and fine-grained experiments. We use desynchronization to shed new light on TLB behavior, examining previously undocumented features such as replacement policies and handling of PCIDs on commodity Intel processors. We also show that such knowledge allows for more and better attacks.
Our results reveal a novel replacement policy on the L2 TLB of modern Intel CPUs as well as behavior indicative of a PCID cache. We use our new insights to design adversarial access patterns that massage the TLB state into evicting a target entry in the minimum number of steps, then examine their impact on several classes of prior TLB-based attacks. Our findings enable practical side channels à la TLBleed over L2, with much finer spatial discrimination and at a sampling rate comparable to L1, as well as an even finer-grained variant that targets both levels. We also show substantial speed gains for other classes of attacks that rely on TLB eviction.
Playing Without Paying: Detecting Vulnerable Payment Verification in Native Binaries of Unity Mobile Games
Chaoshun Zuo and Zhiqiang Lin, The Ohio State University
Modern mobile games often contain in-app purchasing (IAP) for players to purchase digital items such as virtual currency, equipment, or extra moves. In theory, IAP should have been implemented securely; but in practice, we have found that many game developers have failed to do so, particularly by misplacing the trust of payment verification, e.g., by either locally verifying the payment transactions or without using any verification at all, leading to playing without paying vulnerabilities. This paper presents PAYMENTSCOPE, a static binary analysis tool to automatically identify vulnerable IAP implementations in mobile games. Through modeling of its IAP protocols with the SDK provided APIs using a payment-aware data flow analysis, PAYMENTSCOPE directly pinpoints untrusted payment verification vulnerabilities in game native binaries. We have implemented PAYMENTSCOPE on top of binary analysis framework Ghidra, and tested with 39,121 Unity (the most popular game engine) mobile games, with which PAYMENTSCOPE has identified 8,954 (22.89%) vulnerable games. Among them, 8,233 games do not verify the validity of payment transactions and 721 games simply verify the transactions locally. We have disclosed the identified vulnerabilities to developers of vulnerable games, and many of them have acknowledged our findings.
Building an Open, Robust, and Stable Voting-Based Domain Top List
Qinge Xie, Georgia Institute of Technology; Shujun Tang, QI-ANXIN Technology Research Institute; Xiaofeng Zheng, QI-ANXIN Technology Research Institute and Tsinghua University; Qingran Lin, QI-ANXIN Technology Research Institute; Baojun Liu, Tsinghua University; Haixin Duan, QI-ANXIN Technology Research Institute and Tsinghua University; Frank Li, Georgia Institute of Technology
Domain top lists serve as critical resources for the Internet measurement, security, and privacy research communities. Hundreds of prior research studies have used these lists as a set of supposedly popular domains to investigate. However, existing top lists exhibit numerous issues, including a lack of transparency into the list data sources and construction methods, high volatility, and easy ranking manipulation. Despite these flaws, these top lists remain widely used today due to a lack of suitable alternatives.
In this paper, we systematically explore the construction of a domain top list from scratch. Using an extensive passive DNS dataset, we investigate different top list design considerations. As a product of our exploration, we produce a voting-based domain ranking method where we quantify the domain preferences of individual IP addresses, and then determine a global ranking across addresses through a voting mechanism. We empirically evaluate our top list design, demonstrating that it achieves better stability and manipulation resistance than existing top lists, while serving as an open and transparent ranking method that other researchers can use or adapt.
Many Roads Lead To Rome: How Packet Headers Influence DNS Censorship Measurement
Abhishek Bhaskar and Paul Pearce, Georgia Institute of Technology
Internet censorship is widespread, impacting citizens of hundreds of countries around the world. Recent work has developed techniques that can perform widespread, longitudinal measurements of global Internet manipulation remotely and have focused largely on the scale of censorship measurements with minimal focus on reproducibility and consistency.
In this work we explore the role packet headers (e.g., source IP address and source port) have on DNS censorship. By performing a large-scale measurement study building on the techniques deployed by previous and current censorship measurement platforms, we find that choice of ephemeral source port and local source IP address (e.g., x.x.x.7 vs x.x.x.8) influence routing, which in turn influences DNS censorship. We show that 37% of IPs across 56% ASes measured show some change in censorship behavior depending on source port and local source IP. This behavior is frequently all-or-nothing, where choice of header can result in no observable censorship. Such behavior mimics and could be misattributed to geolocation error, packet loss, or network outages. The scale of censorship differences can more than double depending on the lowest 3 bits of the source IP address, consistent with known router load balancing techniques. We also observe smaller-scale censorship variation where only a few domains experience censorship differences based on packet parameters. We lastly find that these variations are persistent; packet retries do not control for observed variation. Our results point to the need for methodological changes in future DNS censorship measurement, which we discuss.
Minefield: A Software-only Protection for SGX Enclaves against DVFS Attacks
Andreas Kogler and Daniel Gruss, Graz University of Technology; Michael Schwarz, CISPA Helmholtz Center for Information Security
Modern CPUs adapt clock frequencies and voltage levels to workloads to reduce energy consumption and heat dissipation. This mechanism, dynamic voltage and frequency scaling (DVFS), is controlled from privileged software but affects all execution modes, including SGX. Prior work showed that manipulating voltage or frequency can fault instructions and thereby subvert SGX enclaves. Consequently, Intel disabled the overclocking mailbox (OCM) required for software undervolting, also preventing benign use for energy saving.
In this paper, we propose Minefield, the first software-level defense against DVFS attacks. The idea of Minefield is not to prevent DVFS faults but to deflect faults to trap instructions and handle them before they lead to harmful behavior. As groundwork for Minefield, we systematically analyze DVFS attacks and observe a timing gap of at least 57.8 us between every OCM transition, leading to random faults over at least 57000 cycles. Minefield places highly fault-susceptible trap instructions in the victim code during compilation. Like redundancy countermeasures, Minefield is scalable and enables enclave developers to choose a security parameter between 0% and almost 100%, yielding a fine-grained security-performance trade-off. Our evaluation shows a density of 0.75, i.e., one trap after every 1-2 instruction, mitigates all known DVFS attacks in 99% on Intel SGX, incurring an overhead of 148.4% on protected enclaves. However, Minefield has no performance effect on the remaining system. Thus, Minefield is a better solution than hardware- or microcode-based patches disabling the OCM interface.
Attacks on Deidentification's Defenses
Aloni Cohen, University of Chicago
Distinguished Paper Award Winner
Quasi-identifier-based deidentification techniques (QI-deidentification) are widely used in practice, including k-anonymity, l-diversity, and t-closeness. We present three new attacks on QI-deidentification: two theoretical attacks and one practical attack on a real dataset. In contrast to prior work, our theoretical attacks work even if every attribute is a quasi-identifier. Hence, they apply to k-anonymity, l-diversity, t-closeness, and most other QI-deidentification techniques.
First, we introduce a new class of privacy attacks called downcoding attacks, and prove that every QI-deidentification scheme is vulnerable to downcoding attacks if it is minimal and hierarchical. Second, we convert the downcoding attacks into powerful predicate singling-out (PSO) attacks, which were recently proposed as a way to demonstrate that a privacy mechanism fails to legally anonymize under Europe's General Data Protection Regulation. Third, we use LinkedIn.com to reidentify 3 students in a k-anonymized dataset published by EdX (and show thousands are potentially vulnerable), undermining EdX's claimed compliance with the Family Educational Rights and Privacy Act.
The significance of this work is both scientific and political. Our theoretical attacks demonstrate that QI-deidentification may offer no protection even if every attribute is treated as a quasi-identifier. Our practical attack demonstrates that even deidentification experts acting in accordance with strict privacy regulations fail to prevent real-world reidentification. Together, they rebut a foundational tenet of QI-deidentification and challenge the actual arguments made to justify the continued use of k-anonymity and other QI-deidentification techniques.
In-Kernel Control-Flow Integrity on Commodity OSes using ARM Pointer Authentication
Sungbae Yoo, Jinbum Park, Seolheui Kim, and Yeji Kim, Samsung Research; Taesoo Kim, Samsung Research and Georgia Institute of Technology
This paper presents an in-kernel, hardware-based control-flow integrity (CFI) protection, called PAL, that utilizes ARM's Pointer Authentication (PA). It provides three important benefits over commercial, state-of-the-art PA-based CFIs like iOS's: 1) enhancing CFI precision via automated refinement techniques, 2) addressing hindsight problems of PA for inkernel uses such as preemptive hijacking and brute-forcing attacks, and 3) assuring the algorithmic or implementation correctness via post validation.
PAL achieves these goals in an OS-agnostic manner, so could be applied to commodity OSes like Linux and FreeBSD. The precision of the CFI protection can be adjusted for better performance or improved for better security with minimal engineering efforts. Our evaluation shows that PAL incurs negligible performance overhead: e.g., <1% overhead for Apache benchmark and 3–5% overhead for Linux perf benchmark on the latest Mac mini (M1). Our post-validation approach helps us ensure the security invariant required for the safe uses of PA inside the kernel, which also reveals new attack vectors on the iOS kernel. PAL as well as the CFI-protected kernels will be open sourced.
Unleash the Simulacrum: Shifting Browser Realities for Robust Extension-Fingerprinting Prevention
Soroush Karami, University of Illinois at Chicago; Faezeh Kalantari, Mehrnoosh Zaeifi, Xavier J. Maso, and Erik Trickel, Arizona State University; Panagiotis Ilia, University of Illinois at Chicago; Yan Shoshitaishvili and Adam Doupé, Arizona State University; Jason Polakis, University of Illinois at Chicago
Online tracking has garnered significant attention due to the privacy risk it poses to users. Among the various approaches, techniques that identify which extensions are installed in a browser can be used for fingerprinting browsers and tracking users, but also for inferring personal and sensitive user data. While preventing certain fingerprinting techniques is relatively simple, mitigating behavior-based extension-fingerprinting poses a significant challenge as it relies on hiding actions that stem from an extension's functionality. To that end, we introduce the concept of DOM Reality Shifting, whereby we split the reality users experience while browsing from the reality that webpages can observe. To demonstrate our approach we develop Simulacrum, a prototype extension that implements our defense through a targeted instrumentation of core Web API interfaces. Despite being conceptually straightforward, our implementation highlights the technical challenges posed by the complex and often idiosyncratic nature and behavior of web applications, modern browsers, and the JavaScript language. We experimentally evaluate our system against a state-of-theart DOM-based extension fingerprinting system and find that Simulacrum readily protects 95.37% of susceptible extensions. We then identify trivial modifications to extensions that enable our defense for the majority of the remaining extensions. To facilitate additional research and protect users from privacy-invasive behaviors we will open-source our system.
Anycast Agility: Network Playbooks to Fight DDoS
A S M Rizvi, USC/ISI; Leandro Bertholdo, University of Twente; João Ceron, SIDN Labs; John Heidemann, USC/ISI
IP anycast is used for services such as DNS and Content Delivery Networks (CDN) to provide the capacity to handle Distributed Denial-of-Service (DDoS) attacks. During a DDoS attack service operators redistribute traffic between anycast sites to take advantage of sites with unused or greater capacity. Depending on site traffic and attack size, operators may instead concentrate attackers in a few sites to preserve operation in others. Operators use these actions during attacks, but how to do so has not been described systematically or publicly. This paper describes several methods to use BGP to shift traffic when under DDoS, and shows that a response playbook can provide a menu of responses that are options during an attack. To choose an appropriate response from this playbook, we also describe a new method to estimate true attack size, even though the operator's view during the attack is incomplete. Finally, operator choices are constrained by distributed routing policies, and not all are helpful. We explore how specific anycast deployment can constrain options in this playbook, and are the first to measure how generally applicable they are across multiple anycast networks.
PolyCruise: A Cross-Language Dynamic Information Flow Analysis
Wen Li, Washington State University, Pullman; Jiang Ming, University of Texas at Arlington; Xiapu Luo, The Hong Kong Polytechnic University; Haipeng Cai, Washington State University, Pullman
Despite the fact that most real-world software systems today are written in multiple programming languages, existing program analysis based security techniques are still limited to single-language code. In consequence, security flaws (e.g., code vulnerabilities) at and across language boundaries are largely left out as blind spots. We present PolyCruise, a technique that enables holistic dynamic information flow analysis (DIFA) across heterogeneous languages hence security applications empowered by DIFA (e.g., vulnerability discovery) for multilingual software. PolyCruise combines a light language-specific analysis that computes symbolic dependencies in each language unit with a language-agnostic online data flow analysis guided by those dependencies, in a way that overcomes language heterogeneity. Extensive evaluation of its implementation for Python-C programs against micro, medium-sized, and large-scale benchmarks demonstrated PolyCruise's practical scalability and promising capabilities. It has enabled the discovery of 14 unknown cross-language security vulnerabilities in real-world multilingual systems such as NumPy, with 11 confirmed, 8 CVEs assigned, and 8 fixed so far. We also contributed the first benchmark suite for systematically assessing multilingual DIFA.
Communication-Efficient Triangle Counting under Local Differential Privacy
Jacob Imola, UC San Diego; Takao Murakami, AIST; Kamalika Chaudhuri, UC San Diego
Triangle counting in networks under LDP (Local Differential Privacy) is a fundamental task for analyzing connection patterns or calculating a clustering coefficient while strongly protecting sensitive friendships from a central server. In particular, a recent study proposes an algorithm for this task that uses two rounds of interaction between users and the server to significantly reduce estimation error. However, this algorithm suffers from a prohibitively high communication cost due to a large noisy graph each user needs to download.
In this work, we propose triangle counting algorithms under LDP with a small estimation error and communication cost. We first propose two-rounds algorithms consisting of edge sampling and carefully selecting edges each user downloads so that the estimation error is small. Then we propose a double clipping technique, which clips the number of edges and then the number of noisy triangles, to significantly reduce the sensitivity of each user's query. Through comprehensive evaluation, we show that our algorithms dramatically reduce the communication cost of the existing algorithm, e.g., from 6 hours to 8 seconds or less at a 20 Mbps download rate, while keeping a small estimation error.
Seeing the Forest for the Trees: Understanding Security Hazards in the 3GPP Ecosystem through Intelligent Analysis on Change Requests
Yi Chen and Di Tang, Indiana University Bloomington; Yepeng Yao, {CAS-KLONAT, BKLONSPT}, Institute of Information Engineering, CAS, and School of Cyber Security, University of Chinese Academy of Sciences; Mingming Zha and XiaoFeng Wang, Indiana University Bloomington; Xiaozhong Liu, Worcester Polytechnic Institute; Haixu Tang and Dongfang Zhao, Indiana University Bloomington
With the recent report of erroneous content in 3GPP specifications leading to real-world vulnerabilities, attention has been drawn to not only the specifications but also the way they are maintained and adopted by manufacturers and carriers. In this paper, we report the first study on this 3GPP ecosystem, for the purpose of understanding its security hazards. Our research leverages 414,488 Change Requests (CRs) that document the problems discovered from specifications and proposed changes, which provides valuable information about the security assurance of the 3GPP ecosystem.
Analyzing these CRs is impeded by the challenge in finding security-relevant CRs (SR-CRs), whose security connections cannot be easily established by even human experts. To identify them, we developed a novel NLP/ML pipeline that utilizes a small set of positively labeled CRs to recover 1,270 high-confidence SR-CRs. Our measurement on them reveals serious consequences of specification errors and their causes, including design errors and presentation issues, particularly the pervasiveness of inconsistent descriptions (misalignment) in security-relevant content. Also important is the discovery of a security weakness inherent to the 3GPP ecosystem, which publishes an SR-CR long before the specification has been fixed and related systems have been patched. This opens an "attack window", which can be as long as 11 years! Interestingly, we found that some recently reported vulnerabilities are actually related to the CRs published years ago. Further, we identified a set of vulnerabilities affecting major carriers and mobile phones that have not been addressed even today. With the trend of SR-CRs not showing any sign of abating, we propose measures to improve the security assurance of the ecosystem, including responsible handling of SR-CRs.
Hyperproofs: Aggregating and Maintaining Proofs in Vector Commitments
Shravan Srinivasan, University of Maryland; Alexander Chepurnoy, Ergo Platform; Charalampos Papamanthou, Yale University; Alin Tomescu, VMware Research; Yupeng Zhang, Texas A&M University
We present Hyperproofs, the first vector commitment (VC) scheme that is efficiently maintainable and aggregatable. Similar to Merkle proofs, our proofs form a tree that can be efficiently maintained: updating all n proofs in the tree after a single leaf change only requires O(logn) time. Importantly, unlike Merkle proofs, Hyperproofs are efficiently aggregatable, anywhere from 10× to 41× faster than SNARK-based aggregation of Merkle proofs. At the same time, an individual Hyperproof consists of only logn algebraic hashes (e.g., 32-byte elliptic curve points) and an aggregation of b such proofs is only O(log(blogn))-sized. Hyperproofs are also reasonably fast to update when compared to Merkle trees with SNARK-friendly hash functions.
As another benefit over Merkle trees, Hyperproofs are homomorphic: digests (and proofs) for two vectors can be homomorphically combined into a digest (and proofs) for their sum. Homomorphism is very useful in emerging applications such as stateless cryptocurrencies. First, it enables unstealability, a novel property that incentivizes proof computation. Second, it makes digests and proofs much more convenient to update.
Finally, Hyperproofs have certain limitations: they are not transparent, have linear-sized public parameters, are slower to verify, and have larger aggregated proofs and slower verification than SNARK-based approaches. Nonetheless, end-to-end, aggregation and verification in Hyperproofs is 10× to 41× faster than in SNARK-based Merkle trees.
Leaky Forms: A Study of Email and Password Exfiltration Before Form Submission
Asuman Senol, imec-COSIC, KU Leuven; Gunes Acar, Radboud University; Mathias Humbert, University of Lausanne; Frederik Zuiderveen Borgesius, Radboud University
Web users enter their email addresses into online forms for a variety of reasons, including signing in or signing up for a service or subscribing to a newsletter. While enabling such functionality, email addresses typed into forms can also be collected by third-party scripts even when users change their minds and leave the site without submitting the form. Email addresses—or identifiers derived from them—are known to be used by data brokers and advertisers for cross-site, cross-platform, and persistent identification of potentially unsuspecting individuals. In order to find out whether access to online forms is misused by online trackers, we present a measurement of email and password collection that occurs before the form submission on the top 100,000 websites. We evaluate the effect of user location, browser configuration, and interaction with consent dialogs by comparing results across two vantage points (EU/US), two browser configurations (desktop/mobile), and three consent modes. Our crawler finds and fills email and password fields, monitors the network traffic for leaks, and intercepts script access to filled input fields. Our analyses show that users' email addresses are exfiltrated to tracking, marketing and analytics domains before form submission and without giving consent on 1,844 websites in the EU crawl and 2,950 websites in the US crawl. While the majority of email addresses are sent to known tracking domains, we further identify 41 tracker domains that are not listed by any of the popular blocklists. Furthermore, we find incidental password collection on 52 websites by third-party session replay scripts.
Using Trātṛ to tame Adversarial Synchronization
Yuvraj Patel, Chenhao Ye, Akshat Sinha, Abigail Matthews, Andrea C. Arpaci-Dusseau, Remzi H. Arpaci-Dusseau, and Michael M. Swift, University of Wisconsin–Madison
We show that Linux containers are vulnerable to a new class of attacks – synchronization attacks – that exploit kernel synchronization to harm application performance, where an unprivileged attacker can control the duration of kernel critical sections to stall victims running in other containers on the same operating system. Furthermore, a subset of these attacks – framing attacks – persistently harm performance by expanding data structures even after the attacker quiesces. We demonstrate three such attacks on the Linux kernel involving the inode cache, the directory cache, and the futex table.
We design Trātṛ, a Linux kernel extension, to detect and mitigate synchronization and framing attacks with low overhead, prevent attacks from worsening, and recover by repairing data structures to their pre-attack state. Using microbenchmarks and real-world workloads, we show that Trātṛ can detect an attack within seconds and recover instantaneously, guaranteeing similar performance to baseline. Our experiments show that Trātṛ can detect simultaneous attacks and mitigate them with minimal overhead.
Security Analysis of Camera-LiDAR Fusion Against Black-Box Attacks on Autonomous Vehicles
R. Spencer Hallyburton and Yupei Liu, Duke University; Yulong Cao and Z. Morley Mao, University of Michigan; Miroslav Pajic, Duke University
To enable safe and reliable decision-making, autonomous vehicles (AVs) feed sensor data to perception algorithms to understand the environment. Sensor fusion with multi-frame tracking is becoming increasingly popular for detecting 3D objects. Thus, in this work, we perform an analysis of camera-LiDAR fusion, in the AV context, under LiDAR spoofing attacks. Recently, LiDAR-only perception was shown vulnerable to LiDAR spoofing attacks; however, we demonstrate these attacks are not capable of disrupting camera-LiDAR fusion. We then define a novel, context-aware attack: frustum attack, and show that out of 8 widely used perception algorithms – across 3 architectures of LiDAR-only and 3 architectures of camera-LiDAR fusion – all are significantly vulnerable to the frustum attack. In addition, we demonstrate that the frustum attack is stealthy to existing defenses against LiDAR spoofing as it preserves consistencies between camera and LiDAR semantics. Finally, we show that the frustum attack can be exercised consistently over time to form stealthy longitudinal attack sequences, compromising the tracking module and creating adverse outcomes on end-to-end AV control.
Automated Detection of Automated Traffic
Cormac Herley, Microsoft Research
We describe a method to separate abuse from legitimate traffic when we have categorical features and no labels are available. Our approach hinges on the observation that, if we could locate them, unattacked bins of a categorical feature x would allow us to estimate the benign distribution of any feature that is independent of x. We give an algorithm that finds these unattacked bins (if they exist) and show how to build an overall classifier that is suitable for very large data volumes and high levels of abuse. The approach is one-sided: our only significant assumptions about abuse are the existence of unattacked bins, and that distributions of abuse traffic do not precisely match those of benign.
We evaluate on two datasets: 3 million requests from a web-server dataset and a collection of 5.1 million Twitter accounts crawled using the public API. The results confirm that the approach is successful at identifying clusters of automated behaviors. On both problems we easily outperform unsupervised methods such as Isolation Forests, and have comparable performance to Botometer on the Twitter dataset.
Ghost Peak: Practical Distance Reduction Attacks Against HRP UWB Ranging
Patrick Leu and Giovanni Camurati, ETH Zurich; Alexander Heinrich, TU Darmstadt; Marc Roeschlin and Claudio Anliker, ETH Zurich; Matthias Hollick, TU Darmstadt; Srdjan Capkun, ETH Zurich; Jiska Classen, TU Darmstadt
We present the first over-the-air attack on IEEE 802.15.4z High-Rate Pulse Repetition Frequency (HRP) Ultra-Wide Band (UWB) distance measurement systems. Specifically, we demonstrate a practical distance reduction attack against pairs of Apple U1 chips (embedded in iPhones and AirTags), as well as against U1 chips inter-operating with NXP and Qorvo UWB chips. These chips have been deployed in a wide range of phones and cars to secure car entry and start and are projected for secure contactless payments, home locks, and contact tracing systems. Our attack operates without any knowledge of cryptographic material, results in distance reductions from 12m (actual distance) to 0m (spoofed distance) with attack success probabilities of up to 4%, and requires only an inexpensive (USD 65) off-the-shelf device. Access control can only tolerate sub-second latencies to not inconvenience the user, leaving little margin to perform time-consuming verifications. These distance reductions bring into question the use of UWB HRP in security-critical applications.
Transferring Adversarial Robustness Through Robust Representation Matching
Pratik Vaishnavi, Stony Brook University; Kevin Eykholt, IBM; Amir Rahmati, Stony Brook University
With the widespread use of machine learning, concerns over its security and reliability have become prevalent. As such, many have developed defenses to harden neural networks against adversarial examples, imperceptibly perturbed inputs that are reliably misclassified. Adversarial training in which adversarial examples are generated and used during training is one of the few known defenses able to reliably withstand such attacks against neural networks. However, adversarial training imposes a significant training overhead and scales poorly with model complexity and input dimension. In this paper, we propose Robust Representation Matching (RRM), a low-cost method to transfer the robustness of an adversarially trained model to a new model being trained for the same task irrespective of architectural differences. Inspired by student-teacher learning, our method introduces a novel training loss that encourages the student to learn the teacher's robust representations. Compared to prior works, RRM is superior with respect to both model performance and adversarial training time. On CIFAR-10, RRM trains a robust model ~1.8X faster than the state-of-the-art. Furthermore, RRM remains effective on higher-dimensional datasets. On Restricted-ImageNet, RRM trains a ResNet50 model ~18X faster than standard adversarial training.
Constant-weight PIR: Single-round Keyword PIR via Constant-weight Equality Operators
Rasoul Akhavan Mahdavi and Florian Kerschbaum, University of Waterloo
Equality operators are an essential building block in tasks over secure computation such as private information retrieval. In private information retrieval (PIR), a user queries a database such that the server does not learn which element is queried. In this work, we propose equality operators for constant-weight codewords. A constant-weight code is a collection of codewords that share the same Hamming weight. Constant-weight equality operators have a multiplicative depth that depends only on the Hamming weight of the code, not the bit-length of the elements. In our experiments, we show how these equality operators are up to 10 times faster than existing equality operators. Furthermore, we propose PIR using the constant-weight equality operator or constant-weight PIR, which is a PIR protocol using an approach previously deemed impractical. We show that for private retrieval of large, streaming data, constant-weight PIR has a smaller communication complexity and lower runtime compared to SEALPIR and MulPIR, respectively, which are two state-of-the-art solutions for PIR. Moreover, we show how constant-weight PIR can be extended to keyword PIR. In keyword PIR, the desired element is retrieved by a unique identifier pertaining to the sought item, e.g., the name of a file. Previous solutions to keyword PIR require one or multiple rounds of communication to reduce the problem to normal PIR. We show that constant-weight PIR is the first practical single-round solution to single-server keyword PIR.
Provably-Safe Multilingual Software Sandboxing using WebAssembly
Jay Bosamiya, Wen Shih Lim, and Bryan Parno, Carnegie Mellon University
Distinguished Paper Award Winner and Second Prize Winner (tie) of the 2022 Internet Defense Prize
Many applications, from the Web to smart contracts, need to safely execute untrusted code. We observe that WebAssembly (Wasm) is ideally positioned to support such applications, since it promises safety and performance, while serving as a compiler target for many high-level languages. However, Wasm's safety guarantees are only as strong as the implementation that enforces them. Hence, we explore two distinct approaches to producing provably sandboxed Wasm code. One draws on traditional formal methods to produce mathematical, machine-checked proofs of safety. The second carefully embeds Wasm semantics in safe Rust code such that the Rust compiler can emit safe executable code with good performance. Our implementation and evaluation of these two techniques indicate that leveraging Wasm gives us provably-safe multilingual sandboxing with performance comparable to standard, unsafe approaches.
ALASTOR: Reconstructing the Provenance of Serverless Intrusions
Pubali Datta, University of Illinois at Urbana-Champaign; Isaac Polinsky, North Carolina State University; Muhammad Adil Inam and Adam Bates, University of Illinois at Urbana-Champaign; William Enck, North Carolina State University
Serverless computing has freed developers from the burden of managing their own platform and infrastructure, allowing them to rapidly prototype and deploy applications. Despite its surging popularity, however, serverless raises a number of concerning security implications. Among them is the difficulty of investigating intrusions – by decomposing traditional applications into ephemeral re-entrant functions, serverless has enabled attackers to conceal their activities within legitimate workflows, and even prevent root cause analysis by abusing warm container reuse policies to break causal paths. Unfortunately, neither traditional approaches to system auditing nor commercial serverless security products provide the transparency needed to accurately track these novel threats.
In this work, we propose ALASTOR, a provenance-based auditing framework that enables precise tracing of suspicious events in serverless applications. ALASTOR records function activity at both system and application layers to capture a holistic picture of each function instances' behavior. It then aggregates provenance from different functions at a central repository within the serverless platform, stitching it together to produce a global data provenance graph of complex function workflows. ALASTOR is both function and language-agnostic, and can easily be integrated into existing serverless platforms with minimal modification. We implement ALASTOR for the OpenFaaS platform and evaluate its performance using the well-established Nordstrom Hello,Retail! application, discovering in the process that ALASTOR imposes manageable overheads (13.74%), in exchange for significantly improved forensic capabilities as compared to commercially-available monitoring tools. To our knowledge, ALASTOR is the first auditing framework specifically designed to satisfy the operational requirements of serverless platforms.
Seeing is Living? Rethinking the Security of Facial Liveness Verification in the Deepfake Era
Changjiang Li, Pennsylvania State University and Zhejiang University; Li Wang, Shandong University; Shouling Ji and Xuhong Zhang, Zhejiang University; Zhaohan Xi, Pennsylvania State University; Shanqing Guo, Shandong University; Ting Wang, Pennsylvania State University
Facial Liveness Verification (FLV) is widely used for identity authentication in many security-sensitive domains and offered as Platform-as-a-Service (PaaS) by leading cloud vendors. Yet, with the rapid advances in synthetic media techniques (e.g., deepfake), the security of FLV is facing unprecedented challenges, about which little is known thus far.
To bridge this gap, in this paper, we conduct the first systematic study on the security of FLV in real-world settings. Specifically, we present LiveBugger, a new deepfake-powered attack framework that enables customizable, automated security evaluation of FLV. Leveraging LiveBugger, we perform a comprehensive empirical assessment of representative FLV platforms, leading to a set of interesting findings. For instance, most FLV APIs do not use anti-deepfake detection; even for those with such defenses, their effectiveness is concerning (e.g., it may detect high-quality synthesized videos but fail to detect low-quality ones). We then conduct an in-depth analysis of the factors impacting the attack performance of LiveBugger: a) the bias (e.g., gender or race) in FLV can be exploited to select victims; b) adversarial training makes deepfake more effective to bypass FLV; c) the input quality has a varying influence on different deepfake techniques to bypass FLV. Based on these findings, we propose a customized, two-stage approach that can boost the attack success rate by up to 70%. Further, we run proof-of-concept attacks on several representative applications of FLV (i.e., the clients of FLV APIs) to illustrate the practical implications: due to the vulnerability of the APIs, many downstream applications are vulnerable to deepfake. Finally, we discuss potential countermeasures to improve the security of FLV. Our findings have been confirmed by the corresponding vendors.
On the Necessity of Auditable Algorithmic Definitions for Machine Unlearning
Anvith Thudi, Hengrui Jia, Ilia Shumailov, and Nicolas Papernot, University of Toronto and Vector Institute
Machine unlearning, i.e. having a model forget about some of its training data, has become increasingly more important as privacy legislation promotes variants of the right-to-be-forgotten. In the context of deep learning, approaches for machine unlearning are broadly categorized into two classes: exact unlearning methods, where an entity has formally removed the data point's impact on the model by retraining the model from scratch, and approximate unlearning, where an entity approximates the model parameters one would obtain by exact unlearning to save on compute costs. In this paper, we first show that the definition that underlies approximate unlearning, which seeks to prove the approximately unlearned model is close to an exactly retrained model, is incorrect because one can obtain the same model using different datasets. Thus one could unlearn without modifying the model at all. We then turn to exact unlearning approaches and ask how to verify their claims of unlearning. Our results show that even for a given training trajectory one cannot formally prove the absence of certain data points used during training. We thus conclude that unlearning is only well-defined at the algorithmic level, where an entity's only possible auditable claim to unlearning is that they used a particular algorithm designed to allow for external scrutiny during an audit.
Might I Get Pwned: A Second Generation Compromised Credential Checking Service
Bijeeta Pal, Cornell University; Mazharul Islam, University of Wisconsin–Madison; Marina Sanusi Bohuk, Cornell University; Nick Sullivan, Luke Valenta, Tara Whalen, and Christopher Wood, Cloudflare; Thomas Ristenpart, Cornell Tech; Rahul Chatterjee, University of Wisconsin–Madison
Credential stuffing attacks use stolen passwords to log into victim accounts. To defend against these attacks, recently deployed compromised credential checking (C3) services provide APIs that help users and companies check whether a username, password pair is exposed. These services however only check if the exact password is leaked, and therefore do not mitigate credential tweaking attacks — attempts to compromise a user account with variants of a user's leaked passwords. Recent work has shown credential tweaking attacks can compromise accounts quite effectively even when the credential stuffing countermeasures are in place.
We initiate work on C3 services that protect users from credential tweaking attacks. The core underlying challenge is how to identify passwords that are similar to their leaked passwords while preserving honest clients' privacy and also preventing malicious clients from extracting breach data from the service. We formalize the problem and explore ways to measure password similarity that balance efficacy, performance, and security. Based on this study, we design "Might I Get Pwned" (MIGP), a new kind of breach alerting service. Our simulations show that MIGP reduces the efficacy of state-of-the-art 1000-guess credential tweaking attacks by 94%. MIGP preserves user privacy and limits potential exposure of sensitive breach entries. We show that the protocol is fast, with response time close to existing C3 services. We worked with Cloudflare to deploy MIGP in practice.
Mitigating Membership Inference Attacks by Self-Distillation Through a Novel Ensemble Architecture
Xinyu Tang, Saeed Mahloujifar, and Liwei Song, Princeton University; Virat Shejwalkar, Milad Nasr, and Amir Houmansadr, University of Massachusetts Amherst; Prateek Mittal, Princeton University
Membership inference attacks are a key measure to evaluate privacy leakage in machine learning (ML) models. It is important to train ML models that have high membership privacy while largely preserving their utility. In this work, we propose a new framework to train privacy-preserving models that induce similar behavior on member and non-member inputs to mitigate membership inference attacks. Our framework, called SELENA, has two major components. The first component and the core of our defense is a novel ensemble architecture for training. This architecture, which we call Split-AI, splits the training data into random subsets, and trains a model on each subset of the data. We use an adaptive inference strategy at test time: our ensemble architecture aggregates the outputs of only those models that did not contain the input sample in their training data. Our second component, Self-Distillation, (self-)distills the training dataset through our Split-AI ensemble, without using any external public datasets. We prove that our Split-AI architecture defends against a family of membership inference attacks, however, our defense does not provide provable guarantees against all possible attackers as opposed to differential privacy. This enables us to improve the utility of models compared to DP. Through extensive experiments on major benchmark datasets we show that SELENA presents a superior trade-off between (empirical) membership privacy and utility compared to the state of the art empirical privacy defenses. In particular, SELENA incurs no more than 3.9% drop in classification accuracy compared to the undefended model while reducing the membership inference attack advantage by a factor of up to 3.7 compared to MemGuard and a factor of up to 2.1 compared to adversarial regularization.
OS-Aware Vulnerability Prioritization via Differential Severity Analysis
Qiushi Wu, University of Minnesota; Yue Xiao and Xiaojing Liao, Indiana University Bloomington; Kangjie Lu, University of Minnesota
The Linux kernel is quickly evolving and extensively customized. This results in thousands of versions and derivatives. Unfortunately, the Linux kernel is quite vulnerable. Each year, thousands of bugs are reported, and hundreds of them are security-related bugs. Given the limited resources, the kernel maintainers have to prioritize patching the more severe vulnerabilities. In practice, Common Vulnerability Scoring System (CVSS)[1] has become the standard for characterizing vulnerability severity. However, a fundamental problem exists when CVSS meets Linux—it is used in a "one for all" manner. The severity of a Linux vulnerability is assessed for only the mainstream Linux, and all affected versions and derivatives will simply honor and reuse the CVSS score. Such an undistinguished CVSS usage results in underestimation or overestimation of severity, which further results in delayed and ignored patching or wastes of the precious resources. In this paper, we propose OS-aware vulnerability prioritization (namely DIFFCVSS), which employs differential severity analysis for vulnerabilities. Specifically, given a severity assessed vulnerability, as well as the mainstream version and a target version of Linux, DIFFCVSS employs multiple new techniques based on static program analysis and natural language processing to differentially identify whether the vulnerability manifests a higher or lower severity in the target version. A unique strength of this approach is that it transforms the challenging and laborious CVSS calculation into automatable differential analysis. We implement DIFFCVSS and apply it to the mainstream Linux and downstream Android systems. The evaluation and user-study results show that DIFFCVSS is able to precisely perform the differential severity analysis, and offers a precise and effective way to identify vulnerabilities that deserve a severity reevaluation.
Efficient Representation of Numerical Optimization Problems for SNARKs
Sebastian Angel, University of Pennsylvania and Microsoft Research; Andrew J. Blumberg, Columbia University; Eleftherios Ioannidis and Jess Woods, University of Pennsylvania
This paper introduces Otti, a general-purpose compiler for (zk)SNARKs that provides support for numerical optimization problems. Otti produces efficient arithmetizations of programs that contain optimization problems including linear programming (LP), semi-definite programming (SDP), and a broad class of stochastic gradient descent (SGD) instances. Numerical optimization is a fundamental algorithmic building block: applications include scheduling and resource allocation tasks, approximations to NP-hard problems, and training of neural networks. Otti takes as input arbitrary programs written in a subset of C that contain optimization problems specified via an easy-to-use API. Otti then automatically produces rank-1 constraint satisfiability (R1CS) instances that express a succinct transformation of those programs. Correct execution of the transformed program implies the optimality of the solution to the original optimization problem. Our evaluation on real benchmarks shows that Otti, instantiated with the Spartan proof system, can prove the optimality of solutions in zero-knowledge in as little as 100 ms—over 4 orders of magnitude faster than existing approaches.
Experimenting with Collaborative zk-SNARKs: Zero-Knowledge Proofs for Distributed Secrets
Alex Ozdemir and Dan Boneh, Stanford University
A zk-SNARK is a powerful cryptographic primitive that provides a succinct and efficiently checkable argument that the prover has a witness to a public NP statement, without revealing the witness. However, in their native form, zk-SNARKs only apply to a secret witness held by a single party. In practice, a collection of parties often need to prove a statement where the secret witness is distributed or shared among them.
We implement and experiment with collaborative zkSNARKs: proofs over the secrets of multiple, mutually distrusting parties. We construct these by lifting conventional zk-SNARKs into secure protocols among N provers to jointly produce a single proof over the distributed witness. We optimize the proof generation algorithm in pairing-based zkSNARKs so that algebraic techniques for multiparty computation (MPC) yield efficient proof generation protocols. For some zk-SNARKs, optimization is more challenging. This suggests MPC "friendliness" as an additional criterion for evaluating zk-SNARKs.
We implement three collaborative proofs and evaluate the concrete cost of proof generation. We find that over a 3Gb/s link, security against a malicious minority of provers can be achieved with approximately the same runtime as a single prover. Security against N −1 malicious provers requires only a 2× slowdown. This efficiency is unusual since most computations slow down by orders of magnitude when securely distributed. This efficiency means that most applications that can tolerate the cost of a single-prover proof should also be able to tolerate the cost of a collaborative proof.
Membership Inference Attacks and Defenses in Neural Network Pruning
Xiaoyong Yuan and Lan Zhang, Michigan Technological Unviersity
Neural network pruning has been an essential technique to reduce the computation and memory requirements for using deep neural networks for resource-constrained devices. Most existing research focuses primarily on balancing the sparsity and accuracy of a pruned neural network by strategically removing insignificant parameters and retraining the pruned model. Such efforts on reusing training samples pose serious privacy risks due to increased memorization, which, however, has not been investigated yet.
In this paper, we conduct the first analysis of privacy risks in neural network pruning. Specifically, we investigate the impacts of neural network pruning on training data privacy, i.e., membership inference attacks. We first explore the impact of neural network pruning on prediction divergence, where the pruning process disproportionately affects the pruned model's behavior for members and non-members. Meanwhile, the influence of divergence even varies among different classes in a fine-grained manner. Enlightened by such divergence, we proposed a self-attention membership inference attack against the pruned neural networks. Extensive experiments are conducted to rigorously evaluate the privacy impacts of different pruning approaches, sparsity levels, and adversary knowledge. The proposed attack shows the higher attack performance on the pruned models when compared with eight existing membership inference attacks. In addition, we propose a new defense mechanism to protect the pruning process by mitigating the prediction divergence based on KL-divergence distance, whose effectiveness has been experimentally demonstrated to effectively mitigate the privacy risks while maintaining the sparsity and accuracy of the pruned models.
Efficient Differentially Private Secure Aggregation for Federated Learning via Hardness of Learning with Errors
Timothy Stevens, Christian Skalka, and Christelle Vincent, University of Vermont; John Ring, MassMutual; Samuel Clark, Raytheon; Joseph Near, University of Vermont
Federated machine learning leverages edge computing to develop models from network user data, but privacy in federated learning remains a major challenge. Techniques using differential privacy have been proposed to address this, but bring their own challenges. Many techniques require a trusted third party or else add too much noise to produce useful models. Recent advances in secure aggregation using multiparty computation eliminate the need for a third party, but are computationally expensive especially at scale. We present a new federated learning protocol that leverages a novel differentially private, malicious secure aggregation protocol based on techniques from Learning With Errors. Our protocol outperforms current state-of-the art techniques, and empirical results show that it scales to a large number of parties, with optimal accuracy for any differentially private federated learning scheme.
OpenVPN is Open to VPN Fingerprinting
Diwen Xue, Reethika Ramesh, and Arham Jain, University of Michigan; Michalis Kallitsis, Merit Network, Inc.; J. Alex Halderman, University of Michigan; Jedidiah R. Crandall, Arizona State University/Breakpointing Bad; Roya Ensafi, University of Michigan
Distinguished Paper Award Winner and First Prize Winner of the 2022 Internet Defense Prize
VPN adoption has seen steady growth over the past decade due to increased public awareness of privacy and surveillance threats. In response, certain governments are attempting to restrict VPN access by identifying connections using "dual use" DPI technology. To investigate the potential for VPN blocking, we develop mechanisms for accurately fingerprinting connections using OpenVPN, the most popular protocol for commercial VPN services. We identify three fingerprints based on protocol features such as byte pattern, packet size, and server response. Playing the role of an attacker who controls the network, we design a two-phase framework that performs passive fingerprinting and active probing in sequence. We evaluate our framework in partnership with a million-user ISP and find that we identify over 85% of OpenVPN flows with only negligible false positives, suggesting that OpenVPN-based services can be effectively blocked with little collateral damage. Although some commercial VPNs implement countermeasures to avoid detection, our framework successfully identified connections to 34 out of 41 "obfuscated" VPN configurations. We discuss the implications of the VPN fingerprintability for different threat models and propose short-term defenses. In the longer term, we urge commercial VPN providers to be more transparent about their obfuscation approaches and to adopt more principled detection countermeasures, such as those developed in censorship circumvention research.
Backporting Security Patches of Web Applications: A Prototype Design and Implementation on Injection Vulnerability Patches
Youkun Shi, Yuan Zhang, Tianhan Luo, and Xiangyu Mao, Fudan University; Yinzhi Cao, Johns Hopkins University; Ziwen Wang, Yudi Zhao, Zongan Huang, and Min Yang, Fudan University
Web vulnerabilities, especially injection-related ones, are popular among web application frameworks (such as Word-Press and Piwigo), which can lead to severe consequences like user information leak and server-side malware execution. One major practice of fixing web vulnerabilities on real-world websites is to apply security patches from the official developers of web frameworks. However, such a practice is challenging because security patches are developed for the latest version of a web framework, but real-world websites often run an old version due to legacy reasons. A direct application of security patches on the old version often fails because web frameworks, especially the code around the vulnerable location, may change between versions.
In this paper, we design a security patch backporting framework and implement a prototype on injection vulnerability patches, called SKYPORT. SKYPORT first identifies safely-backportable patches of injection vulnerabilities and web framework versions in theory and then backports patches to corresponding old versions. In the evaluation, SKYPORT identifies 98 out of 155 security patches targeting legacy injection vulnerabilities, which can be backported to 750 old versions of web application frameworks. Then, SKYPORT successfully backported all of the aforementioned backportable patches to corresponding old versions to correctly fix vulnerabilities. We believe that this is a first-step towards this important research problem and hope our research can draw further attention from the research community in backporting security patches to fix unpatched vulnerabilities in general beyond injection-related ones.
MaDIoT 2.0: Modern High-Wattage IoT Botnet Attacks and Defenses
Tohid Shekari, Georgia Institute of Technology; Alvaro A. Cardenas, University of California, Santa Cruz; Raheem Beyah, Georgia Institute of Technology
The widespread availability of vulnerable IoT devices has resulted in IoT botnets. A particularly concerning IoT botnet can be built around high-wattage IoT devices such as EV chargers because, in large numbers, they can abruptly change the electricity consumption in the power grid. These attacks are called Manipulation of Demand via IoT (MaDIoT) attacks. Previous research has shown that the existing power grid protection mechanisms prevent any large-scale negative consequences to the grid from MaDIoT attacks. In this paper, we analyze this assumption and show that an intelligent attacker with extra knowledge about the power grid and its state, can launch more sophisticated attacks. Rather than attacking all locations at random times, our adversary uses an instability metric that lets the attacker know the specific time and geographical location to activate the high-wattage bots. We call these new attacks MaDIoT 2.0.
Physical-Layer Attacks Against Pulse Width Modulation-Controlled Actuators
Gökçen Yılmaz Dayanıklı, Qualcomm; Sourav Sinha, Virginia Tech; Devaprakash Muniraj, IIT Madras; Ryan M. Gerdes and Mazen Farhood, Virginia Tech; Mani Mina, Iowa State University
Cyber-physical systems (CPS) consist of integrated computational and physical components. The dynamics of physical components (e.g., a robot arm) are controlled by actuators via actuation signals. In this work, we analyze the extent to which intentional electromagnetic interference (IEMI) allows an attacker to alter the actuation signal to jam or control a class of widely used actuators: those that use pulse width modulation (PWM) to encode actuation data (e.g., rotation angle or speed). A theory of False Actuation Injection (FAI) is developed and experimentally validated with IEMI waveforms of certain frequencies and modulations.
Specifically, three attack waveforms, denoted as Block, Block & Rotate, and Full Control, are described that can be utilized by an attacker to block (denial of service) or alter the actuation data encoded in the PWM signal sent by an actuator's legitimate controller. The efficacy of the attack waveforms is evaluated against several PWM-controlled actuators, and it is observed that an attacker can implement denial-of-service attacks on all the tested actuators with Block waveform. Additionally, attackers can take control of servo motors from specific manufacturers (Futaba and HiTec) with reported Block & Rotate, and Full Control waveforms. A coupling model between the attack apparatus and victim PWM-based control system is presented to show that the attacker can utilize magnetic, resonant coupling to mount attacks at an appreciable distance. Indoor and in-flight attacks are demonstrated on the actuators of an unmanned aerial vehicle (UAV), the effects of which are shown to seriously impact the safe operation of said UAV, e.g., change in the flight trajectory. Additionally, the denial of service attacks are demonstrated on other actuators such as DC motors, the rotational speed of which is controlled with PWM, and possible countermeasures (such as optical actuation data transmission) are discussed.
Who Are You (I Really Wanna Know)? Detecting Audio DeepFakes Through Vocal Tract Reconstruction
Logan Blue, Kevin Warren, Hadi Abdullah, Cassidy Gibson, Luis Vargas, Jessica O'Dell, Kevin Butler, and Patrick Traynor, University of Florida
Generative machine learning models have made convincing voice synthesis a reality. While such tools can be extremely useful in applications where people consent to their voices being cloned (e.g., patients losing the ability to speak, actors not wanting to have to redo dialog, etc), they also allow for the creation of nonconsensual content known as deepfakes. This malicious audio is problematic not only because it can convincingly be used to impersonate arbitrary users, but because detecting deepfakes is challenging and generally requires knowledge of the specific deepfake generator. In this paper, we develop a new mechanism for detecting audio deepfakes using techniques from the field of articulatory phonetics. Specifically, we apply fluid dynamics to estimate the arrangement of the human vocal tract during speech generation and show that deepfakes often model impossible or highly-unlikely anatomical arrangements. When parameterized to achieve 99.9% precision, our detection mechanism achieves a recall of 99.5%, correctly identifying all but one deepfake sample in our dataset. We then discuss the limitations of this approach, and how deepfake models fail to reproduce all aspects of speech equally. In so doing, we demonstrate that subtle, but biologically constrained aspects of how humans generate speech are not captured by current models, and can therefore act as a powerful tool to detect audio deepfakes.
Shuffle-based Private Set Union: Faster and More Secure
Yanxue Jia and Shi-Feng Sun, Shanghai Jiao Tong University; Hong-Sheng Zhou, Virginia Commonwealth University; Jiajun Du and Dawu Gu, Shanghai Jiao Tong University
Private Set Union (PSU) allows two players, the sender and the receiver, to compute the union of their input datasets without revealing any more information than the result. While it has found numerous applications in practice, not much research has been carried out so far, especially for large datasets.
In this work, we take shuffling technique as a key to design PSU protocols for the first time. By shuffling receiver's set, we put forward the first protocol, denoted as ΠR PSU, that eliminates the expensive operations in previous works, such as additive homomorphic encryption and repeated operations on the receiver's set. It outperforms the state-of-the-art design by Kolesnikov et al. (ASIACRYPT 2019) in both efficiency and security; the unnecessary leakage in Kolesnikov et al.'s design, can be avoided in our design.
We further extend our investigation to the application scenarios in which both players may hold unbalanced input datasets. We propose our second protocol ΠS PSU, by shuffling the sender's dataset. This design can be viewed as a dual version of our first protocol, and it is suitable in the cases where the sender's input size is much smaller than the receiver's.
Finally, we implement our protocols ΠR PSU and ΠS PSU in C++ on big datasets, and perform a comprehensive evaluation in terms of both scalability and parallelizability. The results demonstrate that our design can obtain a 4-5× improvement over the state-of-the-art by Kolesnikov et al. with a single thread in WAN/LAN settings.
Pacer: Comprehensive Network Side-Channel Mitigation in the Cloud
Aastha Mehta, University of British Columbia (UBC); Mohamed Alzayat, Roberta De Viti, Björn B. Brandenburg, Peter Druschel, and Deepak Garg, Max Planck Institute for Software Systems (MPI-SWS)
Network side channels (NSCs) leak secrets through packet timing and packet sizes. They are of particular concern in public IaaS Clouds, where any tenant may be able to colocate and indirectly observe a victim's traffic shape. We present Pacer, the first system that eliminates NSC leaks in public IaaS Clouds end-to-end. It builds on the principled technique of shaping guest traffic outside the guest to make the traffic shape independent of secrets by design. However, Pacer also addresses important concerns that have not been considered in prior work—it prevents internal side-channel leaks from affecting reshaped traffic, and it respects network flow control, congestion control and loss recovery signals. Pacer is implemented as a paravirtualizing extension to the host hypervisor, requiring modest changes to the hypervisor and the guest kernel and optional, minimal changes to applications. We present Pacer's key abstraction of a cloaked tunnel, describe its design and implementation, and show through an experimental evaluation that Pacer imposes moderate overheads on bandwidth, client latency, and server throughput, while thwarting attacks using state-of-the-art CNN classifiers.
Zero-Knowledge Middleboxes
Paul Grubbs, Arasu Arun, Ye Zhang, Joseph Bonneau, and Michael Walfish, NYU
This paper initiates research on zero-knowledge middleboxes (ZKMBs). A ZKMB is a network middlebox that enforces network usage policies on encrypted traffic. Clients send the middlebox zero-knowledge proofs that their traffic is policy-compliant; these proofs reveal nothing about the client's communication except that it complies with the policy. We show how to make ZKMBs work with unmodified encrypted-communication protocols (specifically TLS 1.3), making ZKMBs invisible to servers. As a contribution of independent interest, we design optimized zero-knowledge proofs for TLS 1.3 session keys.
We apply the ZKMB paradigm to several case studies. Experimental results suggest that in certain settings, performance is in striking distance of practicality; an example is a middlebox that filters domain queries (each query requiring a separate proof) when the client has a long-lived TLS connection with a DNS resolver. In such configurations, the middlebox's overhead is 2–5 ms of running time per proof, and client latency to create a proof is several seconds. On the other hand, clients may have to store hundreds of MBs depending on the underlying zero-knowledge proof machinery, and for some applications, latency is tens of seconds.
TheHuzz: Instruction Fuzzing of Processors Using Golden-Reference Models for Finding Software-Exploitable Vulnerabilities
Rahul Kande, Addison Crump, and Garrett Persyn, Texas A&M University; Patrick Jauernig and Ahmad-Reza Sadeghi, Technische Universität Darmstadt; Aakash Tyagi and Jeyavijayan Rajendran, Texas A&M University
The increasing complexity of modern processors poses many challenges to existing hardware verification tools and methodologies for detecting security-critical bugs. Recent attacks on processors have shown the fatal consequences of uncovering and exploiting hardware vulnerabilities.
Fuzzing has emerged as a promising technique for detecting software vulnerabilities. Recently, a few hardware fuzzing techniques have been proposed. However, they suffer from several limitations, including non-applicability to commonly used hardware description languages (HDLs) like Verilog and VHDL, the need for significant human intervention, and inability to capture many intrinsic hardware behaviors, such as signal transitions and floating wires.
In this paper, we present the design and implementation of a novel hardware fuzzer, TheHuzz, that overcomes the aforementioned limitations and significantly improves the state of the art. We analyze the intrinsic behaviors of hardware designs in HDLs and then measure the coverage metrics that model such behaviors. TheHuzz generates assembly-level instructions to increase the desired coverage values, thereby finding many hardware bugs that are exploitable from software. We evaluate TheHuzz on four popular open-source processors and achieve 1.98× and 3.33× the speed compared to the industry-standard random regression approach and the state-of-the-art hardware fuzzer, DifuzzRTL, respectively. Using TheHuzz, we detected 11 bugs in these processors, including 8 new bugs, and we demonstrate exploits using the detected bugs. We also show that TheHuzz overcomes the limitations of formal verification tools from the semiconductor industry by comparing its findings to those discovered by the Cadence JasperGold tool.
Private Signaling
Varun Madathil and Alessandra Scafuro, North Carolina State University; István András Seres, Eötvös Loránd University; Omer Shlomovits and Denis Varlakov, ZenGo X
Distinguished Paper Award Winner
We introduce the problem of private signaling. In this problem, a sender posts a message on a certain location of a public bulletin board, and then posts a signal that allows only the intended recipient (and no one else) to learn that it is the recipient of the message posted at that location. Besides privacy, two efficiency requirements must be met. First, the sender and recipient do not participate in any out-of-band communication. Second, the overhead of the recipient must be (much) better than scanning the entire board.
Existing techniques, such as server-aided fuzzy message detection (Beck et al., CCS'21), could be employed to solve the private signaling problem. However, this solution leads to a trade-off between privacy and efficiency, where the complexity of the recipient grows with the required privacy. Specifically, this would require a scan of the entire board to obtain full privacy for the recipient.
In this work, we present a server-aided solution to the private signaling problem that guarantees full privacy for all recipients while requiring only constant amount of work for both the recipient and the sender.
Specifically, we provide three contributions: First, we provide a formal definition of private signaling in the Universal Composability (UC) framework and show that it captures several real-world settings where recipient anonymity is desired. Second, we present two server-aided protocols that UC-realize our definitions: one using a single server equipped with a trusted execution environment, and one based on two servers that employ garbled circuits. Third, we provide an open-source implementation of both of our protocols, evaluate their performance, and identify for which sets of parameters they can be practical.
Branch History Injection: On the Effectiveness of Hardware Mitigations Against Cross-Privilege Spectre-v2 Attacks
Enrico Barberis, Pietro Frigo, Marius Muench, Herbert Bos, and Cristiano Giuffrida, Vrije Universiteit Amsterdam
Branch Target Injection (BTI or Spectre v2) is one of the most dangerous transient execution vulnerabilities, as it allows an attacker to abuse indirect branch mispredictions to leak sensitive information. Unfortunately, it also has proven difficult to mitigate, with vendors originally resorting to inefficient software mitigations like retpoline. Recently, efficient hardware mitigations such as Intel eIBRS and Arm CSV2 have been deployed as a replacement in production, isolating the branch target state across privilege domains. The assumption is that this is sufficient to deter practical BTI exploitation. In this paper, we challenge this belief and disclose fundamental design flaws in both Intel and Arm solutions.
We introduce Branch History Injection (BHI or Spectre-BHB), a new primitive to build cross-privilege BTI attacks on systems deploying isolation-based hardware defenses. BHI builds on the observation that, while the branch target state is now isolated across privilege domains, such isolation is not extended to other branch predictor elements tracking the branch history state—ultimately re-enabling cross-privilege attacks. We further analyze the guarantees of a hypothetical isolation-based mitigation which also isolates the branch history and show that, barring a collision-free design, practical same-predictor-mode attacks are still possible. To instantiate our approach, we present end-to-end exploits leaking kernel memory from userland on Intel systems at 160 bytes/s, in spite of existing or hypothetical isolation-based mitigations. We conclude software defenses such as retpoline remain the only practical BTI mitigations in the foreseeable future and the pursuit for efficient hardware mitigations must continue.
Playing for K(H)eaps: Understanding and Improving Linux Kernel Exploit Reliability
Kyle Zeng, Arizona State University; Yueqi Chen, Pennsylvania State University; Haehyun Cho, Arizona State University and Soongsil University; Xinyu Xing, Pennsylvania State University; Adam Doupé, Yan Shoshitaishvili, and Tiffany Bao, Arizona State University
The dynamic of the Linux kernel heap layout significantly impacts the reliability of kernel heap exploits, making exploitability assessment challenging. Though techniques have been proposed to stabilize exploits in the past, little scientific research has been conducted to evaluate their effectiveness and explore their working conditions.
In this paper, we present a systematic study of the kernel heap exploit reliability problem. We first interview kernel security experts, gathering commonly adopted exploitation stabilization techniques and expert opinions about these techniques. We then evaluate these stabilization techniques on 17 real-world kernel heap exploits. The results indicate that many kernel security experts have incorrect opinions on exploitation stabilization techniques. To help the security community better understand exploitation stabilization, we inspect our experiment results and design a generic kernel heap exploit model. We use the proposed exploit model to interpret the exploitation unreliability issue and analyze why stabilization techniques succeed or fail. We also leverage the model to propose a new exploitation technique. Our experiment indicates that the new stabilization technique improves Linux kernel exploit reliability by 14.87% on average. Combining this newly proposed technique with existing stabilization approaches produces a composite stabilization method that achieves a 135.53% exploitation reliability improvement on average, outperforming exploit stabilization by professional security researchers by 67.86%.
Are Your Sensitive Attributes Private? Novel Model Inversion Attribute Inference Attacks on Classification Models
Shagufta Mehnaz; The Pennsylvania State University; Sayanton V. Dibbo and Ehsanul Kabir, Dartmouth College; Ninghui Li and Elisa Bertino, Purdue University
Increasing use of machine learning (ML) technologies in privacy-sensitive domains such as medical diagnoses, lifestyle predictions, and business decisions highlights the need to better understand if these ML technologies are introducing leakage of sensitive and proprietary training data. In this paper, we focus on model inversion attacks where the adversary knows non-sensitive attributes about records in the training data and aims to infer the value of a sensitive attribute unknown to the adversary, using only black-box access to the target classification model. We first devise a novel confidence score-based model inversion attribute inference attack that significantly outperforms the state-of-the-art. We then introduce a label-only model inversion attack that relies only on the model's predicted labels but still matches our confidence score-based attack in terms of attack effectiveness. We also extend our attacks to the scenario where some of the other (non-sensitive) attributes of a target record are unknown to the adversary. We evaluate our attacks on two types of machine learning models, decision tree and deep neural network, trained on three real datasets. Moreover, we empirically demonstrate the disparate vulnerability of model inversion attacks, i.e., specific groups in the training dataset (grouped by gender, race, etc.) could be more vulnerable to model inversion attacks.
Stalloris: RPKI Downgrade Attack
Tomas Hlavacek and Philipp Jeitner, Fraunhofer Institute for Secure Information Technology SIT and National Research Center for Applied Cybersecurity ATHENE; Donika Mirdita, Fraunhofer Institute for Secure Information Technology SIT, National Research Center for Applied Cybersecurity ATHENE, and Technische Universität Darmstadt; Haya Shulman, Fraunhofer Institute for Secure Information Technology SIT, National Research Center for Applied Cybersecurity ATHENE, and Goethe-Universität Frankfurt; Michael Waidner, Fraunhofer Institute for Secure Information Technology SIT, National Research Center for Applied Cybersecurity ATHENE, and Technische Universität Darmstadt
We demonstrate the first downgrade attacks against RPKI. The key design property in RPKI that allows our attacks is the tradeoff between connectivity and security: when networks cannot retrieve RPKI information from publication points, they make routing decisions in BGP without validating RPKI. We exploit this tradeoff to develop attacks that prevent the retrieval of the RPKI objects from the public repositories, thereby disabling RPKI validation and exposing the RPKI-protected networks to prefix hijack attacks.
We demonstrate experimentally that at least 47% of the public repositories are vulnerable against a specific version of our attacks, a rate-limiting off-path downgrade attack. We also show that all the current RPKI relying party implementations are vulnerable to attacks by a malicious publication point. This translates to 20.4% of the IPv4 address space.
We provide recommendations for preventing our downgrade attacks. However, resolving the fundamental problem is not straightforward: if the relying parties prefer security over connectivity and insist on RPKI validation when ROAs cannot be retrieved, the victim AS may become disconnected from many more networks than just the one that the adversary wishes to hijack. Our work shows that the publication points are a critical infrastructure for Internet connectivity and security. Our main recommendation is therefore that the publication points should be hosted on robust platforms guaranteeing a high degree of connectivity.
V'CER: Efficient Certificate Validation in Constrained Networks
David Koisser and Patrick Jauernig, Technical University Darmstadt; Gene Tsudik, University of California, Irvine; Ahmad-Reza Sadeghi, Technical University Darmstadt
We address the challenging problem of efficient trust establishment in constrained networks, i.e., networks that are composed of a large and dynamic set of (possibly heterogeneous) devices with limited bandwidth, connectivity, storage, and computational capabilities. Constrained networks are an integral part of many emerging application domains, from IoT meshes to satellite networks. A particularly difficult challenge is how to enforce timely revocation of compromised or faulty devices. Unfortunately, current solutions and techniques cannot cope with idiosyncrasies of constrained networks, since they mandate frequent real-time communication with centralized entities, storage and maintenance of large amounts of revocation information, and incur considerable bandwidth overhead.
To address the shortcomings of existing solutions, we design V'CER, a secure and efficient scheme for certificate validation that augments and benefits a PKI for constrained networks. V'CER utilizes unique features of Sparse Merkle Trees (SMTs) to perform lightweight revocation checks, while enabling collaborative operations among devices to keep them up-to-date when connectivity to external authorities is limited. V'CER can complement any PKI scheme to increase its flexibility and applicability, while ensuring fast dissemination of validation information independent of the network routing or topology. V'CER requires under 3KB storage per node covering 106 certificates. We developed and deployed a prototype of V'CER on an in-orbit satellite and our large-scale simulations demonstrate that V'CER decreases the number of requests for updates from external authorities by over 93%, when nodes are intermittently connected.
Oops... Code Execution and Content Spoofing: The First Comprehensive Analysis of OpenDocument Signatures
Simon Rohlmann, Christian Mainka, Vladislav Mladenov, and Jörg Schwenk, Ruhr University Bochum
OpenDocument is one of the major standards for interoperable office documents. Supported by office suites like Apache OpenOffice, LibreOffice, and Microsoft Office, the OpenDocument Format (ODF) is available for text processing, spreadsheets, and presentations on all major desktop and mobile operating systems.
When it comes to governmental and business use cases, OpenDocument signatures can protect the integrity of a document's content, for example, for contracts, amendments, or bills. Moreover OpenDocument signatures also protect document's macros. Since the risks of using macros in documents is well-known, modern office applications only enable their execution if a trusted entity signs the macro code. Thus, the security of ODF documents often depends on the correct signature verification.
In this paper, we conduct the first comprehensive analysis of OpenDocument signatures and reveal numerous severe threats. We identified five new attacks and evaluated them against 16 office applications on Windows, macOS, Linux, iOS, Android, and two online services. Our investigation revealed 12 out of 18 applications to be vulnerable for macro code execution, although the application only executes macros signed by trusted entities. For 17 of 18 applications, we could spoof the content in a signed ODF document while keeping the signature valid and trusted. Finally, we showed that attackers possessing a signed ODF could alter and forge the signature creation time in 16 of 18 applications.
Our research was acknowledged by Microsoft, Apache OpenOffice, and LibreOffice during the coordinated disclosure.
Identity Confusion in WebView-based Mobile App-in-app Ecosystems
Lei Zhang, Zhibo Zhang, and Ancong Liu, Fudan University; Yinzhi Cao, Johns Hopkins University; Xiaohan Zhang, Yanjun Chen, Yuan Zhang, Guangliang Yang, and Min Yang, Fudan University
Distinguished Paper Award Winner
Mobile applications (apps) often delegate their own functions to other parties, which makes them become a super ecosystem hosting these parties. Therefore, such mobile apps are being called super-apps, and the delegated parties are subsequently called sub-apps, behaving like "app-in-app". Sub-apps not only load (third-party) resources like a normal app, but also have access to the privileged APIs provided by the super-app. This leads to an important research question—determining who can access these privileged APIs.
Real-world super-apps, according to our study, adopt three types of identities—namely web domains, sub-app IDs, and capabilities—to determine privileged API access. However, existing identity checks of these three types are often not well designed, leading to a disobey of the least privilege principle. That is, the granted recipient of a privileged API is broader than intended, thus defined as an "identity confusion" in this paper. To the best of our knowledge, no prior works have studied this type of identity confusion vulnerability.
In this paper, we perform the first systematic study of identity confusion in real-world app-in-app ecosystems. We find that confusions of the aforementioned three types of identities are widespread among all 47 studied super-apps. More importantly, such confusions lead to severe consequences such as manipulating users' financial accounts and installing malware on a smartphone. We responsibly reported all of our findings to developers of affected super-apps, and helped them to fix their vulnerabilities.
How Machine Learning Is Solving the Binary Function Similarity Problem
Andrea Marcelli, Mariano Graziano, Xabier Ugarte-Pedrero, and Yanick Fratantonio, Cisco Systems, Inc.; Mohamad Mansouri and Davide Balzarotti, EURECOM
The ability to accurately compute the similarity between two pieces of binary code plays an important role in a wide range of different problems. Several research communities such as security, programming language analysis, and machine learning, have been working on this topic for more than five years, with hundreds of papers published on the subject. One would expect that, by now, it would be possible to answer a number of research questions that go beyond very specific techniques presented in papers, but that generalize to the entire research field. Unfortunately, this topic is affected by a number of challenges, ranging from reproducibility issues to opaqueness of research results, which hinders meaningful and effective progress.
In this paper, we set out to perform the first measurement study on the state of the art of this research area. We begin by systematizing the existing body of research. We then identify a number of relevant approaches, which are representative of a wide range of solutions recently proposed by three different research communities. We re-implemented these approaches and created a new dataset (with binaries compiled with different compilers, optimizations settings, and for three different architectures), which enabled us to perform a fair and meaningful comparison. This effort allowed us to answer a number of research questions that go beyond what could be inferred by reading the individual research papers. By releasing our entire modular framework and our datasets (with associated documentation), we also hope to inspire future work in this interesting research area.
FLAME: Taming Backdoors in Federated Learning
Thien Duc Nguyen and Phillip Rieger, Technical University of Darmstadt; Huili Chen, University of California San Diego; Hossein Yalame, Helen Möllering, and Hossein Fereidooni, Technical University of Darmstadt; Samuel Marchal, Aalto University and F-Secure; Markus Miettinen, Technical University of Darmstadt; Azalia Mirhoseini, Google; Shaza Zeitouni, Technical University of Darmstadt; Farinaz Koushanfar, University of California San Diego; Ahmad-Reza Sadeghi and Thomas Schneider, Technical University of Darmstadt
Federated Learning (FL) is a collaborative machine learning approach allowing participants to jointly train a model without having to share their private, potentially sensitive local datasets with others. Despite its benefits, FL is vulnerable to so-called backdoor attacks, in which an adversary injects manipulated model updates into the federated model aggregation process so that the resulting model will provide targeted false predictions for specific adversary-chosen inputs. Proposed defenses against backdoor attacks based on detecting and filtering out malicious model updates consider only very specific and limited attacker models, whereas defenses based on differential privacy-inspired noise injection significantly deteriorate the benign performance of the aggregated model. To address these deficiencies, we introduce FLAME, a defense framework that estimates the sufficient amount of noise to be injected to ensure the elimination of backdoors. To minimize the required amount of noise, FLAME uses a model clustering and weight clipping approach. This ensures that FLAME can maintain the benign performance of the aggregated model while effectively eliminating adversarial backdoors. Our evaluation of FLAME on several datasets stemming from application areas including image classification, word prediction, and IoT intrusion detection demonstrates that FLAME removes backdoors effectively with a negligible impact on the benign performance of the models.