In this paper, we present the design of Greystar, an innovative defense system for combating the growing SMS spam traffic in cellular networks. By exploiting the fact that most SMS spammers select targets randomly from the finite phone number space, Greystar monitors phone numbers from the grey phone space (which are associated with data only devices like laptop data cards and machine-to-machine communication devices like electricity meters) and employs a novel statistical model to detect spam numbers based on their footprints on the grey phone space. Evaluation using five month SMS call detail records from a large US cellular carrier shows that Greystar can detect thousands of spam numbers each month with very few false alarms and 15% of the detected spam numbers have never been reported by spam recipients. Moreover, Greystar is much faster in detecting SMS spam than existing victim spam reports, reducing spam traffic by 75% during peak hours.

Mobile telecommunication has become an important part of our daily lives. Yet, industry standards such as GSM often exclude scenarios with active attackers. Devices participating in communication are seen as trusted and non-malicious. By implementing our own baseband firmware based on OsmocomBB, we violate this trust and are able to evaluate the impact of a rogue device with regard to the usage of broadcast information. Through our analysis we show two new attacks based on the paging procedure used in cellular networks. We demonstrate that for at least GSM, it is feasible to hijack the transmission of mobile terminated services such as calls, perform targeted denial of service attacks against single subscribers and as well against large geographical regions within a metropolitan area.

Web and smartphone applications commonly embed third-party user interfaces like advertisements and social media widgets. However, this capability comes with security implications, both for the embedded interfaces and the host page or application. While browsers have evolved over time to address many of these issues, mobile systems like Android—which do not yet support true cross-application interface embedding—present an opportunity to redesign support for secure embedded user interfaces from scratch. In this paper, we explore the requirements for a system to support secure embedded user interfaces by systematically analyzing existing systems like browsers, smartphones, and research systems. We describe our experience modifying Android to support secure interface embedding and evaluate our implementation using case studies that rely on embedded interfaces, such as advertisement libraries, Facebook social plugins (e.g., the “Like” button), and access control gadgets. We provide concrete techniques and reflect on lessons learned for secure embedded user interfaces.

In this paper we tackle the challenge of providing a generic security architecture for the Android OS that can serve as a flexible and effective ecosystem to instantiate different security solutions. In contrast to prior work our security architecture, termed FlaskDroid, provides mandatory access control simultaneously on both Android’s middleware and kernel layers. The alignment of policy enforcement on these two layers is non-trivial due to their completely different semantics. We present an efficient policy language (inspired by SELinux) tailored to the specifics of Android’s middleware semantics. We show the flexibility of our architecture by policy-driven instantiations of selected security models such as the existing work Saint as well as a new privacy-protecting, user-defined and fine-grained per-app access control model. Other possible instantiations include phone booth mode, or dual persona phone. Finally we evaluate our implementation on SE Android 4.0.4 illustrating its efficiency and effectiveness.

Among anonymity systems, DC-nets have long held attraction for their resistance to traffic analysis attacks, but practical implementations remain vulnerable to internal disruption or “jamming” attacks, which require time-consuming detection procedures to resolve. We present Verdict, the first practical anonymous group communication system built using proactively verifiable DC-nets: participants use public-key cryptography to construct DC-net ciphertexts, and use zero-knowledge proofs of knowledge to detect and exclude misbehavior before disruption. We compare three alternative constructions for verifiable DC-nets: one using bilinear maps and two based on simpler ElGamal encryption. While verifiable DC-nets incur higher computational overheads due to the public-key cryptography involved, our experiments suggest that Verdict is practical for anonymous group messaging or microblogging applications, supporting groups of 100 clients at 1 second per round or 1000 clients at 10 seconds per round. Furthermore, we show how existing symmetric-key DC-nets can “fall back” to a verifiable DC-net to quickly identify misbehavior, speeding up previous detections schemes by two orders of magnitude.

Cloud storage service providers such as Dropbox, Mozy, and others perform deduplication to save space by only storing one copy of each file uploaded. Should clients conventionally encrypt their files, however, savings are lost. Message-locked encryption (the most prominent manifestation of which is convergent encryption) resolves this tension. However it is inherently subject to brute-force attacks that can recover files falling into a known set. We propose an architecture that provides secure deduplicated storage resisting brute-force attacks, and realize it in a system called DupLESS. In DupLESS, clients encrypt under message-based keys obtained from a key-server via an oblivious PRF protocol. It enables clients to store encrypted data with an existing service, have the service perform deduplication on their behalf, and yet achieves strong confidentiality guarantees. We show that encryption for deduplicated storage can achieve performance and space savings close to that of using the storage service with plaintext data.

Fake identities and Sybil accounts are pervasive in today’s online communities. They are responsible for a growing number of threats, including fake product reviews, malware and spam on social networks, and astroturf political campaigns. Unfortunately, studies show that existing tools such as CAPTCHAs and graph-based Sybil detectors have not proven to be effective defenses.

In this paper, we describe our work on building a practical system for detecting fake identities using server-side clickstream models. We develop a detection approach that groups “similar” user clickstreams into behavioral clusters, by partitioning a similarity graph that captures distances between clickstream sequences. We validate our clickstream models using ground-truth traces of 16,000 real and Sybil users from Renren, a large Chinese social network with 220M users. We propose a practical detection system based on these models, and show that it provides very high detection accuracy on our clickstream traces. Finally, we worked with collaborators at Renren and LinkedIn to test our prototype on their server-side data. Following positive results, both companies have expressed strong interest in further experimentation and possible internal deployment.

We perform an empirical study to better understand two well-known vulnerability rewards programs, or VRPs, which software vendors use to encourage community participation in finding and responsibly disclosing software vulnerabilities. The Chrome VRP has cost approximately $580,000 over 3 years and has resulted in 501 bounties paid for the identification of security vulnerabilities. The Firefox VRP has cost approximately $570,000 over the last 3 years and has yielded 190 bounties. 28% of Chrome’s patched vulnerabilities appearing in security advisories over this period, and 24% of Firefox’s, are the result of VRP contributions. Both programs appear economically efficient, comparing favorably to the cost of hiring full-time security researchers. The Chrome VRP features low expected payouts accompanied by high potential payouts, while the Firefox VRP features fixed payouts. Finding vulnerabilities for VRPs typically does not yield a salary comparable to a full-time job; the common case for recipients of rewards in either program is that they have received only one reward. Firefox has far more critical-severity vulnerabilities than Chrome, which we believe is attributable to an architectural difference between the two browsers.

Control-Flow Integrity (CFI) has been recognized as an important low-level security property. Its enforcement can defeat most injected and existing code attacks, including those based on Return-Oriented Programming (ROP). Previous implementations of CFI have required compiler support or the presence of relocation or debug information in the binary. In contrast, we present a technique for applying CFI to stripped binaries on x86/Linux. Ours is the first work to apply CFI to complex shared libraries such as glibc. Through experimental evaluation, we demonstrate that our CFI implementation is effective against control-flow hijack attacks, and eliminates the vast majority of ROP gadgets. To achieve this result, we have developed robust techniques for disassembly, static analysis, and transformation of large binaries. Our techniques have been tested on over 300MB of binaries (executables and shared libraries).

Low-level Inlined Reference Monitors (IRM) such as control-flow integrity and software-based fault isolation can foil numerous software attacks. Conventionally, those IRMs are implemented through binary rewriting or transformation on equivalent low-level programs that are tightly coupled with a specific Instruction Set Architecture (ISA). Resulting implementations have poor retargetability to different ISAs. This paper introduces an IRM-implementation framework at a compiler intermediate-representation (IR) level. The IR-level framework enables easy retargetability to different ISAs, but raises the challenge of how to preserve security at the low level, as the compiler backend might invalidate the assumptions at the IR level. We propose a constraint language to encode the assumptions and check whether they still hold after the backend transformations and optimizations. Furthermore, an independent verifier is implemented to validate the security of low-level code. We have implemented the framework inside LLVM to enforce the policy of control-flow integrity and data sandboxing for both reads and writes. Experimental results demonstrate that it incurs modest runtime overhead of 19.90% and 25.34% on SPECint2000 programs for x86-32 and x86-64, respectively

Computing devices with touch-screens have experienced unprecedented growth in recent years. Such an evolutionary advance has been facilitated by various applications that are heavily relying on multi-touch gestures. In addition, picture gesture authentication has been recently introduced as an alternative login experience to text-based password on such devices. In particular, the new Microsoft Windows 8TM operating system adopts such an alternative authentication to complement traditional text-based authentication. In this paper, we present an empirical analysis of picture gesture authentication on more than 10,000 picture passwords collected from over 800 subjects through online user studies. Based on the findings of our user studies, we also propose a novel attack framework that is capable of cracking passwords on previously unseen pictures in a picture gesture authentication system. Our approach is based on the concept of selection function that models users’ password selection processes. Our evaluation results show the proposed approach could crack a considerable portion of collected picture passwords under different settings.

Augmented reality (AR) applications sense the environment, then render virtual objects on human senses. Examples include smartphone applications that annotate storefronts with reviews and XBox Kinect games that show "avatars" mimicking human movements. No current OS has special support for such applications. As a result, permissions for AR applications are necessarily coarse-grained: applications must ask for access to raw sensor feeds, such as video and audio. These raw feeds expose significant additional information beyond what applications need, including sensitive information such as the user’s location, face, or surroundings.

Instead of exposing raw sensor data to applications directly, we introduce a new OS abstraction: the recognizer. A recognizer takes raw sensor data as input and exposes higher-level objects, such as a skeleton or a face, to applications. We propose a fine-grained permission system where applications request permissions at the granularity of recognizer objects. We analyze 87 shipping AR applications and find that a set of four core recognizers covers almost all current apps. We also introduce privacy goggles, a visualization of sensitive data exposed to an application. Surveys of 962 people establish a clear "privacy ordering" over recognizers and demonstrate that privacy goggles are effective at communicating application capabilities. We build a prototype on Windows that exposes nine recognizers to applications, including the Kinect skeleton tracker. Our prototype incurs negligible overhead for single applications, while improving performance of concurrent applications and enabling secure offloading of heavyweight recognizer computation.

In this paper we propose Sancus, a security architecture for networked embedded devices. Sancus supports extensibility in the form of remote (even third-party) software installation on devices while maintaining strong security guarantees. More specifically, Sancus can remotely attest to a software provider that a specific software module is running uncompromised, and can authenticate messages from software modules to software providers. Software modules can securely maintain local state, and can securely interact with other software modules that they choose to trust. The most distinguishing feature of Sancus is that it achieves these security guarantees without trusting any infrastructural software on the device. The Trusted Computing Base (TCB) on the device is only the hardware. Moreover, the hardware cost of Sancus is low.

We describe the design of Sancus, and develop and evaluate a prototype FPGA implementation of a Sancus-enabled device. The prototype extends an MSP430 processor with hardware support for the memory access control and cryptographic functionality required to run Sancus. We also develop a C compiler that targets our device and that can compile standard C modules to Sancus protected software modules.

Kernel rootkits undermine the integrity of system by manipulating its operating system kernel. External hardware-based monitors can serve as a root of trust that is resilient to rootkit attacks. The existing external hardware-based approaches lack an event-triggered verification scheme for mutable kernel objects. To address the issue, we present KI-Mon, a hardware-based platform for event-triggered kernel integrity monitor. A refined form of bus traffic monitoring efficiently verifies the update values of the objects, and callback verification routines can be programmed and executed for a designated event space. We have built a KI-Mon prototype to demonstrate the efficacy of KI-Mon’s event-triggered mechanism in terms of performance overhead for the monitored host system and the processor usage of the KI-Mon processor.

Application markets such as Apple’s App Store and Google’s Play Store have played an important role in the popularity of smartphones and mobile devices. However, keeping malware out of application markets is an ongoing challenge. While recent work has developed various techniques to determine what applications do, no work has provided a technical approach to answer, what do users expect? In this paper, we present the first step in addressing this challenge. Specifically, we focus on permissions for a given application and examine whether the application description provides any indication for why the application needs a permission. We present WHYPER, a framework using Natural Language Processing (NLP) techniques to identify sentences that describe the need for a given permission in an application description. WHYPER achieves an average precision of 82.8%, and an average recall of 81.5% for three permissions (address book, calendar, and record audio) that protect frequently used security and privacy sensitive resources. These results demonstrate great promise in using NLP techniques to bridge the semantic gap between user expectations and application functionality, further aiding the risk assessment of mobile applications.

Apple adopts the mandatory app review and code signing mechanisms to ensure that only approved apps can run on iOS devices. In this paper, we present a novel attack method that fundamentally defeats both mechanisms. Our method allows attackers to reliably hide malicious behavior that would otherwise get their app rejected by the Apple review process. Once the app passes the review and is installed on an end user’s device, it can be instructed to carry out the intended attacks.

The key idea is to make the apps remotely exploitable and subsequently introduce malicious control flows by rearranging signed code. Since the new control flows do not exist during the app review process, such apps, namely Jekyll apps, can stay undetected when reviewed and easily obtain Apple’s approval.

We implemented a proof-of-concept Jekyll app and successfully published it in App Store. We remotely launched the attacks on a controlled group of devices that installed the app. The result shows that, despite running inside the iOS sandbox, Jekyll app can successfully perform many malicious tasks, such as stealthily posting tweets, taking photos, stealing device identity information, sending email and SMS, attacking other apps, and even exploiting kernel vulnerabilities.

Note: Updated version contains the corrected acknowledgements

The Web’s principal security policy is the Same-Origin Policy (SOP), which enforces origin-based isolation of mutually distrusting Web applications. Since the early days, the SOP was repeatedly undermined with variants of the DNS Rebinding attack, allowing untrusted script code to gain illegitimate access to protected network resources. To counter these attacks, the browser vendors introduced countermeasures, such as DNS Pinning, to mitigate the attack. In this paper, we present a novel DNS Rebinding attack method leveraging the HTML5 Application Cache. Our attack allows reliable DNS Rebinding attacks, circumventing all currently deployed browser-based defense measures. Furthermore, we analyze the fundamental problem which allows DNS Rebinding to work in the first place: The SOP’s main purpose is to ensure security boundaries of Web servers. However, the Web servers themselves are only indirectly involved in the corresponding security decision. Instead, the SOP relies on information obtained from the domain name system, which is not necessarily controlled by the Web server’s owners. This mismatch is exploited by DNS Rebinding. Based on this insight, we propose a light-weight extension to the SOP which takes Web server provided information into account. We successfully implemented our extended SOP for the Chromium Web browser and report on our implementation’s interoperability and security properties.

We present new attacks and robust countermeasures for security-sensitive components, such as single sign-on APIs and client-side cryptographic libraries, that need to be safely deployed on untrusted web pages. We show how failing to isolate such components leaves them vulnerable to attacks both from the hosting website and other components running on the same page. These attacks are not prevented by browser security mechanisms alone, because they are caused by code interacting within the same origin. To mitigate these attacks, we propose to combine fine-grained component isolation at the JavaScript level with cryptographic mechanisms. We present Defensive JavaScript (DJS), a subset of the language that guarantees the behavior integrity of scripts even when loaded in a hostile environment. We give a sound type system, type inference tool, and build defensive libraries for cryptography and data encodings. We show the effectiveness of our solution by implementing several applications using defensive patterns that fix some of our original attacks. We present a model extraction tool to analyze the security properties of our applications using a cryptographic protocol verifier.

The talk will open with an entertaining demonstration of what is widely considered the best casino-cheating move ever that made millions and fooled state-of-the-art video surveillance technology only because the move was so simple-stupid. Marcus will move on to cover how modern-day casino cheats use computers to scam tables and how casinos battle back with their own. He will discuss roulette computers, RFID technology, hacking into online gaming, and hacking into land casinos’ security systems.

For 25 years, Richard Marcus directed sophisticated casino-cheating teams across the entire world. He developed and carried out a series of cheating moves that not only fooled surveillance video cameras but also turned them into his allies. Today Marcus serves as an international casino game-protection consultant training casino floor and surveillance personnel. He has published five books and written dozens of magazine articles on the subject, and has been a guest on several network TV news and entertainment shows including NBC Today and Entertainment Tonight.