The full Proceedings published by USENIX for the symposium are available for download below. Individual papers can also be downloaded from the presentation pages. Copyright to the individual works is retained by the author[s].
Full Proceedings
SOUPS 2026 Full Proceedings (PDF, 120.68 MB)
SOUPS 2026 Errata Slip #1
View the full program on the SOUPS 2026 website.
Monday, August 24
Authentication & Credentials
An Analysis of the Security, Usability, and Automation Capabilities of Password Update Processes on Top-Ranked Websites
Alexander Krause, Jacques Suray, Lea Schmüser, Marten Oltrogge, Oliver Wiese, Maximillian Golla, and Sascha Fahl, CISPA Helmholtz Center for Information Security
Password updates are a critical part of the password lifecycle and are recommended following exposure of reused passwords or suspected compromise. However, password update processes are often cumbersome, require manual password creation, and involve inconsistent website workflows that hinder reliable automation by password managers.
In this work, we conduct the first in-depth, systematic analysis of 111 password update processes deployed on top-ranked websites. We provide novel insights into their overall security, usability, and automation capabilities, and contribute to authentication security research by improving the understanding of password update processes. Websites often deploy highly diverse, complex, and confusing password update processes that are not supported by password managers. Processes are often challenging to use, and end-users struggle to transfer experience and knowledge across websites. Notably, security measures designed to enhance security often hinder password manager automation. We conclude our work by discussing our findings and giving recommendations for web developers, the web standardization community, and security researchers.
Passkeys in the Wild: A Systematic Study of FIDO2 User Experience Consistency Across Websites
Bernhardt Ramat, Dave Kartchner, Michael Clark, and Kent Seamons, Brigham Young University
Passkeys have emerged as a phishing-resistant alternative to passwords, yet little is known about how consistently they are implemented across the web. Prior usability research suggests that cross-site inconsistencies increase cognitive load and hinder adoption, as users carry expectations from one website to another. To address this knowledge gap, we conducted a systematic analysis of passkey user experiences across 111 websites using a comparative framework of 28 factors. Cluster analysis reveals that core passkey support is converging, but supporting features remain uneven. Future research can use our framework for longitudinal comparisons and studies targeting specific factors of interest.
How Users Enter Generated Passwords on Non-Desktop Devices
John Sadik, Joshoua Rodriguez, Hector Santos Villalobos, and Scott Ruoti, The University of Tennessee, Knoxville
Generated passwords remain a powerful tool for account security, yet they are often avoided due to fear of manual entry. Developing better generated passwords requires a fine-grained understanding of how they are entered in practice, but no such public dataset currently exists. We contribute the first such dataset, containing nearly 80,000 keystrokes from 4,916 password entries collected from 99 participants using five different input modalities. Our analysis reveals that generated password entry is driven by physical kinematic costs, such as layout switching, rather than the interface-level treatments in our study. Using these insights, we design two novel device-aware password generation algorithms as proof-of-concept models optimizing for travel distance and rhythmic entry flow. We then leverage machine learning models trained on our dataset to analytically validate these algorithms, finding that they reduce the predicted entry time by 27.55% and 31.28% respectively compared to standard randomly generated passwords while maintaining sufficient cryptographic entropy. We release our research artifacts to promote the development of new password generation algorithms, and discuss the how these insights can inform better password authentication designs.
Not All Is Lost: Partial Recovery & Memorability in Self-Sovereign Digital Identity
Sushanth Ambati, Dylan Jablonski, Charalampos Papachristou, Brian Pugliese, Jack Myers, and Nikolay Ivanov, Rowan University
The shift toward Self-Sovereign Digital Identity (SSDI) promises to mitigate the epidemic of massive personal data leaks, but it fundamentally relies on users securely managing high-strength private keys. Traditional backup methods, such as hardware wallets, paper records, and cloud vaults, inevitably compromise either security, locality, or availability. While human memory offers a uniquely theft-proof and local alternative, the cognitive burden of memorizing strong cryptographic keys remains a critical usability barrier. Crucially, prior usable security research has largely treated key memorability as a binary outcome — either perfectly recalled or completely forgotten. We challenge this paradigm through a lab-based randomized controlled trial (N=92) of 128-bit (NIST-minimum) keys. Comparing standard 12-word BIP-39 mnemonics against a graphical key scheme, we demonstrate how combinatorial search can leverage fragmented human memory to reconstruct keys without degrading theoretical security bounds. We present an empirical evaluation of how cognitive-science principles affect memorability of full 128-bit system-assigned keys, paired with a novel computational recovery technique. Our findings reveal an advantage for graphical keys, suggesting that augmenting BIP-39 mnemonics with graphical cues and combinatorial recovery of partially forgotten keys may offer a promising path forward.
Families & Shared Devices
SoK: The Design Space of Usable Privacy Interventions for Parents: A Systematization of Knowledge
Ann-Kristin Lieberknecht, Goethe University Frankfurt; Sascha Loebner, University of Hamburg
Privacy and security risks increasingly emerge through everyday family technology use, requiring families to navigate complex decisions about children’s data and digital participation. Parents play a central role in these processes, interpreting risks, configuring technologies, and negotiating privacy practices with children while adapting to rapidly evolving digital environments. Research addressing these challenges spans multiple communities, including usable privacy, family human–computer interaction, and digital literacy, yet remains fragmented across studies of parental experiences, behavioral determinants, and intervention design. This paper presents a systematization of knowledge synthesizing how research has approached supporting parental privacy practices. We compile a focused corpus combining evaluated interventions, parent-focused empirical pre-studies, and quantitative investigations of behavioral determinants. We develop an analytic framework distinguishing intended parental learning outcomes from the design features through which support is operationalized, enabling comparison across heterogeneous evidence. Mapping interventions against this framework and relating them to empirical patterns reveals how knowledge about parents’ privacy challenges is selectively translated into implemented forms of support. The results highlight recurring concentrations and gaps in parental privacy interventions and illuminate the practical challenges of enabling privacy practices within everyday family life.
"They are not my children to post": Examining Non-Parental Sharenting Practices in In-home Childcare
Meghna Gupta, University of Washington; Sophie Stephenson, University of Wisconsin–Madison; Apu Kapadia, Indiana University Bloomington; Julie Kientz and Franziska Roesner, University of Washington
Taking and sharing photos is a common aspect of caregiving, often used to document and share the care-recipient’s milestones. Although extensively studied in the context of ’sharenting’, where parents share photos of their children, these practices also extend to other caregivers, including paid professionals such as nannies and babysitters. Despite the prevalence of photo taking and sharing, there remains a limited understanding of the complexities and dynamics of these behaviors within professional caregiving contexts. Examining these practices is essential to understanding the unique privacy implications of non-parental figures taking and sharing photos of children.
We conducted 15 semi-structured interviews with parents and in-home childcare workers in the U.S. to explore how they perceive caregivers’ taking and sharing photos of children in their care. Our findings reveal misalignment and limited negotiations among parents and childcare workers regarding taking and sharing photos–exacerbated by underlying assumptions, power imbalances, and blurred personal-professional boundaries. These dynamics significantly shape these practices, and result in internal turmoils among both parents and childcare workers. We conclude with implications and recommendations for future research for fostering better communication and sanctions in context of non-parental sharenting.
Sharing Digital Devices is Normal and Cultural: Privacy and Security Challenges in Collectivist Immigrant Households
S. Shanza, University of Waterloo; Ameemah Humayun, Lahore University of Management Sciences; Urs Hengartner and Leah Zhang-Kennedy, University of Waterloo
In collectivist households, sharing digital devices is both a cultural norm and often necessary. After migration to more individualistic cultures such as Canada, these sharing practices persist but become entangled with new cultural norms and highly digitized systems. We report findings from a qualitative study of 18 nuclear families (19 parents, 22 children) mainly from South-Asia examining how migration and sociocultural factors shape security and privacy risks in multi-user households and how current devices support or hinder safe sharing. Shared-device use exposed families to data breaches, reputational harms, cultural and religious boundary violations, and intergenerational tensions. At the same time, coping strategies were often fragile or abandoned when they threatened family harmony. Our findings reveal a structural mismatch between single-user device assumptions and collectivist immigrants’ household realities. Designing for shared-device contexts is therefore essential for building more inclusive digital infrastructures that support multicultural families and others who rely on device sharing.
From Thrift Stores to Digital Storefronts: Users' Perspectives on Privacy and Security of Online Second-Hand Shopping in Germany
Darya Zarkalam and Anna Lena Rotthaler, Paderborn University; Lucy Simko, Barnard College; Yasemin Acar, Paderborn University
The emergence of online second-hand shopping platforms such as eBay, Craigslist, and Facebook Marketplace has added new ways for consumers to shop second-hand. Although these platforms offer some features for security and safety, users still encounter fraud, scams, and personal safety risks when (trying to) purchase second-hand goods online. Here, we seek to understand users’ privacy, safety, and security practices when using online second-hand shopping platforms, both technical and interpersonal. We conducted semi-structured interviews with 20 users of online second-hand shopping platforms in Germany, to learn their security and privacy perceptions and concerns, experiences, and protective strategies. Our study identifies key security and privacy concerns, including scams, personal safety, and the misuse of personal information. We find that platform-integrated features, such as user reviews and payment systems, are perceived as effective for mitigation. Furthermore, we find that trust in platforms is shaped by social factors. Based on these insights, we make recommendations for platforms to increase user safety.
Tuesday, August 25
Privacy Notices, Permissions & Labels
How Effective are Privacy Labels at Informing Users About App Data Handling Practices?
Sophia Walsh, University of Bristol; Lukasz Piwek, University of Bath; Kopo Marvin Ramokapane, University of Bristol
Mobile app privacy labels aim to enhance transparency by helping users understand app data-handling practices. However, limited empirical evidence demonstrates whether they achieve this goal. Using a card-matching activity with think-aloud protocols, we examined how users interpret iOS Privacy Nutrition Labels and Android Data Safety Labels. Participants (n=16) matched privacy labels to corresponding apps, revealing their existing knowledge of app data-handling practices and engagement patterns. Thematic analysis generated three themes: (1) Understanding – how users comprehend app data-handling practices; (2) Salience – which label elements participants notice; and (3) Impact – how awareness of actual data practices influences users. Our findings extend existing work on technical terminology comprehension, providing rich qualitative insights into user assumptions, heuristic reasoning, and emotional responses. While privacy labels can advance transparency goals, their current presentation undermines their utility. Future label designs should aim to dispel rather than compound misconceptions, develop heuristic abstractions to communicate information effectively, and reduce negative emotional impacts.
I don't know what I've all granted. Does it really matter? – Understanding Users' Awareness of Different Permission Types on Android
Verena Winterhalter, Sarah Prange, and Anouk Moreno, LMU Munich; Harel Israel Berger, Ariel University; Florian Alt, LMU Munich and University of the Bundeswehr Munich
In this paper, we present the results of an in-the-wild study (N = 77) using an Android study app that captures real permission configurations, in-situ experience sampling of participants’ perceptions of which permissions apps hold, and risk awareness for permission combinations during everyday use. Such understanding matters because Android attacks can exploit individual permissions and, in particular, risky combinations. At the same time, it is unclear how often users grant such permissions (and combinations) and whether they are aware of this threat. We found that users frequently misjudge app permission states, regardless of permission type. This also holds for permissions that form part of risky combinations. 56.76% of responses mismatched the actual permission state, including 60.70% for Runtime permissions, 51.19% for Install-time permissions, and 47.67% for permissions which are part of risky combinations reported in prior security literature. Because awareness of granted permissions is a prerequisite for informed action, these gaps hinder users’ ability to mitigate privacy and security risks. We discuss the implications for permission user interfaces and platform mechanisms to better surface risks arising from permission combinations.
Nudging Developers Toward Privacy: Evaluating the Impact of Personalized App Review Reports
Sai Teja Peddinti, Google; Omer Akgul, RSAC Research; Michelle L. Mazurek, University of Maryland; Nina Taft, Google
Mobile application developers often struggle to create accurate privacy notices or implement robust privacy practices due to limited expertise or resources. While users share unsolicited privacy feedback in app reviews, and prior research has characterized this privacy feedback, uncovering developer reactions to this feedback remains unexplored. This study explores whether personalized privacy review reports—summarizing real user feedback for a developer’s own app—can effectively nudge them toward planning privacy improvements. We surveyed 42 app developers, presenting them with reports containing privacy themes, temporal trends, peer benchmarks, and emotion distributions derived from their apps’ reviews. Our findings indicate that these privacy report interventions proved highly effective, with 76% (32 of 42) of participants finding at least one section of the report useful. Furthermore, exposure to the report increased the participants’ intent to pursue privacy-relevant actions – such as reorganizing the UI, enhancing privacy communications, or adding/removing features – with 69% (29 of 42) of participants indicating an increased intent to do so. Almost all developers expressed a desire to receive such privacy reports periodically or on demand. These results indicate that making this style of report broadly available across the industry could foster a more privacy-conscious mobile ecosystem.
Security Operations & Practitioners
Like a Hammer, It Can Build, It Can Break: Large Language Model Uses, Perceptions, and Adoption in Cybersecurity Operations on Reddit
Souradip Nath, Chih-Yi Huang, and Aditi Ganapathi, Arizona State University; Kashyap Thimmaraju, Technische Universität Berlin; Jaron Mink and Gail-Joon Ahn, Arizona State University
Large language models (LLMs) have recently emerged as promising tools for augmenting Security Operations Center (SOC) workflows, with vendors increasingly marketing autonomous AI solutions for SOCs. However, there remains a limited empirical understanding of how such tools are used, perceived, and adopted by real-world security practitioners. To address this gap, we conduct a mixed-methods analysis of discussions in cybersecurity-focused forums to learn how a diverse group of practitioners use and perceive modern LLM tools for security operations. More specifically, we analyzed 892 posts between December 2022 and September 2025 from three cybersecurity-focused forums on Reddit, and, using a combination of qualitative coding and statistical analysis, examined how security practitioners discuss LLM tools across three dimensions: (1) their stated tools and use cases, (2) the perceived pros and cons of each tool across a set of critical factors, and (3) their adoption of such tools and the expected impacts on the cybersecurity industry and individual analysts. Overall, our findings reveal nuanced patterns in LLM tools adoption, highlighting independent use of LLMs for low-risk, productivity-oriented tasks, alongside active interest around enterprise-grade, security-focused LLM platforms. Although practitioners report meaningful gains in efficiency and effectiveness in LLM-assisted workflows, persistent issues with reliability, verification overheads, and security risks sharply constrain the autonomy granted to LLM tools. Based on these results, we also provide recommendations for developing and adopting LLM tools to ensure the security of organizations and the safety of cybersecurity practitioners.
The Impact of Emerging AI Practices on the Cybersecurity Workforce
Miuyin Yong Wong, Alan Luo, Yunze Zhao, and Shubham Bhatnagar, University of Maryland; Fabian Monrose, Georgia Institute of Technology; Michelle L. Mazurek, University of Maryland
Large language models (LLMs) have frequently been proposed as useful tools for cybersecurity tasks like malware detection, reverse engineering, and incident response. However, little is known about whether and how cybersecurity practitioners have adopted LLMs in enterprise settings, how they perceive these tools and communicate about them with peers, and what benefits or challenges they encounter in practice. To fill this gap, we conducted 28 semi-structured interviews with cybersecurity professionals from 26 organizations, with varying experience levels. We find that although LLM use is common and frequent, organizational policies governing this use are often vague, inconsistent, and unclear to individual practitioners. Practitioners face important challenges related to the fundamental limitations of LLMs—such as unreliable outputs and context window limits—as well as domain-specific challenges, including poor performance on specific cybersecurity tasks and restrictive guardrails that inhibit penetration testing. Other key concerns included data leakage and productivity loss. Beyond technical limitations, we find that confusion around informal guidance and limited peer communication is creating a psychological barrier, influencing practitioners’ adoption of LLMs and altering the traditional collaborative culture and apprenticeship knowledge-transfer model of the cybersecurity community.
From Preventive to Reactive: How AI Coding Assistants Transform Developers' Security Awareness
Faisal Haque Bappy and Tahrim Hossain, University of Maryland Baltimore County; Sidratul Muntaher Meheraj, Annoor Sharara Akhand, and Tasfia Tabassum, University of Dhaka; Tarannum Shaila Zaman, University of Maryland Baltimore County; Raiful Hasan, Kent State University; Tariqul Islam, University of Maryland Baltimore County
AI coding assistants are now central to professional software development, yet their impact on how developers think about and practice security remains poorly understood. While prior work has documented vulnerability rates in AI-generated code, a more fundamental question persists: how do these tools transform security awareness in authentic, ongoing development practice? We conducted semi-structured interviews with 15 professional software engineers and observed them completing security-relevant coding tasks with AI assistance, spanning 3 experience cohorts defined by their relationship to AI tools during professional formation. We find that AI coding assistants reorganize rather than eliminate security thinking, shifting it from the act of writing code to the act of reviewing it. This transition from preventive to reactive security is structurally encouraged by interaction models that frame code generation as a functional task, leaving security as an afterthought. Notably, none of our coding session participants specified security requirements in their initial prompts, even when they possessed the relevant knowledge, revealing a decoupling of security awareness from security behavior. We further document informal coping strategies developers had independently invented to manage AI security risk, none of which are supported by current tools or organizations, and find that the experience cohort did not reliably predict security performance. This paper contributes a practice-grounded account of how AI-assisted development reshapes the human side of secure coding, offering empirical foundations for the design of more security-aware tools, training programs, and organizational policies.
I See DNS People: DNS Resolver Security, Through Operator Perspectives and Practices
Wisdom Obinna, Georgetown University; Katharina Barlage, LMU Munich; Florian Alt, LMU Munich and University of the Bundeswehr Munich; Harel Israel Berger, Ariel University
We study how DNS resolver operators and security practitioners make DNS security decisions. Prior research has measured DNS security adoption rates and misconfiguration, but offers limited insight into why operators make the choices they do. To address this gap, we conducted semi-structured interviews with 24 DNS resolver operators and security practitioners from 14 countries, covering service providers, enterprises, and government organizations. Our analysis identifies three interconnected barriers impeding wider adoption. First, there is a gap between how complex DNSSEC is perceived and how complex it is in practice: operators frequently describe DNSSEC as intimidating prior to deployment, yet majority who deployed it reported the process simpler than expected. Two factors help explain why this gap persists. DNS suffers from systematic deprioritization: because DNS "just works," organizations tend to pay less attention to DNS security until failure. Secondly, expertise gaps, combined with limited resources and coordination costs, amplify the prior two factors. Grounded in practitioners’ own recommendations, we suggest actionable interventions to reduce the complexity of DNSSEC, such as clearer documentation and visible success stories, while also addressing its deprioritization through compliance mandates or security-as-a-mission framing.
Phishing & Social Engineering
“I didn’t know I would be this excited not to be scammed.” Exploring Emotional and Behavioral Responses During Phishing Attacks
Raphael Weidhaas, Aalto University; Alexandra von Preuschen, CISPA Helmholtz Center for Information Security; Verena Distler, Aalto University
Humans remain vulnerable to phishing attacks, despite awareness, training, and technical measures. The role of emotions during phishing attacks is largely unexplored. Building on the appraisal theory of emotion, we investigated (1) which emotions are invoked during a phishing attack, (2) related cognitive appraisals, (3) behavioral responses during an attack, and (4) reflections about behavior throughout the attack. We conducted a qualitative study (N=41), simulating a workplace scenario during which a phishing attack occurs. We combined observation, interviews, and questionnaires. Emotions differed based on whether the phish was detected or not. The phish was frequently appraised as an interruption, invoking negative emotions. Not clicking on the phishing link was often appraised as a possible mistake, which may lead to reputational damage. Many participants reflected detecting the phish intuitively. This work contributes to an empirical understanding of emotions and behavior during phishing and provides research avenues and implications for how anti-phishing measures could benefit from emotion-aware solutions.
Anxious and Aware: Examining the Effects of Social Anxiety on Social Engineering Resilience and Vulnerability
Martin Dechant, Victoria Woo, Florence Tomlin, Bettina Moltrecht, and Mark Warner, University College London
Social engineering attacks manipulate individuals into acting against their own interests, by exploiting users’ cognitive capabilities. Less is known about how characteristics of social anxiety, a trait that exists on a spectrum ranging from mild situational anxiety to a clinical threshold affecting daily life, may influence susceptibility to social engineering attacks. The present study starts to address this gap. We conducted an online survey (N=42) to identify real-world social engineering tactics, which informed role-playing interviews (N = 20) that compared responses in individuals on the higher and lower end of the SoAnx spectrum. Our findings show that characteristics of SoAnx can create vulnerabilities and benefits depending on the type of attack. Participants suggested technology solutions, ranging from educational resources and automation to social community support. We contribute new qualitative insights into how SoAnx may affect both vulnerability and resilience against social engineering attacks.
Quantifying Risk Perception and Scam Response Among International and Domestic US University Students
Alexandra Xinran Li, Elijah Bouma-Sims, Lily Klucinec, and Ray Liu, Carnegie Mellon University; Ayesha Binte Mostofa, University of Massachusetts, Amherst; Arjun Arunasalam, Florida International University; Lorrie Faith Cranor, Carnegie Mellon University; Pubali Datta, University of Massachusetts, Amherst; Lucy Simko, Barnard College, Columbia University; Karen Sowon, Indiana University
We conducted an online survey of 1,074 students from five U.S. universities to examine and quantify the differences between domestic and international students’ scam encounters, engagement behaviors, reporting tendencies, and exposure to anti-scam training. International students were significantly more likely than domestic students to say they would engage with scammers when presented with suspicious phone-call and text-message scenarios, but also more likely to disengage in our suspicious-rental email scenario. We found no difference in their ability to detect suspicious cues, but international students may require more evidence before disengaging in certain scenarios. While domestic students are more likely to encounter generic scams, international students reported more loss from authority-based, rental, and "unknown language" scams. International students reported a greater reliance on social media and peer support to avoid scams. They were significantly more likely to report incidents to law enforcement than domestic students, but less likely to report overall, use call filters, or online platform tools. We provide recommendations for improving both scam training offered by universities and digital scam reporting platforms.
Experiences with Digital Scams Post-Incarceration in the U.S.
Yael Eiger, University of Washington; Candice Baughman, Interaction Transition; Rory Andes, Communities of Belonging; Bryan Glant, Project Rise; Franziska Roesner, University of Washington
When people are released from prison in the United States, they reenter into a highly digitized world that they have little to no experience with or preparation for. This lack of experience makes them potentially vulnerable to digital scams and phishing attempts upon release. Towards addressing this digital divide, we ran a workshop to teach and learn about digital security and privacy needs unique to the formerly incarcerated community (in our area of the U.S.). At this workshop, we deployed a survey to both participants (N=20) and volunteers (N=12). We later followed up with a more targeted survey specifically asking about scams and scam susceptibility among formerly incarcerated people (N=20, 4 of whom also attend the workshop). We additionally include personal narratives from three formerly incarcerated researchers (co-authors on the paper) about their experiences with scams. Combining all of this data (N=48 unique respondents), we find that scam exposure and victimization in this community is common, and we document a variety of risk factors: digital literacy challenges, heightened coercive potential of scams which purport to be from law enforcement or governmental bodies, loneliness during incarceration, financial instability post-incarceration, and a fear of stigmatization and repercussions from reporting. At the same time, we find that participants recount some mitigating factors, including belonging to a community that faces common challenges and therefore builds strong community networks to turn to for help. We close by identify gaps between best practices to avoid scam victimization and the realities of being currently or formerly incarcerated in the U.S., and making recommendations.
At-Risk & Vulnerable Users
Usability Determines Safety for At-Risk Users: Evaluating Hidden Device Detectors for Intimate Partner Surveillance
Akhil Polamarasetty, Leonie Maria Tanczer, Enrico Costanza, and Kevin Chetty, University College London (UCL)
Hidden surveillance devices pose serious risks to survivors of intimate partner surveillance (IPS), yet consumer detection tools remain poorly understood from the user’s perspective. Across in-room trials with 34 participants using 19 tools spanning eight interface types, we evaluated whether people can turn detector output into successful search, verification, and identification. Effectiveness depended on the actionability of feedback, not on sensing capability alone. Tools that presented a navigable list of detected devices with clear next steps outperformed signal-strength designs across every outcome, even when they sensed far less. Emotional impact proved equally consequential: false alarms heightened anxiety, and opaque failures led users to blame themselves. We offer design recommendations that treat detectors as decision-support tools and propose reusable dimensions for evaluating them.
Goals, Risks, and Safety Practices in Online Labor Abuse Disclosures
Veronica A. Rivera, Max Planck Institute for Security and Privacy, Stanford University, and Georgia Institute of Technology; Tracy Li, Stanford University; Alex Ozdemir, Max Planck Institute for Security and Privacy and Georgia Institute of Technology; Catherine Han and Zakir Durumeric, Stanford University; Elissa M. Redmiles, Georgetown University
Survivors of workplace labor abuse disclose sensitive experiences online to build collective power, seek support, and warn others. Doing this without triggering retaliation or further abuse requires achieving targeted visibility: reaching supportive audiences while remaining hidden from abusers. Existing social media tools offer limited support for navigating the tensions between visibility, trust-building, and obscurity, and current threat models for at-risk groups do not account for these dynamics. We report findings from interviews with 17 survivors of labor abuse who disclosed their experiences online. Survivors disclose to different platforms and audiences to manage risk and evade adversaries. Yet, these strategies often fail to meaningfully reduce risk and can also undermine their goals. We extend existing frameworks of at-risk technology use by foregrounding targeted visibility and conclude with implications for designing systems that better support survivors of labor abuse and other interpersonal harms.
“I feel as though my privacy is being violated”: Privacy Risks and Barriers in Instant Messaging for Blind and Low-Vision Users
Sohana Akter, University of Texas at San Antonio; Wejdan Al Sarih, University of Bristol; Blessy Kalluri, University of Texas at San Antonio; Kopo M. Ramokapane, University of Bristol; Taslima Akter, University of Texas at San Antonio
Instant messaging applications are central to social and professional life. Yet managing privacy around their use remains a challenge, particularly for blind and low-vision (BLV) users. We conducted an exploratory qualitative study with 20 BLV participants to examine their privacy concerns and threats, barriers to exercising privacy agency, and the strategies they adopt to protect their privacy on instant messaging platforms. Our findings show that BLV users often cannot verify what content or profile information is visible to others, are frequently constrained by default settings beyond their control, and sometimes must expose private content to access basic accessibility support, such as by sharing images with AI tools for interpretation. Furthermore, participants reported ongoing difficulty discovering, understanding, and navigating the privacy features that do exist. These findings reveal critical gaps at the intersection of accessibility and privacy in instant messaging applications and we conclude by discussing design implications for more inclusive and privacy-preserving platforms.
"Don't Let Them Get To You": Understanding the Role of TikTok as a Source of Support for Cyberbullying Victims
Saba Iqbal, Warda Usman, and Daniel Zappala, Brigham Young University
Cyberbullying poses a significant threat to teenagers and young adults, who often seek support and advice online. Little is known about the support and advice shared online for victims of cyberbullying and whether it is helpful. In this study, we used qualitative thematic analysis to examine a sample of 150 TikTok videos that give support or advice for cyberbullying victims. We identified 33 unique pieces of support and advice across 5 categories, with differences in tone and delivery among different types of videos. We then evaluated the quality of the support and advice in a survey of 493 young adults who had prior experiences with cyberbullying. Young adults perceive 97% of the support and advice as comprehensible and 60% as effective, but only 37% of the advice is seen as actionable. However, they did identify 6 pieces of advice that are both effective and actionable, as well as 4 types of emotional support that are effective. Participants held nuanced views of TikTok as a platform for advice. We discuss insights from our study that may lead to better support for youth.
Reproductive Security & Privacy Advice on TikTok after the Overturn of Roe
Harshini Sri Ramulu, Rachel Gonzalez Rodriguez, and Yasemin Acar, Paderborn University; Lucy Simko, Barnard College
In 2022, the Supreme Court of the United States overturned Roe v. Wade, a seminal case linking the right to privacy to the right to reproductive self determination at the federal level. Reproductive self determination in the US is now regulated at the state level, with vast differences across states. With the current landscape of online tracking, people who may become pregnant are at risk of prosecution based on data from their digital footprint, including online searches, period trackers, and fitness trackers. After the overturn of Roe v. Wade, social media creators reacted on TikTok, including by giving security and privacy advice regarding reproductive health under the new legal situation.
To empower vulnerable users to protect their security and privacy, we need to understand the landscape of security and privacy advice immediately after significant shifts in legislation. We qualitatively analyze 92 TikTok videos giving advice on reproductive security and privacy after the overturn of Roe v. Wade. We find that content creators connected general-purpose security advice (like using encrypted messengers) to reproductive privacy, and that domain-specific advice (like abandoning period tracking apps) emerged. Though each piece of advice was often sound, it collectively lacked nuance, actionability, completeness, and practicality due to the complexities of the legal, technical, and interpersonal threat landscape. We provide recommendations for advice-givers, social media platforms, and the security community, towards stronger, more actionable, and more complete communication of domain-specific security and privacy advice during crisis.
Wednesday, August 26
Usable Security Mechanisms & Emerging Threats
IdentitySign: Design and User Study of a Prototype Application for Digitally Signing Documents Using an Identity Wallet
Yorick Last, Paderborn University; Hanna Schraffenberger, Daniel Ostkamp, and Jorrit Geels, Radboud University; Patricia Arias Cabarcos, European Commission Joint Research Centre (JRC)
When signing digital documents, users often insert scanned or digitized handwritten signatures, which provide no assurance of a document’s authenticity and integrity. Cryptographic signatures offer such guarantees, but are rarely adopted due to persistent usability challenges. This paper explores the potential of digital identity wallets for securely and intuitively signing documents with verified personal data. To this end, we developed IdentitySign, an open-source prototype application, and evaluated it through a user study (N=74). Our results indicate good usability but also reveal potential pitfalls for this and similar wallet-based signature applications, such as unjustified trust in documents with signatures that are cryptographically valid but signed by the wrong signer. We provide guidelines and insights for creating user-centric, secure, and trustworthy digital signage applications using identity wallets.
Is It Real? Exploiting Virtual-Physical Discrimination Vulnerability in Mixed Reality
Xueyang Wang and Xihuan Yao, Tsinghua University; Yanming Xiu, Duke University; Xin Yi, Tsinghua University and Beijing Academy of Artificial Intelligence; Maria Gorlatova, Duke University; Hewu Li, Tsinghua University
Consumer mixed reality (MR) headsets seamlessly blend virtual content into physical environments with sufficient fidelity that users may be unable to distinguish virtual objects from physical ones. We identify this virtual-physical discrimination vulnerability as an exploitable security primitive. Through speculative design workshops with 12 experts from cybersecurity and MR/HCI, we develop a taxonomy of virtual-physical confusion attacks and implement four proof-of-concept attacks on Apple Vision Pro, evaluating them with 26 participants in realistic MR tasks. All four attacks altered user behavior, with success rates ranging from 85% to 100%, producing misdirected interactions, misjudged object identities, biased purchasing decisions, and altered navigation paths. Notably, the most successful attacks were also the hardest to detect according to participants’ subjective ratings. Even participants who recognized virtual content still complied behaviorally, and no participant attributed anomalous events to adversarial causes. We propose platform-level provenance, interaction gating, and user education as countermeasures.
Clear, Actionable and Confidence-Inspiring Recommendations? Comparative Study of AI-Generated and Human-Written PT Reports
Katarina Galanska, Masaryk University; Maria Pibilota Murumaa, University of Tartu; Vashek Matyas and Agata Kruzikova, Masaryk University; Mike Just, Heriot-Watt University; Tomas Cerny, The University of Arizona
Penetration testing reports typically include remediation recommendations to guide organizations in addressing identified vulnerabilities. The usability of such recommendations influences whether the reports are effectively understood and acted upon. Yet, the usability implications of different recommendation formulations, including AI-generated content, remain underexplored. This paper presents a mixed methods study examining how participants perceive different versions of PT recommendations. In our study, total of 598 IT professionals and cybersecurity students evaluated five recommendation variants for each of two vulnerability descriptions. These variants included versions written by an experienced penetration tester or generated by AI, and differed in the level of detail provided, as well as in the inclusion or exclusion of an explicit responsibility role tag. Study participants evaluated each recommendation using scaled ratings for three aspects: clarity, actionability, and confidence in applying the guidance. They also selected their preferred recommendation in pairwise comparisons and provided reasoning.
Results reveal some significant differences in perceived usability across the recommendation variants. Recommendations with clear steps and an organized structure were rated higher for clarity, actionability, and confidence. Presenting guidance as a sequence of actionable steps helps users quickly understand what to do, reduces ambiguity, and decreases the time needed to interpret and implement the recommendation. These results suggest that both human and AI authors of PT reports should prioritize structured presentation to make security guidance easier to understand and act on.
Scams, Fraud & Misinformation
"You Have Been Selected as the Winner": Characterizing User-Reported Scams on TikTok
Smirity Kaushik, The George Washington University; Kyle Beadle, University College London; Gauri Nayak, Cornell University; Madelyn Rose Sanfilippo, University of Illinois at Urbana–Champaign; Mainack Mondal, Indian Institute of Technology, Kharagpur; Yang Wang, University of Illinois at Urbana–Champaign; Sai Teja Peddinti, Google; Jingjie Li, University of Edinburgh; Yixin Zou, Max Planck Institute for Security and Privacy
Short-form video platforms (SVPs) such as TikTok have grown rapidly in popularity. While online scams have been extensively studied, the extent to which they take new forms on SVPs and the discourses around them remain understudied. Using TikTok as a case study, we analyzed 150 videos in which content creators reported scam experiences and offered anti-scam advice. We focus on how TikTok users (creators, followers, and commenters) discuss scams, rather than analyzing scams. Our analysis surfaces six types of scams, including creator impersonation and account badge verification scams that target TikTok’s influencer-follower ecosystem. Scammers also exploit platform-specific features (e.g., direct messaging or the "For You Page") to lure victims. In response, TikTok users share strategies to identify scammer profiles and communication cues, building community support through anti-scam advice. Based on our findings, we offer recommendations for systemizing platform support to combat scams and leveraging the influencer ecosystem to raise awareness.
Source-Level Disengagement: A Usable Security Defense Against Misinformation
Zaid Hakami, Florida International University and Jazan University; Yuzhou Feng and Bogdan Carbunar, Florida International University
Misinformation defenses on social media mostly focus on individual posts, and do not address repeated exposure from the same misinformation-sharing accounts. We frame source-level disengagement as a usable security defense and study how interface design shapes adoption and escalation of such controls. We built TimeLineR, an instrumented timeline environment, and used it to run a within-subject lab study with 29 participants. We design and compare three interface designs that vary the visibility and timing of source-level defensive actions, and introduce a Filter action that hides misinformation-labeled posts from an account while preserving other content. We find that actionability increases defensive uptake: in-context controls substantially increased adoption in our controlled setting. Users select relationship-appropriate defenses, with the introduced Filter acting as a stable middle-ground endpoint.
A Penny for Your Prompts: Experiments Detecting and Mitigating LLM Usage by Survey Respondents
Zane Xu and Nathan Malkin, New Jersey Institute of Technology
Large language models are increasingly used by participants on crowdsourcing platforms when responding to surveys, potentially undermining the validity of collected data. Our study aims to quantify the prevalence of this behavior and investigate methods to detect and prevent it. In a series of surveys (N = 250), we examined conditions such as platform choice, survey length, requests not to use AI, and disabling copy-paste functionality. We were able to identify distinct characteristics of LLM-assisted responses and found that their frequency varied widely, from under 10% on Prolific to over 80% on Mechanical Turk. Mitigation measures reduced LLM usage but did not necessarily improve data quality. No participants employed browser-use agents at the time of our survey, but we report on our own detection experiments. We recommend that researchers actively screen survey responses for LLM usage by recording and analyzing keystroke data and crafting instructions and questions aimed at AI.
SoK: Mapping Threats to Defenses in Online Survey Fraud
Shiza Ali, Wellington Esposito Barbosa, and Matthias Fassl, The George Washington University; Aditi Ganapathi and Jaron Mink, Arizona State University; Adam J. Aviv, The George Washington University
Online surveys and recruitment mechanisms, including crowd-sourcing platforms such as Prolific and MTurk, as well as social media-based recruitment, have become core infrastructure for human-subjects research. At the same time, their accessibility has made studies increasingly vulnerable to large-scale fraud. Prior work on survey fraud is extensive yet fragmented: different communities use inconsistent definitions, conflate inattentive responding with intentional or automated attacks, and deploy mitigation techniques without explicit threat models. This paper presents a Systematization of Knowledge (SoK) on online survey fraud. Based on a structured review of 124 papers across multiple disciplines that explicitly conceptualize, evaluate, or advance fraud-related mechanisms, we synthesize how fraud is conceptualized, where it arises across the survey lifecycle, and how detection and mitigation strategies are proposed and reported in this literature. Our analysis reveals three recurring gaps: (1) a lack of consistent and explicit definitions that distinguish inattentive responding from adversarial human and automated fraud; (2) systematic misalignment between fraud threats and reported defenses, particularly when recruitment-stage attacks are addressed only post hoc; and (3) inconsistent reporting practices that limit interpretability and reproducibility.
GenAI Privacy & Risks
AI've Got a Bad Feeling About This: A Privacy Threat Modeling Framework for GenAI
Qianying Liao, Jonah Bellemans, and Laurens Sion, DistriNet, KU Leuven; Xue Jiang, Dmitrii Usynin, and Xuebing Zhou, Huawei Heisenberg Research Center (Munich); Dimitri Van Landuyt, LIRIS, KU Leuven; Lieven Desmet and Wouter Joosen, DistriNet, KU Leuven
As generative AI (GenAI) systems become increasingly prevalent across various technological stacks, the question of how such systems handle sensitive and personal data flows becomes increasingly important. Specifically, both the ability to harness and process large swaths of information as well as their stochastic nature raise key concerns related to both security and privacy. Unfortunately, while some of the traditional security threat modeling can effectively identify certain violations, privacy-related issues are often overlooked. To respond to these challenges, we introduce a novel domain-specific privacy threat modeling framework to support the privacy threat analysis of GenAI-based applications. This framework is constructed through a two-pronged approach: (i) a systematic review of the emerging literature on GenAI privacy threats, and (ii) a case-driven application to a representative Chatbot system. These efforts yield a foundational GenAI privacy threat modeling framework built on LINDDUN. The new framework extends three out of the seven privacy threat types of LINDDUN and introduces 100 GenAI examples to the entire knowledge base. Its effectiveness is validated on an AI Agent system, which demonstrates that a comprehensive privacy analysis can be supported by the new framework.
Understanding U.S. Users' Security and Privacy Transparency Needs for Consumer-Facing Generative AI
Jiaxun Cao and Yu Dong, Duke University; Chunxi Zhan, Duke Kunshan University; Rithvik Neti, Duke University; Sai Teja Peddinti, Google; Pardis Emami-Naeini, Duke University
Users increasingly rely on consumer-facing generative AI (GenAI) for tasks ranging from everyday needs to sensitive use cases. Yet, it remains unclear whether and how existing security and privacy (S&P) communications in GenAI tools shape users’ adoption decisions and experiences. Understanding how users seek, interpret, and evaluate S&P information is critical for designing usable transparency that users can trust and act on. We conducted semi-structured interviews and design sessions with 21 U.S. GenAI users. Our findings suggest that available S&P information rarely drove initial adoption in practice, as participants often perceived it as incomplete, ineffective, or not credible. Instead, they relied on rough proxies (e.g., popularity) to infer S&P practices. After adoption, S&P uncertainty constrained participants’ willingness to use GenAI tools, especially for high-stakes purposes, and, in some cases, contributed to discontinued use. Participants therefore called for transparency that supports decisions and actions through trustworthy information (e.g., independent evaluations) and usable interfaces (e.g., on-demand disclosure). We categorize participants’ desired design practices into five dimensions to facilitate systematic future investigation into best practices. We conclude with recommendations for researchers, designers, and policymakers to improve S&P transparency in consumer-facing GenAI.
Maybe... I Don't Really Wanna Clone: Attitudes and Anticipated Harms of (Consensual) After-Death Cloning of One's Own and Voices of Entrusted Others
Jennifer Vander Loop and Filipo Sharevski, DePaul University; Lucy Davies and Partha Das Chowdhury, University of Bristol
The memories of people after their death could take on a "life of their own" with the help of GenAI voice cloning services. A "voice doppelgänger" thus could generate novel content that might cause harm in various ways for which the voice owner has no control over. To explore after-death voice cloning reckoning, we devised a two-stage mixed-method study. In the first stage, we developed ten after-death vignettes of (consensual) voice cloning through interviews with n1=25 participants. In the second stage, we evaluated the attitudes and anticipated harms of cloning one’s own and voices of entrusted others relative to each of these vignettes through a n2=919 participant survey. A "live forever" sentiment prevailed across all vignettes, though our participants were more likely to consensually use the voice of late family members than to give consent for their own voice to be used after their death. The anticipated abuse of the cloned after-death voice ranged from economic, cultural, social, and even harms to mind, reputation and physical well-being. We contextualize the threat of these harms towards a thanatosensitive design of security safeguards against rogue "voice doppelgänger" misuse.
Security Advice & Education
Who Can Actually Follow IT-Security Advice? Exploring the Usability of IT-Security Advice
Dennis Lawo, Jenny Hindrichs, and Gunnar Stevens, University of Siegen
Individuals are increasingly expected to rely on publicly available IT-security advice to protect themselves against cyberattacks and data breaches. However, little is known about how usable such IT-security advice is in practice for ordinary users. In this paper, we empirically evaluate the perceived usability of IT-security advice drawn from multiple publicly available sources. Based on an adapted System Usability Scale survey and qualitative analysis, we explored the perceived usability of IT security advice. Our results show substantial variation in usability across pieces of advice. User characteristics such as age, education, and self-assessed IT-knowledge are associated with differences in perceived usability, whereas traditional text-metrics show limited explanatory power. We discuss implications for the user-centred design and evaluation of IT-security advice. This work contributes to the usable security literature by providing empirical evidence on policy usability and by highlighting the role of user differences in the perceived usability of IT-security advice.
CyQured: Design, Development, and Empirical Evaluation of a Tabletop Game for Personal Cybersecurity Education
Utsho Das and Argha Pratim Saha, Shahjalal University of Science and Technology; Md Sadek Ferdous, BRAC University; Md Masum, Shahjalal University of Science and Technology; Farida Chowdhury, BRAC University
As reliance on connected home devices grows, human behavior remains a critical vulnerability that is often inadequately addressed by traditional, overly technical educational methods. To bridge this gap, we introduce CyQured, a tabletop board game that models a digital home ecosystem of 16 devices, enabling players to simulate 16 STRIDE-inspired attacks and defenses through structured gameplay. We evaluated CyQured with 50 university students, 13 with prior cybersecurity coursework and 37 without, including 14 with limited tabletop gaming experience. The evaluation was conducted using a mixed-methods approach that included knowledge assessments and user-centered evaluations, including the SUS, TAM, and NASA-TLX scales. While novices without domain or tabletop gaming experience reported higher cognitive load and borderline usability scores, knowledge gains were observed across all cohorts. Thematic analysis revealed that novices encountered initial accessibility barriers and elevated cognitive load due to unfamiliar cybersecurity terminology and gameplay mechanics; however, high engagement and social interaction motivated them to persist. Collectively, these findings suggest that game-based simulations can effectively engage diverse learners in understanding home cybersecurity, while also underscoring the need for accessible game design to support players without prior domain knowledge or tabletop gaming experience.
Nice to Know You're Not Alone: Co-designing Community-centered Online Safety and Privacy Education with Librarians
Tanisha Afnan, Sheza Naveed, Griffin Christie, and Jackie Hu, University of Michigan; Byron M. Lowens, Indiana University Indianapolis; Allison McDonald, Boston University; Florian Schaub, University of Michigan
Online safety and privacy literacy education is often offered in community settings and spaces, such as libraries, which allow for educational offerings tailored to community needs. However, such efforts are often resource constrained and place substantial burdens and responsibility on individual educators. We present findings from a co-design study with librarians offering online safety and privacy education in public and academic libraries to address their challenges in bringing these trainings to their communities. Fourteen participants across four co-design workshops engaged in guided prompts, discussions, and activities to develop innovative and sustainable ideas for enhancing community centered online safety and privacy education efforts. Educators also ideated promising directions for independent learning resources that could deepen patrons’ understanding of relevant topics on their own time after a training workshop. We discuss the implications of our findings for strengthening community-centered online safety and privacy education and resources.
Security versus Productivity: A Case Study of Email Management Practices with Job Task Analysis
Philip Shumway and Yi Ting Chua, University of Tulsa; Simon Parkin, Delft University of Technology
Employees routinely perform both productivity and security tasks when managing email, particularly in small organizations without dedicated cybersecurity teams. These responsibilities can create friction but are rarely analyzed within the structure of the job itself. This study applies task analysis (TA) to map security and productivity email tasks within a small U.S. publishing company. Using subject-expert task elicitation and employee ratings across five validated task attributes (time spent, criticality of error, difficulty to learn, feasibility, and importance), we examine how security and productivity tasks are structured within the role. Security tasks are rated higher in criticality and learning difficulty and lower in feasibility, but not more time-consuming overall. Correlation patterns show that task-attribute alignment shifts with external email volume, with high-volume employees exhibiting stronger coupling between security and productivity demands. These findings demonstrate how TA can identify structural sources of security–productivity tension and inform context-sensitive training and workflow design.