arXiv subjects

Kassem Fawaz

Publications and source records attributed to Kassem Fawaz.

At least 37 records · Page 2Linked to original sources

arXiv2025

Do Large Code Models Understand Programming Concepts? Counterfactual Analysis for Code Predicates

Large Language Models' success on text generation has also made them better at code generation and coding tasks. While a lot of work has demonstrated their remarkable performance on tasks such as code completion and editing, it is still unclear as to why. We help bridge this gap by exploring to what degree auto-regressive models understand the logical constructs of the underlying programs. We propose Counterfactual Analysis for Programming Concept Predicates (CACP) as a counterfactual testing framework to evaluate whether Large Code Models understand programming concepts. With only black-box access to the model, we use CACP to evaluate ten popular Large Code Models for four different programming concepts. Our findings suggest that current models lack understanding of concepts such as data flow and control flow.

cs.SE↗

arXiv2024

Evaluating the Impact of Warning Modalities and False Alarms in Pedestrian Crossing Alert System

With the steadily increasing pedestrian fatalities, pedestrian safety is a growing concern, especially in urban environments. Advanced Driver Assistance Systems (ADAS) have been developed to mitigate road user risks by predicting potential pedestrian crossings and issuing timely driver alerts. However, there is limited understanding of how drivers respond to different modalities of alerts, particularly in the presence of false alarms. In this study, we utilized a full-scale driving simulator to compare the effectiveness of different alert modalities, audio-visual (AV), visual-tactile (VT), and audio-visual-tactile (AVT), in alerting drivers to various pedestrian jaywalking events. Our findings reveal that, compared to no alerts, multimodal alerts significantly increased the number of vehicles stopped for pedestrians and the distance to pedestrians when stopped. However, the false alarms negatively impacted driver trust, with some drivers exhibiting excessive caution, alert fatigue and anxiety, even including one instance where a driver fully stopped when no pedestrian was present.

cs.HC↗

arXiv2024

Reliable Heading Tracking for Pedestrian Road Crossing Prediction Using Commodity Devices

Pedestrian heading tracking enables applications in pedestrian navigation, traffic safety, and accessibility. Previous works, using inertial sensor fusion or machine learning, are limited in that they assume the phone is fixed in specific orientations, hindering their generalizability. We propose a new heading tracking algorithm, the Orientation-Heading Alignment (OHA), which leverages a key insight: people tend to carry smartphones in certain ways due to habits, such as swinging them while walking. For each smartphone attitude during this motion, OHA maps the smartphone orientation to the pedestrian heading and learns such mappings efficiently from coarse headings and smartphone orientations. To anchor our algorithm in a practical scenario, we apply OHA to a challenging task: predicting when pedestrians are about to cross the road to improve road user safety. In particular, using 755 hours of walking data collected since 2020 from 60 individuals, we develop a lightweight model that operates in real-time on commodity devices to predict road crossings. Our evaluation shows that OHA achieves 3.4 times smaller heading errors across nine scenarios than existing methods. Furthermore, OHA enables the early and accurate detection of pedestrian crossing behavior, issuing crossing alerts 0.35 seconds, on average, before pedestrians enter the road range.

eess.SP↗

arXiv2024

PolicyLR: A Logic Representation For Privacy Policies

Privacy policies are crucial in the online ecosystem, defining how services handle user data and adhere to regulations such as GDPR and CCPA. However, their complexity and frequent updates often make them difficult for stakeholders to understand and analyze. Current automated analysis methods, which utilize natural language processing, have limitations. They typically focus on individual tasks and fail to capture the full context of the policies. We propose PolicyLR, a new paradigm that offers a comprehensive machine-readable representation of privacy policies, serving as an all-in-one solution for multiple downstream tasks. PolicyLR converts privacy policies into a machine-readable format using valuations of atomic formulae, allowing for formal definitions of tasks like compliance and consistency. We have developed a compiler that transforms unstructured policy text into this format using off-the-shelf Large Language Models (LLMs). This compiler breaks down the transformation task into a two-stage translation and entailment procedure. This procedure considers the full context of the privacy policy to infer a complex formula, where each formula consists of simpler atomic formulae. The advantage of this model is that PolicyLR is interpretable by design and grounded in segments of the privacy policy. We evaluated the compiler using ToS;DR, a community-annotated privacy policy entailment dataset. Utilizing open-source LLMs, our compiler achieves precision and recall values of 0.91 and 0.88, respectively. Finally, we demonstrate the utility of PolicyLR in three privacy tasks: Policy Compliance, Inconsistency Detection, and Privacy Comparison Shopping.

cs.CR↗

arXiv2024

A Picture is Worth 500 Labels: A Case Study of Demographic Disparities in Local Machine Learning Models for Instagram and TikTok

Mobile apps have embraced user privacy by moving their data processing to the user's smartphone. Advanced machine learning (ML) models, such as vision models, can now locally analyze user images to extract insights that drive several functionalities. Capitalizing on this new processing model of locally analyzing user images, we analyze two popular social media apps, TikTok and Instagram, to reveal (1) what insights vision models in both apps infer about users from their image and video data and (2) whether these models exhibit performance disparities with respect to demographics. As vision models provide signals for sensitive technologies like age verification and facial recognition, understanding potential biases in these models is crucial for ensuring that users receive equitable and accurate services. We develop a novel method for capturing and evaluating ML tasks in mobile apps, overcoming challenges like code obfuscation, native code execution, and scalability. Our method comprises ML task detection, ML pipeline reconstruction, and ML performance assessment, specifically focusing on demographic disparities. We apply our methodology to TikTok and Instagram, revealing significant insights. For TikTok, we find issues in age and gender prediction accuracy, particularly for minors and Black individuals. In Instagram, our analysis uncovers demographic disparities in the extraction of over 500 visual concepts from images, with evidence of spurious correlations between demographic features and certain concepts.

cs.LG↗

arXiv2024

PRP: Propagating Universal Perturbations to Attack Large Language Model Guard-Rails

Large language models (LLMs) are typically aligned to be harmless to humans. Unfortunately, recent work has shown that such models are susceptible to automated jailbreak attacks that induce them to generate harmful content. More recent LLMs often incorporate an additional layer of defense, a Guard Model, which is a second LLM that is designed to check and moderate the output response of the primary LLM. Our key contribution is to show a novel attack strategy, PRP, that is successful against several open-source (e.g., Llama 2) and closed-source (e.g., GPT 3.5) implementations of Guard Models. PRP leverages a two step prefix-based attack that operates by (a) constructing a universal adversarial prefix for the Guard Model, and (b) propagating this prefix to the response. We find that this procedure is effective across multiple threat models, including ones in which the adversary has no access to the Guard Model at all. Our work suggests that further advances are required on defenses and Guard Models before they can be considered effective.

cs.CR↗

arXiv2024

I see an IC: A Mixed-Methods Approach to Study Human Problem-Solving Processes in Hardware Reverse Engineering

Trust in digital systems depends on secure hardware, often assured through Hardware Reverse Engineering (HRE). This work develops methods for investigating human problem-solving processes in HRE, an underexplored yet critical aspect. Since reverse engineers rely heavily on visual information, eye tracking holds promise for studying their cognitive processes. To gain further insights, we additionally employ verbal thought protocols during and immediately after HRE tasks: Concurrent and Retrospective Think Aloud. We evaluate the combination of eye tracking and Think Aloud with 41 participants in an HRE simulation. Eye tracking accurately identifies fixations on individual circuit elements and highlights critical components. Based on two use cases, we demonstrate that eye tracking and Think Aloud can complement each other to improve data quality. Our methodological insights can inform future studies in HRE, a specific setting of human-computer interaction, and in other problem-solving settings involving misleading or missing information.

cs.HC↗

arXiv2023

Limitations of Face Image Generation

Text-to-image diffusion models have achieved widespread popularity due to their unprecedented image generation capability. In particular, their ability to synthesize and modify human faces has spurred research into using generated face images in both training data augmentation and model performance assessments. In this paper, we study the efficacy and shortcomings of generative models in the context of face generation. Utilizing a combination of qualitative and quantitative measures, including embedding-based metrics and user studies, we present a framework to audit the characteristics of generated faces conditioned on a set of social attributes. We applied our framework on faces generated through state-of-the-art text-to-image diffusion models. We identify several limitations of face image generation that include faithfulness to the text prompt, demographic disparities, and distributional shifts. Furthermore, we present an analytical model that provides insights into how training data selection contributes to the performance of generative models.

cs.CV↗

arXiv2023

Human-Producible Adversarial Examples

Visual adversarial examples have so far been restricted to pixel-level image manipulations in the digital world, or have required sophisticated equipment such as 2D or 3D printers to be produced in the physical real world. We present the first ever method of generating human-producible adversarial examples for the real world that requires nothing more complicated than a marker pen. We call them $\textbf{adversarial tags}$. First, building on top of differential rendering, we demonstrate that it is possible to build potent adversarial examples with just lines. We find that by drawing just $4$ lines we can disrupt a YOLO-based model in $54.8\%$ of cases; increasing this to $9$ lines disrupts $81.8\%$ of the cases tested. Next, we devise an improved method for line placement to be invariant to human drawing error. We evaluate our system thoroughly in both digital and analogue worlds and demonstrate that our tags can be applied by untrained humans. We demonstrate the effectiveness of our method for producing real-world adversarial examples by conducting a user study where participants were asked to draw over printed images using digital equivalents as guides. We further evaluate the effectiveness of both targeted and untargeted attacks, and discuss various trade-offs and method limitations, as well as the practical and ethical implications of our work. The source code will be released publicly.

cs.CV↗

arXiv2023

Stateful Defenses for Machine Learning Models Are Not Yet Secure Against Black-box Attacks

Recent work has proposed stateful defense models (SDMs) as a compelling strategy to defend against a black-box attacker who only has query access to the model, as is common for online machine learning platforms. Such stateful defenses aim to defend against black-box attacks by tracking the query history and detecting and rejecting queries that are "similar" and thus preventing black-box attacks from finding useful gradients and making progress towards finding adversarial attacks within a reasonable query budget. Recent SDMs (e.g., Blacklight and PIHA) have shown remarkable success in defending against state-of-the-art black-box attacks. In this paper, we show that SDMs are highly vulnerable to a new class of adaptive black-box attacks. We propose a novel adaptive black-box attack strategy called Oracle-guided Adaptive Rejection Sampling (OARS) that involves two stages: (1) use initial query patterns to infer key properties about an SDM's defense; and, (2) leverage those extracted properties to design subsequent query patterns to evade the SDM's defense while making progress towards finding adversarial inputs. OARS is broadly applicable as an enhancement to existing black-box attacks - we show how to apply the strategy to enhance six common black-box attacks to be more effective against current class of SDMs. For example, OARS-enhanced versions of black-box attacks improved attack success rate against recent stateful defenses from almost 0% to to almost 100% for multiple datasets within reasonable query budgets.

cs.CR↗

arXiv2023

Exposing and Addressing Security Vulnerabilities in Browser Text Input Fields

In this work, we perform a comprehensive analysis of the security of text input fields in web browsers. We find that browsers' coarse-grained permission model violates two security design principles: least privilege and complete mediation. We further uncover two vulnerabilities in input fields, including the alarming discovery of passwords in plaintext within the HTML source code of the web page. To demonstrate the real-world impact of these vulnerabilities, we design a proof-of-concept extension, leveraging techniques from static and dynamic code injection attacks to bypass the web store review process. Our measurements and case studies reveal that these vulnerabilities are prevalent across various websites, with sensitive user information, such as passwords, exposed in the HTML source code of even high-traffic sites like Google and Cloudflare. We find that a significant percentage (12.5\%) of extensions possess the necessary permissions to exploit these vulnerabilities and identify 190 extensions that directly access password fields. Finally, we propose two countermeasures to address these risks: a bolt-on JavaScript package for immediate adoption by website developers allowing them to protect sensitive input fields, and a browser-level solution that alerts users when an extension accesses sensitive input fields. Our research highlights the urgent need for improved security measures to protect sensitive user information online.

cs.CR↗

arXiv2023

D4: Detection of Adversarial Diffusion Deepfakes Using Disjoint Ensembles

Detecting diffusion-generated deepfake images remains an open problem. Current detection methods fail against an adversary who adds imperceptible adversarial perturbations to the deepfake to evade detection. In this work, we propose Disjoint Diffusion Deepfake Detection (D4), a deepfake detector designed to improve black-box adversarial robustness beyond de facto solutions such as adversarial training. D4 uses an ensemble of models over disjoint subsets of the frequency spectrum to significantly improve adversarial robustness. Our key insight is to leverage a redundancy in the frequency domain and apply a saliency partitioning technique to disjointly distribute frequency components across multiple models. We formally prove that these disjoint ensembles lead to a reduction in the dimensionality of the input subspace where adversarial deepfakes lie, thereby making adversarial deepfakes harder to find for black-box attacks. We then empirically validate the D4 method against several black-box attacks and find that D4 significantly outperforms existing state-of-the-art defenses applied to diffusion-generated deepfake detection. We also demonstrate that D4 provides robustness against adversarial deepfakes from unseen data distributions as well as unseen generative techniques.

cs.LG↗

arXiv2023

Theoretically Principled Trade-off for Stateful Defenses against Query-Based Black-Box Attacks

Adversarial examples threaten the integrity of machine learning systems with alarming success rates even under constrained black-box conditions. Stateful defenses have emerged as an effective countermeasure, detecting potential attacks by maintaining a buffer of recent queries and detecting new queries that are too similar. However, these defenses fundamentally pose a trade-off between attack detection and false positive rates, and this trade-off is typically optimized by hand-picking feature extractors and similarity thresholds that empirically work well. There is little current understanding as to the formal limits of this trade-off and the exact properties of the feature extractors/underlying problem domain that influence it. This work aims to address this gap by offering a theoretical characterization of the trade-off between detection and false positive rates for stateful defenses. We provide upper bounds for detection rates of a general class of feature extractors and analyze the impact of this trade-off on the convergence of black-box attacks. We then support our theoretical findings with empirical evaluations across multiple datasets and stateful defenses.

cs.LG↗

arXiv2023

Unpacking Privacy Labels: A Measurement and Developer Perspective on Google's Data Safety Section

Google has mandated developers to use Data Safety Sections (DSS) to increase transparency in data collection and sharing practices. In this paper, we present a comprehensive analysis of Google's Data Safety Section (DSS) using both quantitative and qualitative methods. We conduct the first large-scale measurement study of DSS using apps from Android Play store (n=1.1M). We find that there are internal inconsistencies within the reported practices. We also find trends of both over and under-reporting practices in the DSSs. Next, we conduct a longitudinal study of DSS to explore how the reported practices evolve over time, and find that the developers are still adjusting their practices. To contextualize these findings, we conduct a developer study, uncovering the process that app developers undergo when working with DSS. We highlight the challenges faced and strategies employed by developers for DSS submission, and the factors contributing to changes in the DSS. Our research contributes valuable insights into the complexities of implementing and maintaining privacy labels, underlining the need for better resources, tools, and guidelines to aid developers. This understanding is crucial as the accuracy and reliability of privacy labels directly impact their effectiveness.

cs.CY↗

arXiv2023

How Interactions Influence Users' Security Perception of Virtual Reality Authentication?

Users readily embrace the rapid advancements in virtual reality (VR) technology within various everyday contexts, such as gaming, social interactions, shopping, and commerce. In order to facilitate transactions and payments, VR systems require access to sensitive user data and assets, which consequently necessitates user authentication. However, there exists a limited understanding regarding how users' unique experiences in VR contribute to their perception of security. In our study, we adopt a research approach known as ``technology probe'' to investigate this question. Specifically, we have designed probes that explore the authentication process in VR, aiming to elicit responses from participants from multiple perspectives. These probes were seamlessly integrated into the routine payment system of a VR game, thereby establishing an organic study environment. Through qualitative analysis, we uncover the interplay between participants' interaction experiences and their security perception. Remarkably, despite encountering unique challenges in usability during VR interactions, our participants found the intuitive virtualized authentication process beneficial and thoroughly enjoyed the immersive nature of VR. Furthermore, we observe how these interaction experiences influence participants' ability to transfer their pre-existing understanding of authentication into VR, resulting in a discrepancy in perceived security. Moreover, we identify users' conflicting expectations, encompassing their desire for an enjoyable VR experience alongside the assurance of secure VR authentication. Building upon our findings, we propose recommendations aimed at addressing these expectations and alleviating potential conflicts.

cs.CR↗

arXiv2023

Tubes Among Us: Analog Attack on Automatic Speaker Identification

Recent years have seen a surge in the popularity of acoustics-enabled personal devices powered by machine learning. Yet, machine learning has proven to be vulnerable to adversarial examples. A large number of modern systems protect themselves against such attacks by targeting artificiality, i.e., they deploy mechanisms to detect the lack of human involvement in generating the adversarial examples. However, these defenses implicitly assume that humans are incapable of producing meaningful and targeted adversarial examples. In this paper, we show that this base assumption is wrong. In particular, we demonstrate that for tasks like speaker identification, a human is capable of producing analog adversarial examples directly with little cost and supervision: by simply speaking through a tube, an adversary reliably impersonates other speakers in eyes of ML models for speaker identification. Our findings extend to a range of other acoustic-biometric tasks such as liveness detection, bringing into question their use in security-critical settings in real life, such as phone banking.

cs.LG↗

arXiv2023

The Overview of Privacy Labels and their Compatibility with Privacy Policies

Privacy nutrition labels provide a way to understand an app's key data practices without reading the long and hard-to-read privacy policies. Recently, the app distribution platforms for iOS(Apple) and Android(Google) have implemented mandates requiring app developers to fill privacy nutrition labels highlighting their privacy practices such as data collection, data sharing, and security practices. These privacy labels contain very fine-grained information about the apps' data practices such as the data types and purposes associated with each data type. This provides us with a unique vantage point from which we can understand apps' data practices at scale.

cs.CY↗

arXiv2022

SkillFence: A Systems Approach to Practically Mitigating Voice-Based Confusion Attacks

Voice assistants are deployed widely and provide useful functionality. However, recent work has shown that commercial systems like Amazon Alexa and Google Home are vulnerable to voice-based confusion attacks that exploit design issues. We propose a systems-oriented defense against this class of attacks and demonstrate its functionality for Amazon Alexa. We ensure that only the skills a user intends execute in response to voice commands. Our key insight is that we can interpret a user's intentions by analyzing their activity on counterpart systems of the web and smartphones. For example, the Lyft ride-sharing Alexa skill has an Android app and a website. Our work shows how information from counterpart apps can help reduce dis-ambiguities in the skill invocation process. We build SkilIFence, a browser extension that existing voice assistant users can install to ensure that only legitimate skills run in response to their commands. Using real user data from MTurk (N = 116) and experimental trials involving synthetic and organic speech, we show that SkillFence provides a balance between usability and security by securing 90.83% of skills that a user will need with a False acceptance rate of 19.83%.

cs.CR↗