arXiv subjects

Patrick McDaniel

Publications and source records attributed to Patrick McDaniel.

At least 19 recordsLinked to original sources

arXiv2026

MESA: Prioritizing Vulnerable Communication Channels for Securing Multi-Agent Systems

Multi-agent systems (MAS) are increasingly used to automate complex, distributed workflows. However, their inter-agent communication channels introduce new attack surfaces that remain poorly understood and are difficult to defend against. In this paper, we address how defenders should prioritize limited security effort to protect vulnerable communication channels before attacks are observed. This is motivated by our observation that the channel-level attack impact is highly non-uniform: a single compromised edge can account for up to 75% of total attack success. We introduce Mesa, a label-free framework for proactively ranking which MAS edges are most security-critical -- that is, most likely to affect the system's decision if compromised. Mesa combines six graph-theoretic metrics and two dynamic probes (ablation and masking) without requiring attack traces. We evaluate Mesa against a dynamic misinformation attack pipeline across three diverse MAS scenarios, eight network topologies, and five open-source LLMs from Qwen, Llama, and Gemma families. Mesa rankings correlate strongly with empirical per-edge attack success rate, achieving mean Spearman $\rho=+0.60$ (peaking at $+0.73$). In resource-constrained defense deployment, monitoring the top 10% of Mesa-ranked edges intercepts about 3x the successful attacks as random allocation. We further test Mesa under varying attacker and defender models and LangGraph workflows and characterize its limits under adaptive attacks and high-redundancy graphs. Overall, our results show that edge-level risk in MAS is often concentrated and predictable, allowing proactive hardening of multi-agent infrastructures.

cs.CR↗

arXiv2026

Longitudinal Adoption and Deprecation of the Privacy Sandbox Web APIs

While several web actors have been trying to reduce web tracking for years, it remains unclear how to achieve both desirable levels of utility and privacy. In 2019, Google launched the Privacy Sandbox initiative to balance that trade-off and find privacy alternatives to common use cases such as advertising. Yet, in late 2025, Google canceled the project and deprecated most of the newly introduced APIs. Despite its end, the Privacy Sandbox represents a unique opportunity to learn about how the ecosystem reacted to the proposed changes. In this paper, we present a longitudinal measurement and analysis study of the Privacy Sandbox APIs to characterize their adoption and deprecation over the past seven years by different web actors. Leveraging historical HTTP Archive crawls and public Chrome telemetry data, we offer the largest study of its kind into the prevalence of each Privacy Sandbox feature, during their entire respective lifetime (5+ years for some), on popular websites (CrUX top 100k), and as experienced by Chrome users during their browsing journey. Our results showcase an adoption that remained limited and uneven across the years; only few web actors implemented very specific APIs, and in disparate manners. We motivate our interpretation of these results by considering the incentives (interest, resources, timeline, etc.) and risks (potential trade-offs, privacy violations, and legal exposure, etc.) for these actors. Finally, our analysis also yields a few actionable recommendations for the next generation of web privacy-enhancing technologies.

cs.CR↗

arXiv2026

Longitudinal Analyses of SAST Tools: A CodeQL Case Study

Open-source software (OSS) pipelines rely on automated static analysis tools to prevent the introduction of vulnerabilities in code. However, there is limited understanding of the efficacy of these tools across the OSS ecosystem over time. In this paper, we introduce a novel method to evaluate static application security testing (SAST) tools through longitudinal measurements and perform the largest academic study of CodeQL -- the most prevalent static analysis tool from GitHub -- on OSS codebases. We apply our apparatus on 114 versions of CodeQL over time on 3993 CVEs from 1622 repositories to measure key properties of the tool, culminating in more than 20 billion lines of code analyzed. First, we measure its effectiveness, i.e., its ability to detect vulnerabilities before they are fixed. Then, we determine whether these detections were actionable through two measures of the distance between findings and vulnerability location either over the entire codebase or within the vulnerable file. Finally, we study the stability of CodeQL by examining how vulnerability detections hold across versions and the evolution of CodeQL on the accuracy-precision trade-off. We find that CodeQL identifies a total of 171 CVEs, and that for 83 of them, a CodeQL version prior to the fix could detect it. Such detections are in general actionable if findings are triaged across files, as for 50% of the 171 detections, more than 50% of findings in the vulnerable file are located in the vulnerable location. Finally, we show that CVE detections are not monotonic across versions as 21 CVEs were no longer detected following a version change and 17 that were never redetected. Our study shows that using SAST tools is a matter of best practice as they prevent numerous vulnerabilities from being introduced, but that developers should be aware of changes that may leave blind spots in detections upon updates of the tool.

cs.CR↗

arXiv2026

The Role of Learning in Attacking ML-based Network Intrusion Detection

Machine learning (ML)-based network intrusion detection is susceptible to attacks that perturb malicious network flows to evade detection. Existing approaches to evaluating the robustness of these models rely on gradient-based optimization that are computationally expensive and restricted to differentiable model architectures. This limits their practicality for continuous, large-scale evaluation. To address this, we develop lightweight adversarial agents trained via reinforcement learning (RL) that decouples the cost of learning an evasion strategy from the cost of executing it. These agents learn offline to perturb malicious NetFlow records to evade surrogate intrusion detection models, encoding the resulting strategy into a reusable policy that requires no gradient computation at deployment. We evaluate our approach on four NetFlow datasets spanning enterprise, cloud, and IoT environments against diverse model architectures, including non-differentiable classifiers that gradient-based methods cannot evaluate directly. Agents achieve up to 58.1% attack success at 0.31ms per attack demonstrating up to 1,042X improvement in throughput (attack success per ms) over gradient-based methods. On non-differentiable targets, gradient-based methods lose over 59% of their effectiveness to surrogate transfer, while the RL agent evaluates these models directly at 29.8% attack success. We further conduct a comprehensive transferability study on ML-based intrusion detection, evaluating agent generalization across unseen model architectures and traffic distributions. Our results establish lightweight RL agents as a practical and scalable tool for continuous ML robustness evaluation across diverse network intrusion detection environments.

cs.CR↗

arXiv2025

Technical Report: The Need for a (Research) Sandstorm through the Privacy Sandbox

The Privacy Sandbox, launched in 2019, is a series of proposals from Google to reduce ``cross-site and cross-app tracking while helping to keep online content and services free for all''. Over the years, Google implemented, experimented, and deprecated some of these APIs into their own products (Chrome, Android, etc.) which raised concerns about the potential of these mechanisms to fundamentally disrupt the advertising, mobile, and web ecosystems. As a result, it is paramount for researchers to understand the consequences that these new technologies, and future ones, will have on billions of users if and when deployed. In this report, we outline our call for privacy, security, usability, and utility evaluations of these APIs, our efforts materialized through the creation and operation of Privacy Sandstorm (https://privacysandstorm.github.io); a research portal to systematically gather resources (overview, analyses, artifacts, etc.) about such proposals. We find that our inventory provides a better visibility and broader perspective on the research findings in that space than what Google lets show through official channels.

cs.CR↗

arXiv2025

It's a Feature, Not a Bug: Secure and Auditable State Rollback for Confidential Cloud Applications

Replay and rollback attacks threaten cloud application integrity by reintroducing authentic yet stale data through an untrusted storage interface to compromise application decision-making. Prior security frameworks mitigate these attacks by enforcing forward-only state transitions (state continuity) with hardware-backed mechanisms, but they categorically treat all rollback as malicious and thus preclude legitimate rollbacks used for operational recovery from corruption or misconfiguration. We present Rebound, a general-purpose security framework that preserves rollback protection while enabling policy-authorized legitimate rollbacks of application binaries, configuration, and data. Key to Rebound is a reference monitor that mediates state transitions, enforces authorization policy, guarantees atomicity of state updates and rollbacks, and emits a tamper-evident log that provides transparency to applications and auditors. We analyze Rebound's security properties and show through an application case study -- with software deployment workflows in GitLab CI -- that it enables robust control over binary, configuration, and raw data versioning with low end-to-end overhead.

cs.CR↗

arXiv2025

LibIHT: A Hardware-Based Approach to Efficient and Evasion-Resistant Dynamic Binary Analysis

Dynamic program analysis is invaluable for malware detection, debugging, and performance profiling. However, software-based instrumentation incurs high overhead and can be evaded by anti-analysis techniques. In this paper, we propose LibIHT, a hardware-assisted tracing framework that leverages on-CPU branch tracing features (Intel Last Branch Record and Branch Trace Store) to efficiently capture program control-flow with minimal performance impact. Our approach reconstructs control-flow graphs (CFGs) by collecting hardware generated branch execution data in the kernel, preserving program behavior against evasive malware. We implement LibIHT as an OS kernel module and user-space library, and evaluate it on both benign benchmark programs and adversarial anti-instrumentation samples. Our results indicate that LibIHT reduces runtime overhead by over 150x compared to Intel Pin (7x vs 1,053x slowdowns), while achieving high fidelity in CFG reconstruction (capturing over 99% of execution basic blocks and edges). Although this hardware-assisted approach sacrifices the richer semantic detail available from full software instrumentation by capturing only branch addresses, this trade-off is acceptable for many applications where performance and low detectability are paramount. Our findings show that hardware-based tracing captures control flow information significantly faster, reduces detection risk and performs dynamic analysis with minimal interference.

cs.CR↗

arXiv2025

A Practical Guideline and Taxonomy to LLVM's Control Flow Integrity

Memory corruption vulnerabilities remain one of the most severe threats to software security. They often allow attackers to achieve arbitrary code execution by redirecting a vulnerable program's control flow. While Control Flow Integrity (CFI) has gained traction to mitigate this exploitation path, developers are not provided with any direction on how to apply CFI to real-world software. In this work, we establish a taxonomy mapping LLVM's forward-edge CFI variants to memory corruption vulnerability classes, offering actionable guidance for developers seeking to deploy CFI incrementally in existing codebases. Based on the Top 10 Known Exploited Vulnerabilities (KEV) list, we identify four high-impact vulnerability categories and select one representative CVE for each. We evaluate LLVM's CFI against each CVE and explain why CFI blocks exploitation in two cases while failing in the other two, illustrating its potential and current limitations. Our findings support informed deployment decisions and provide a foundation for improving the practical use of CFI in production systems.

cs.CR↗

arXiv2025

ARMOR: Aligning Secure and Safe Large Language Models via Meticulous Reasoning

Large Language Models have shown impressive generative capabilities across diverse tasks, but their safety remains a critical concern. Existing post-training alignment methods, such as SFT and RLHF, reduce harmful outputs yet leave LLMs vulnerable to jailbreak attacks, especially advanced optimization-based ones. Recent system-2 approaches enhance safety by adding inference-time reasoning, where models assess potential risks before producing responses. However, we find these methods fail against powerful out-of-distribution jailbreaks, such as AutoDAN-Turbo and Adversarial Reasoning, which conceal malicious goals behind seemingly benign prompts. We observe that all jailbreaks ultimately aim to embed a core malicious intent, suggesting that extracting this intent is key to defense. To this end, we propose ARMOR, which introduces a structured three-step reasoning pipeline: (1) analyze jailbreak strategies from an external, updatable strategy library, (2) extract the core intent, and (3) apply policy-based safety verification. We further develop ARMOR-Think, which decouples safety reasoning from general reasoning to improve both robustness and utility. Evaluations on advanced optimization-based jailbreaks and safety benchmarks show that ARMOR achieves state-of-the-art safety performance, with an average harmful rate of 0.002 and an attack success rate of 0.06 against advanced optimization-based jailbreaks, far below other reasoning-based models. Moreover, ARMOR demonstrates strong generalization to unseen jailbreak strategies, reducing their success rate to zero. These highlight ARMOR's effectiveness in defending against OOD jailbreak attacks, offering a practical path toward secure and reliable LLMs.

cs.CR↗

arXiv2025

Doxing via the Lens: Revealing Location-related Privacy Leakage on Multi-modal Large Reasoning Models

Recent advances in multi-modal large reasoning models (MLRMs) have shown significant ability to interpret complex visual content. While these models enable impressive reasoning capabilities, they also introduce novel and underexplored privacy risks. In this paper, we identify a novel category of privacy leakage in MLRMs: Adversaries can infer sensitive geolocation information, such as a user's home address or neighborhood, from user-generated images, including selfies captured in private settings. To formalize and evaluate these risks, we propose a three-level visual privacy risk framework that categorizes image content based on contextual sensitivity and potential for location inference. We further introduce DoxBench, a curated dataset of 500 real-world images reflecting diverse privacy scenarios. Our evaluation across 11 advanced MLRMs and MLLMs demonstrates that these models consistently outperform non-expert humans in geolocation inference and can effectively leak location-related private information. This significantly lowers the barrier for adversaries to obtain users' sensitive geolocation information. We further analyze and identify two primary factors contributing to this vulnerability: (1) MLRMs exhibit strong reasoning capabilities by leveraging visual clues in combination with their internal world knowledge; and (2) MLRMs frequently rely on privacy-related visual clues for inference without any built-in mechanisms to suppress or avoid such usage. To better understand and demonstrate real-world attack feasibility, we propose GeoMiner, a collaborative attack framework that decomposes the prediction process into two stages: clue extraction and reasoning to improve geolocation performance while introducing a novel attack perspective. Our findings highlight the urgent need to reassess inference-time privacy risks in MLRMs to better protect users' sensitive information.

cs.CR↗

arXiv2025

Efficient Storage Integrity in Adversarial Settings

Storage integrity is essential to systems and applications that use untrusted storage (e.g., public clouds, end-user devices). However, known methods for achieving storage integrity either suffer from high (and often prohibitive) overheads or provide weak integrity guarantees. In this work, we demonstrate a hybrid approach to storage integrity that simultaneously reduces overhead while providing strong integrity guarantees. Our system, partially asynchronous integrity checking (PAC), allows disk write commitments to be deferred while still providing guarantees around read integrity. PAC delivers a 5.5X throughput and latency improvement over the state of the art, and 85% of the throughput achieved by non-integrity-assuring approaches. In this way, we show that untrusted storage can be used for integrity-critical workloads without meaningfully sacrificing performance.

cs.CR↗

arXiv2025

On the Robustness Tradeoff in Fine-Tuning

Fine-tuning has become the standard practice for adapting pre-trained models to downstream tasks. However, the impact on model robustness is not well understood. In this work, we characterize the robustness-accuracy trade-off in fine-tuning. We evaluate the robustness and accuracy of fine-tuned models over 6 benchmark datasets and 7 different fine-tuning strategies. We observe a consistent trade-off between adversarial robustness and accuracy. Peripheral updates such as BitFit are more effective for simple tasks -- over 75% above the average measured by the area under the Pareto frontiers on CIFAR-10 and CIFAR-100. In contrast, fine-tuning information-heavy layers, such as attention layers via Compacter, achieves a better Pareto frontier on more complex tasks -- 57.5% and 34.6% above the average on Caltech-256 and CUB-200, respectively. Lastly, we observe that the robustness of fine-tuning against out-of-distribution data closely tracks accuracy. These insights emphasize the need for robustness-aware fine-tuning to ensure reliable real-world deployments.

cs.LG↗

arXiv2025

Adversarial Agents: Black-Box Evasion Attacks with Reinforcement Learning

Attacks on machine learning models have been extensively studied through stateless optimization. In this paper, we demonstrate how a reinforcement learning (RL) agent can learn a new class of attack algorithms that generate adversarial samples. Unlike traditional adversarial machine learning (AML) methods that craft adversarial samples independently, our RL-based approach retains and exploits past attack experience to improve the effectiveness and efficiency of future attacks. We formulate adversarial sample generation as a Markov Decision Process and evaluate RL's ability to (a) learn effective and efficient attack strategies and (b) compete with state-of-the-art AML. On two image classification benchmarks, our agent increases attack success rate by up to 13.2% and decreases the average number of victim model queries per attack by up to 16.9% from the start to the end of training. In a head-to-head comparison with state-of-the-art image attacks, our approach enables an adversary to generate adversarial samples with 17% more success on unseen inputs post-training. From a security perspective, this work demonstrates a powerful new attack vector that uses RL to train agents that attack ML models efficiently and at scale.

cs.CR↗

arXiv2025

Alignment and Adversarial Robustness: Are More Human-Like Models More Secure?

A small but growing body of work has shown that machine learning models which better align with human vision have also exhibited higher robustness to adversarial examples, raising the question: can human-like perception make models more secure? If true generally, such mechanisms would offer new avenues toward robustness. In this work, we conduct a large-scale empirical analysis to systematically investigate the relationship between representational alignment and adversarial robustness. We evaluate 114 models spanning diverse architectures and training paradigms, measuring their neural and behavioral alignment and engineering task performance across 105 benchmarks as well as their adversarial robustness via AutoAttack. Our findings reveal that while average alignment and robustness exhibit a weak overall correlation, specific alignment benchmarks serve as strong predictors of adversarial robustness, particularly those that measure selectivity toward texture or shape. These results suggest that different forms of alignment play distinct roles in model robustness, motivating further investigation into how alignment-driven approaches can be leveraged to build more secure and perceptually-grounded vision models.

cs.CV↗

arXiv2025

Deserialization Gadget Chains are not a Pathological Problem in Android:an In-Depth Study of Java Gadget Chains in AOSP

Inter-app communication is a mandatory and security-critical functionality of operating systems, such as Android. On the application level, Android implements this facility through Intents, which can also transfer non-primitive objects using Java's Serializable API. However, the Serializable API has a long history of deserialization vulnerabilities, specifically deserialization gadget chains. Research endeavors have been heavily directed towards the detection of deserialization gadget chains on the Java platform. Yet, there is little knowledge about the existence of gadget chains within the Android platform. We aim to close this gap by searching gadget chains in the Android SDK, Android's official development libraries, as well as frequently used third-party libraries. To handle this large dataset, we design a gadget chain detection tool optimized for soundness and efficiency. In a benchmark on the full Ysoserial dataset, it achieves similarly sound results to the state-of-the-art in significantly less time. Using our tool, we first show that the Android SDK contains almost the same trampoline gadgets as the Java Class Library. We also find that one can trigger Java native serialization through Android's Parcel API. Yet, running our tool on the Android SDK and 1,200 Android dependencies, in combination with a comprehensive sink dataset, yields no security-critical gadget chains. This result opposes the general notion of Java deserialization gadget chains being a widespread problem. Instead, the issue appears to be more nuanced, and we provide a perspective on where to direct further research.

cs.CR↗

arXiv2025

Targeting Alignment: Extracting Safety Classifiers of Aligned LLMs

Alignment in large language models (LLMs) is used to enforce guidelines such as safety. Yet, alignment fails in the face of jailbreak attacks that modify inputs to induce unsafe outputs. In this paper, we introduce and evaluate a new technique for jailbreak attacks. We observe that alignment embeds a safety classifier in the LLM responsible for deciding between refusal and compliance, and seek to extract an approximation of this classifier: a surrogate classifier. To this end, we build candidate classifiers from subsets of the LLM. We first evaluate the degree to which candidate classifiers approximate the LLM's safety classifier in benign and adversarial settings. Then, we attack the candidates and measure how well the resulting adversarial inputs transfer to the LLM. Our evaluation shows that the best candidates achieve accurate agreement (an F1 score above 80%) using as little as 20% of the model architecture. Further, we find that attacks mounted on the surrogate classifiers can be transferred to the LLM with high success. For example, a surrogate using only 50% of the Llama 2 model achieved an attack success rate (ASR) of 70% with half the memory footprint and runtime -- a substantial improvement over attacking the LLM directly, where we only observed a 22% ASR. These results show that extracting surrogate classifiers is an effective and efficient means for modeling (and therein addressing) the vulnerability of aligned models to jailbreaking attacks. The code is available at https://github.com/jcnf0/targeting-alignment.

cs.CR↗

arXiv2024

Err on the Side of Texture: Texture Bias on Real Data

Bias significantly undermines both the accuracy and trustworthiness of machine learning models. To date, one of the strongest biases observed in image classification models is texture bias-where models overly rely on texture information rather than shape information. Yet, existing approaches for measuring and mitigating texture bias have not been able to capture how textures impact model robustness in real-world settings. In this work, we introduce the Texture Association Value (TAV), a novel metric that quantifies how strongly models rely on the presence of specific textures when classifying objects. Leveraging TAV, we demonstrate that model accuracy and robustness are heavily influenced by texture. Our results show that texture bias explains the existence of natural adversarial examples, where over 90% of these samples contain textures that are misaligned with the learned texture of their true label, resulting in confident mispredictions.

cs.CV↗

arXiv2024

AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs

In this paper, we propose AutoDAN-Turbo, a black-box jailbreak method that can automatically discover as many jailbreak strategies as possible from scratch, without any human intervention or predefined scopes (e.g., specified candidate strategies), and use them for red-teaming. As a result, AutoDAN-Turbo can significantly outperform baseline methods, achieving a 74.3% higher average attack success rate on public benchmarks. Notably, AutoDAN-Turbo achieves an 88.5 attack success rate on GPT-4-1106-turbo. In addition, AutoDAN-Turbo is a unified framework that can incorporate existing human-designed jailbreak strategies in a plug-and-play manner. By integrating human-designed strategies, AutoDAN-Turbo can even achieve a higher attack success rate of 93.4 on GPT-4-1106-turbo.

cs.CR↗