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.

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

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.

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 formally prove 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.

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

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.

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

Merkle hash trees are the standard method to protect the integrity and freshness of stored data. However, hash trees introduce additional compute and I/O costs on the I/O critical path, and prior efforts have not fully characterized these costs. In this paper, we quantify performance overheads of storage-level hash trees in realistic settings. We then design an optimized tree structure called Dynamic Merkle Trees (DMTs) based on an analysis of root causes of overheads. DMTs exploit patterns in workloads to deliver up to a 2.2x throughput and latency improvement over the state of the art. Our novel approach provides a promising new direction to achieve integrity guarantees in storage efficiently and at scale.

Cloud file systems offer organizations a scalable and reliable file storage solution. However, cloud file systems have become prime targets for adversaries, and traditional designs are not equipped to protect organizations against the myriad of attacks that may be initiated by a malicious cloud provider, co-tenant, or end-client. Recently proposed designs leveraging cryptographic techniques and trusted execution environments (TEEs) still force organizations to make undesirable trade-offs, consequently leading to either security, functional, or performance limitations. In this paper, we introduce BFS, a cloud file system that leverages the security capabilities provided by TEEs to bootstrap new security protocols that deliver strong security guarantees, high-performance, and a transparent POSIX-like interface to clients. BFS delivers stronger security guarantees and up to a 2.5X speedup over a state-of-the-art secure file system. Moreover, compared to the industry standard NFS, BFS achieves up to 2.2X speedups across micro-benchmarks and incurs <1X overhead for most macro-benchmark workloads. BFS demonstrates a holistic cloud file system design that does not sacrifice an organizations' security yet can embrace all of the functional and performance advantages of outsourcing.

The powerful capabilities of LLMs stem from their rich training data and high-quality labeled datasets, making the training of strong LLMs a resource-intensive process, which elevates the importance of IP protection for such LLMs. Compared to gathering high-quality labeled data, directly sampling outputs from these fully trained LLMs as training data presents a more cost-effective approach. This practice—where a suspect model is fine-tuned using high-quality data derived from these LLMs, thereby gaining capabilities similar to the target model—can be seen as a form of IP infringement against the original LLM. In recent years, LLM watermarks have been proposed and used to detect whether a text is AI-generated. Intuitively, if data sampled from a watermarked LLM is used for training, the resulting model would also be influenced by this watermark. This raises the question: can we directly use such watermarks to detect IP infringement of LLMs? In this paper, we explore the potential of LLM watermarks for detecting model infringement. We find that there are two issues with direct detection: (1) The queries used to sample output from the suspect LLM have a significant impact on detectability. (2) The watermark that is easily learned by LLMs exhibits instability regarding the watermark's hash key during detection. To address these issues, we propose LIDet, a detection method that leverages available anchor LLMs to select suitable queries for sampling from the suspect LLM. Additionally, it adapts the detection threshold to mitigate detection failures caused by different hash keys. To demonstrate the effectiveness of this approach, we construct a challenging model set containing multiple suspect LLMs on which direct detection methods struggle to yield effective results. Our method achieves over 90\% accuracy in distinguishing between infringing and clean models, demonstrating the feasibility of using LLM watermarks to detect LLM IP infringement.

Jailbreak attacks serve as essential red-teaming tools, proactively assessing whether LLMs can behave responsibly and safely in adversarial environments. Despite diverse strategies (e.g., cipher, low-resource language, persuasions, and so on) that have been proposed and shown success, these strategies are still manually designed, limiting their scope and effectiveness as a red-teaming tool. 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.

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

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.

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.

Texture data serves as a valuable tool for interpreting the high-level features models learn, uncovering biases, and identifying security vulnerabilities. However, works in this space have been limited by small texture datasets and synthesis methods that struggle to scale in the diversity and specificity required for these tasks. In this work, we introduce an extensible methodology for generating high-quality, diverse texture images, which we use to create the Prompted Textures Dataset (PTD), a new texture dataset spanning 246,285 images across 56 texture classes. Our comparison against real texture data demonstrates that PTD is more diverse while maintaining quality. Additionally, human evaluations confirm that every stage in our methodology enhances texture quality, yielding a 3.4% increase in quality and a 4.5% increase in representativeness overall. Our dataset is available for download at https://zenodo.org/records/15359142.

Honeypots are essential to measuring and defending network services, but the shift to public clouds has eroded their effectiveness visibility over the past decade. By adapting to new deployment realities, we can regain this visibility and defend real-world deployments.

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.

Public clouds necessitate dynamic resource allocation and sharing. However, the dynamic allocation of IP addresses can be abused by adversaries to source malicious traffic, bypass rate limiting systems, and even capture traffic intended for other cloud tenants. As a result, both the cloud provider and their customers are put at risk, and defending against these threats requires a rigorous analysis of tenant behavior, adversarial strategies, and cloud provider policies. In this paper, we develop a practical defense for IP address allocation through such an analysis. We first develop a statistical model of cloud tenant deployment behavior based on literature and measurement of deployed systems. Through this, we analyze IP allocation policies under existing and novel threat models. In response to our stronger proposed threat model, we design IP scan segmentation, an IP allocation policy that protects the address pool against adversarial scanning even when an adversary is not limited by number of cloud tenants. Through empirical evaluation on both synthetic and real-world allocation traces, we show that IP scan segmentation reduces adversaries' ability to rapidly allocate addresses, protecting both address space reputation and cloud tenant data. In this way, we show that principled analysis and implementation of cloud IP address allocation can lead to substantial security gains for tenants and their users.

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

Internet of Things (IoT) platforms possess several properties that can potentially jeopardize the safety and security of users, such as distributed deployment, processing sensitive information, and exposed networks. This issue is further exacerbated by the physical nature of IoT allowing devices to compromise the confidentiality and integrity of both persons and property.Some IoT problems can be addressed by taking preventative measures before they happen, which requires making predictions about the future. We design ProvPredictor to gather provenance information in order to train a model to make predictions about potentially unsafe behaviors in the future. To demonstrate the effectiveness of ProvPredictor, we create a realistic deployment using IFTTT, a web-based IoT platform, using the most common IFTTT compatible services and applications in a home environment. We additionally use Agriculture datasets to show how ProvPredictor can operate in industrial systems. We train ProvPredictor on the generated provenance data and find that ProvPredictor can predict violations with over 90