The recovery of types from stripped binaries is a key to exact decompilation, yet its practical realization suffers. For composite structures in particular, both layout and semantic fidelity are required to enable end-to-end reconstruction. Many existing approaches either synthesize layouts or infer names post-hoc, which weakens downstream usability. This is further aggravated by an excessive runtime overhead that is especially prohibitive in automated environments. We present XTRIDE, an improved n-gram-based approach that focuses on practicality: highly optimized throughput and actionable confidence scores allow for deployment in automated pipelines. When compared to the state of the art in struct recovery, our method achieves comparable performance while being between 70 and 2300 times faster. As our inference is grounded in real-world types, we achieve the highest ratio of fully-correct struct layouts. With an optimized training regimen, our model outperforms the current state of the art on the DIRT dataset by 5.09 percentage points, achieving 90.15

The behavior of LLMs does not depend solely on the model itself. Components of the inference system, such as the inference engine, attention backend, and hardware platform, subtly influence how inputs are processed. These components differ in their implementations and thereby induce small numerical deviations across systems when running the same model. While prior work has established the theoretical existence of such deviations, their security implications have remained unexplored. In this paper, we show that these deviations are characteristic of specific components and propagate to observable textual outputs, exposing the inference system to any party that can query the model. Building on this observation, we introduce a fingerprinting method that analyzes the prompt-response behavior of LLMs to identify components of the inference system. Our empirical evaluation demonstrates that the inference engine, attention backend, and underlying hardware platform can be identified reliably, even when the LLM is operated at non-zero temperature. We show that preventing fingerprinting is fundamentally hard, as it would require eliminating numerical differences between hardware and software stacks. We therefore propose partial mitigations and discuss their impact.

Machine learning models are routinely deployed on a wide range of computing hardware. Although such hardware is typically expected to produce identical results, differences in its design can lead to small numerical variations during inference. In this work, we show that these variations can be exploited to create backdoors in machine learning models. The core idea is to shape the model's decision function such that it yields different predictions for the same input when executed on different hardware. This effect is achieved by locally moving the decision boundary close to a target input and then refining numerical deviations to flip the prediction on selected hardware. We empirically demonstrate that these hardware-triggered backdoors can be created reliably across common GPU accelerators. Our findings reveal a novel attack vector affecting the use of third-party models, and we investigate different defenses to counter this threat.

Autonomous agents based on large language models (LLMs) are rapidly emerging as a general-purpose technology, with recent systems such as OpenClaw extending their capabilities through broad tool use, third-party skills, and deeper integration into user environments. At the same time, these agentic systems introduce substantial security risks by combining unconstrained capabilities with access to sensitive user data. In this work, we investigate the security of LLM-based agents through the lens of operating systems. We argue that both face strikingly similar challenges in isolating resources, separating privileges, and mediating communication. Guided by this perspective, we survey the current landscape of open-source agents, derive a unified agent architecture, and systematically analyze potential attack vectors. To validate this analysis, we conduct a case study evaluating four widely used OpenClaw-like agents. Even under modest attacker capabilities, we find that several protection mechanisms fail in practice and that secure operation requires detailed system knowledge and careful configuration. However, we also observe that while some agentic capabilities remain insecure by design, many vulnerabilities can be mitigated using well-established techniques from operating system security. We conclude with a set of recommendations for the secure design of agentic systems.

Adversarial examples in machine learning are typically generated using gradients, obtained either directly through access to the model or approximated via queries to it. In this paper, we propose a much simpler approach to craft adversarial examples, drawing inspiration from insights of explainable machine learning. In particular, we design adversarial image filters that are based on classic edge detection algorithms but optimized to deceive learning models. The resulting untargeted attacks are transferable and require only a single pass over the input. Empirically, we find that 3x3 filters already enable success rates between 30

Different age-related regulations have been proposed to protect minors from harmful content and interactions online. Automated age estimation is central to enforcing such regulations, and vision-language models (VLMs) achieve state-of-the-art performance on this task. However, we find that the zero-shot nature of VLM-based age estimation produces an unexpected side effect we call the identity shortcut: Instead of estimating age from visual features, VLMs tend to identify the depicted person and infer their age from memorized knowledge. This phenomenon leads to substantially incorrect predictions when non-celebrities are misidentified as celebrities. It also produces deceptively high robustness to noise and adversarial perturbations on celebrity images, which dominate popular benchmarks. To mitigate this, we propose an activation steering method that suppresses the shortcut by intervening on the hidden states of the VLM. This method improves age estimation accuracy for both memorized and unseen identities, reducing mean absolute error by up to 25

The benchmarks used to evaluate AI agents in security-critical roles suffer from crucial weaknesses. Building on recent empirical evidence, we characterize three core challenges that undermine security evaluations: benchmark vulnerabilities, temporal staleness, and runtime uncertainty. We then outline practical directions toward building more robust and trustworthy evaluation frameworks.

Human mobility prediction models, which forecast the next location in a user's trajectory, are increasingly deployed in urban analytics, navigation, and personalized services. Yet, little is known about their potential to memorize and expose sensitive user trajectories from training data. While memorization has been extensively studied in language models, mobility prediction poses unique challenges: training sequences encode human behavior at various spatial and temporal scales, creating privacy risks at different granularities. In this paper, we conduct the first systematic audit of memorization in mobility prediction models. While prior work has shown that privacy leaks can arise from such models, we systematically assess and quantify memorization risks at scale. We identify key challenges, including the lack of a randomness space, the multi-scale structure of trajectories, and user-specific behavioral diversity. To address these challenges, we introduce a framework to quantify mobility memorization at different levels of granularity: individual locations, anchor pairs, and subtrajectory segments. We also develop user-grounded reference sets to assess how likely a model is to prefer training data over realistic alternatives. Our evaluation across multiple models and datasets reveals pervasive memorization patterns that correlate with user regularity and increase the risk of data extraction at inference time. Our findings call for mandatory privacy auditing in mobility prediction models.

Large language models now power autonomous agents capable of complex, multi-step tasks in different environments. Accurate and reliable execution of these tasks requires the agent to predict the results of its actions. Recent research proposes to enhance predictive capabilities via specially trained environment simulators-world models. While world models can improve performance, they can also mislead agents into executing harmful actions, creating significant security and privacy risks. In this paper, we raise security concerns regarding the usage of world models in agentic systems. We discover a range of world model specific vulnerabilities, which can be exploited in terminal-based agents to execute malicious code or extract sensitive data. To facilitate future development, we introduce a security benchmark dataset designed for text-based world models. We argue that some risks are intrinsic to approximate world modeling, and show that attackers can induce mispredictions in agentic pipelines with up to 95% success rate, possibly resulting in unintended command execution, denial of service, drainage of wallet and private information extraction. Finally, we provide practical recommendations for practitioners to mitigate the discovered harms and harden agentic systems.

Feature Visualization (FV) is a widely used technique for interpreting concepts learned by Deep Neural Networks (DNNs), which synthesizes input patterns that maximally activate a given feature. Despite its popularity, the trustworthiness of FV explanations has received limited attention. We introduce Gradient Slingshots, a novel method that enables FV manipulation without modifying model architecture or significantly degrading performance. By shaping new trajectories in off-distribution regions of a feature's activation landscape, we coerce the optimization process to converge to a predefined visualization. We evaluate our approach on several DNN architectures, demonstrating its ability to replace faithful FVs with arbitrary targets. These results expose a critical vulnerability: auditors relying solely on FV may accept entirely fabricated explanations. To mitigate this risk, we propose a straightforward defense and quantitatively demonstrate its effectiveness.

Video calls have become an essential part of remote work. They enable employees to collaborate from different locations, including their homes. Transmitting video from the personal living environment, however, poses a privacy risk: Colleagues may gain insight into private information through details in the background. To limit this risk, video conferencing services implement virtual backgrounds that conceal the real environment during a video call. Unfortunately, this protection suffers from imperfections and pixels from the environment occasionally become visible. In this paper, we investigate this privacy leak. We analyze the virtual background techniques used in two major video conferencing services (Zoom and Google) and determine how pixels of the environment leak. Based on this analysis, we propose a reconstruction attack: This attack removes the virtual background by re-purposing the video conferencing software and uses semantic segmentation to filter out the video caller. As a result, only pixels leaking from the environment remain and can be aggregated into a reconstructed image. We examine the efficacy of this attack in a quantitative and qualitative evaluation. In comparison to previous studies, our attack recovers at least 53% more leaked pixels from a video call, exposing larger areas of the environment. We thus conclude that virtual backgrounds currently do not provide an adequate protection in practice.

Linear algebra is a cornerstone of neural network inference. The efficiency of popular frameworks, such as TensorFlow and PyTorch, critically depends on backend libraries providing highly optimized matrix multiplications and convolutions. A diverse range of these backends exists across platforms, including Intel MKL, Nvidia CUDA, and Apple Accelerate. Although these backends provide equivalent functionality, subtle variations in their implementations can lead to seemingly negligible differences during inference. In this paper, we investigate these minor discrepancies and demonstrate how they can be selectively amplified by adversaries. Specifically, we introduce , inputs to models that elicit conflicting predictions depending on the employed backend library. These inputs can even be constructed with integer values, creating a vulnerability exploitable from real-world input domains. We analyze the prevalence and extent of the underlying attack surface and propose corresponding defenses to mitigate this threat.

Motion sensors in mobile devices enable device fingerprinting through hardware-induced variations in sensed data. Although the feasibility of this identification technique has been demonstrated across numerous studies, the literature remains fragmented in terms of experimental setups and evaluation metrics—hindering a comprehensive understanding of its effectiveness and limitations. In this work, we provide the first systematization of the motion sensor fingerprinting landscape, structuring the pipeline into distinct stages and identifying key design parameters and countermeasures. Building on this, we develop a unified evaluation framework to assess each parameter in isolation under realistic conditions. Our results show that motion-based fingerprinting remains effective across diverse settings and classifier architectures, yet current countermeasures fail to provide reliable protection and often degrade data utility. We release our dataset to foster reproducibility and future work in this underexplored yet persistent privacy threat.

Large language models (LLMs) excel in many tasks of software engineering, yet progress in leveraging them for vulnerability discovery has stalled in recent years. To understand this phenomenon, we investigate LLMs through the lens of classic code metrics. Surprisingly, we find that a classifier trained solely on these metrics performs on par with state-of-the-art LLMs for vulnerability discovery. A root-cause analysis reveals a strong correlation and a causal effect between LLMs and code metrics: When the value of a metric is changed, LLM predictions tend to shift by a corresponding magnitude. This dependency suggests that LLMs operate at a similarly shallow level as code metrics, limiting their ability to grasp complex patterns and fully realize their potential in vulnerability discovery. Based on these findings, we derive recommendations on how research should more effectively address this challenge.

AI-based systems, such as Google's GenCast, have recently redefined the state of the art in weather forecasting, offering more accurate and timely predictions of both everyday weather and extreme events. While these systems are on the verge of replacing traditional meteorological methods, they also introduce new vulnerabilities into the forecasting process. In this paper, we investigate this threat and present a novel attack on autoregressive diffusion models, such as those used in GenCast, capable of manipulating weather forecasts and fabricating extreme events, including hurricanes, heat waves, and intense rainfall. The attack introduces subtle perturbations into weather observations that are statistically indistinguishable from natural noise and change less than 0.1 % of the measurements-comparable to tampering with data from a single meteorological satellite. As modern forecasting integrates data from nearly one hundred satellites and many other sources operated by different countries, our findings highlight a critical security risk with the potential to cause large-scale disruptions and undermine public trust in weather forecasting.

Audio content recognition is an emerging technology that forms the basis for mobile services, such as automatic song recognition, second-screen synchronization, and broadcast monitoring. The technology utilizes audio fingerprints, short patterns that are extracted from audio recordings of a smartphone and enable the identification of specific content. These fingerprints are generally considered privacy-friendly, as they contain minimal information of the original signal. As a result, mobile applications have emerged in the past few years that silently monitor user habits by collecting such audio fingerprints in the background. In this paper, we systematically examine whether audio fingerprints leak sensitive information from the recording environment and potentially violate the privacy of smartphone users. To this end, we analyze three popular audio recognition solutions and develop attacks to infer sensitive information from their fingerprints. To the best of our knowledge, we are the first to show that the identification of speakers and words in the fingerprints is possible. Based on our analysis, we conclude that current audio fingerprints do not sufficiently protect privacy and should be used with great caution.

Statically analyzing dynamically-typed code is a challenging endeavor, as even seemingly trivial tasks such as determining the targets of procedure calls are non-trivial without knowing the types of objects at compile time. Addressing this challenge, gradual typing is increasingly added to dynamically-typed languages, a prominent example being TypeScript that introduces static typing to JavaScript. Gradual typing improves the developer's ability to verify program behavior, contributing to robust, secure and debuggable programs. In practice, however, users only sparsely annotate types directly. At the same time, conventional type inference faces performance-related challenges as program size grows. Statistical techniques based on machine learning offer faster inference, but although recent approaches demonstrate overall improved accuracy, they still perform significantly worse on user-defined types than on the most common built-in types. Limiting their real-world usefulness even more, they rarely integrate with user-facing applications. We propose CodeTIDAL5, a Transformer-based model trained to reliably predict type annotations. For effective result retrieval and re-integration, we extract usage slices from a program's code property graph. Comparing our approach against recent neural type inference systems, our model outperforms the current state-of-the-art by 7.85% on the ManyTypes4TypeScript benchmark, achieving 71.27% accuracy overall. Furthermore, we present JoernTI, an integration of our approach into Joern, an open source static analysis tool, and demonstrate that the analysis benefits from the additional type information. As our model allows for fast inference times even on commodity CPUs, making our system available through Joern leads to high accessibility and facilitates security research.

Deep learning has enabled remarkable progress in binary code analysis. In particular, pre-trained embeddings of assembly code have become a gold standard for solving analysis tasks, such as measuring code similarity or recognizing functions. These embeddings are capable of learning a vector representation from unlabeled code. In contrast to natural language processing, however, label information is not scarce for many tasks in binary code analysis. For example, labeled training data for function boundaries, optimization levels, and argument types can be easily derived from debug information provided by a compiler. Consequently, the main motivation of embeddings does not transfer directly to binary code analysis. In this paper, we explore the role of pre-trained embeddings from a critical perspective. To this end, we systematically evaluate recent embeddings for assembly code on five downstream tasks using a corpus of 1.2 million functions from the Debian distribution. We observe that several embeddings perform similarly when sufficient labeled data is available, and that differences reported in prior work are hardly noticeable. Surprisingly, we find that end-to-end learning without pre-training performs best on average, which calls into question the need for specialized embeddings. By varying the amount of labeled data, we eventually derive guidelines for when embeddings offer advantages and when end-to-end learning is preferable for binary code analysis.

Backdoors pose a serious threat to machine learning, as they can compromise the integrity of security-critical systems, such as self-driving cars. While different defenses have been proposed to address this threat, they all rely on the assumption that the hardware on which the learning models are executed during inference is trusted. In this paper, we challenge this assumption and introduce a backdoor attack that completely resides within a common hardware accelerator for machine learning. Outside of the accelerator, neither the learning model nor the software is manipulated, so that current defenses fail. To make this attack practical, we overcome two challenges: First, as memory on a hardware accelerator is severely limited, we introduce the concept of a minimal backdoor that deviates as little as possible from the original model and is activated by replacing a few model parameters only. Second, we develop a configurable hardware trojan that can be provisioned with the backdoor and performs a replacement only when the specific target model is processed. We demonstrate the practical feasibility of our attack by implanting our hardware trojan into the Xilinx Vitis AI DPU, a commercial machine-learning accelerator. We configure the trojan with a minimal backdoor for a traffic-sign recognition system. The backdoor replaces only 30 (0.069%) model parameters, yet it reliably manipulates the recognition once the input contains a backdoor trigger. Our attack expands the hardware circuit of the accelerator by 0.24% and induces no run-time overhead, rendering a detection hardly possible. Given the complex and highly distributed manufacturing process of current hardware, our work points to a new threat in machine learning that is inaccessible to current security mechanisms and calls for hardware to be manufactured only in fully trusted environments.