The widespread reliance on open-source software dramatically increases the risk of vulnerability exploitation, underscoring the need for effective and scalable vulnerability detection (VD). Existing VD techniques, whether traditional machine learning-based or LLM-based approaches like prompt engineering, supervised fine-tuning, or off-policy preference optimization, remain fundamentally limited in their ability to perform context-aware analysis: They depend on fixed inputs or static preference datasets, cannot adaptively explore repository-level dependencies, and are constrained by function-level benchmarks that overlook critical vulnerability context. This paper introduces Vulnerability-Adaptive Policy Optimization (VULPO), an on-policy LLM reinforcement learning framework for context-aware VD. To support training and evaluation, we first construct ContextVul, a new dataset that augments high-quality function-level samples with lightweight method to extract repository-level context information. We then design multi-dimensional reward structuring that jointly captures prediction correctness, vulnerability localization accuracy, and the semantic relevance of vulnerability analysis, thereby guiding the model toward comprehensive contextual reasoning. To address the asymmetric difficulty of different vulnerability cases and mitigate reward hacking, VULPO incorporates label-level and sample-level difficulty-adaptive reward scaling, encouraging the model to explore challenging cases while maintaining balanced reward distribution. Extensive experiments demonstrate the superiority of our VULPO framework in context-aware VD: Our VULPO-4B substantially outperforms existing VD baselines based on prompt engineering and off-policy optimization, improving F1 by 85% over Qwen3-4B and achieving performance comparable to a 150x larger-scale model, DeepSeek-R1-0528.

AI systems are increasingly able to autonomously conduct realistic software engineering tasks, and may soon be deployed to automate machine learning (ML) R&D itself. Frontier AI systems may be deployed in safety-critical settings, including to help ensure the safety of future systems. Unfortunately, frontier and future systems may not be sufficiently trustworthy, and there is evidence that these systems may even be misaligned with their developers or users. Therefore, we investigate the capabilities of AI agents to act against the interests of their users when conducting ML engineering, by sabotaging ML models, sandbagging their performance, and subverting oversight mechanisms. First, we extend MLE-Bench, a benchmark for realistic ML tasks, with code-sabotage tasks such as implanting backdoors and purposefully causing generalisation failures. Frontier agents make meaningful progress on our sabotage tasks. In addition, we study agent capabilities to sandbag on MLE-Bench. Agents can calibrate their performance to specified target levels below their actual capability. To mitigate sabotage, we use LM monitors to detect suspicious agent behaviour, and we measure model capability to sabotage and sandbag without being detected by these monitors. Overall, monitors are capable at detecting code-sabotage attempts but our results suggest that detecting sandbagging is more difficult. Additionally, aggregating multiple monitor predictions works well, but monitoring may not be sufficiently reliable to mitigate sabotage in high-stakes domains. Our benchmark is implemented in the UK AISI's Inspect framework and we make our code publicly available at https://github.com/TeunvdWeij/ctrl-alt-deceit

The rapid expansion of Internet of Things (IoT) devices has transformed industries and daily life by enabling widespread connectivity and data exchange. However, this increased interconnection has introduced serious security vulnerabilities, making IoT systems more exposed to sophisticated cyber attacks. This study presents a novel ensemble learning architecture designed to improve IoT attack detection. The proposed approach applies advanced machine learning techniques, specifically the Extra Trees Classifier, along with thorough preprocessing and hyperparameter optimization. It is evaluated on several benchmark datasets including CICIoT2023, IoTID20, BotNeTIoT L01, ToN IoT, N BaIoT, and BoT IoT. The results show excellent performance, achieving high recall, accuracy, and precision with very low error rates. These outcomes demonstrate the model efficiency and superiority compared to existing approaches, providing an effective and scalable method for securing IoT environments. This research establishes a solid foundation for future progress in protecting connected devices from evolving cyber threats.

Anomaly detection is critical in domains such as cybersecurity and finance, especially when working with large-scale tabular data. Yet, unsupervised anomaly detection-where no labeled anomalies are available-remains challenging because traditional deep learning methods model a single global distribution, assuming all samples follow the same behavior. In contrast, real-world data often contain heterogeneous contexts (e.g., different users, accounts, or devices), where globally rare events may be normal within specific conditions. We introduce a contextual learning framework that explicitly models how normal behavior varies across contexts by learning conditional data distributions $P(\mathbf{Y} \mid \mathbf{C})$ rather than a global joint distribution $P(\mathbf{X})$. The framework encompasses (1) a probabilistic formulation for context-conditioned learning, (2) a principled bilevel optimization strategy for automatically selecting informative context features using early validation loss, and (3) theoretical grounding through variance decomposition and discriminative learning principles. We instantiate this framework using a novel conditional Wasserstein autoencoder as a simple yet effective model for tabular anomaly detection. Extensive experiments across eight benchmark datasets demonstrate that contextual learning consistently outperforms global approaches-even when the optimal context is not intuitively obvious-establishing a new foundation for anomaly detection in heterogeneous tabular data.

Machine learning based predictions are increasingly used in sensitive decision-making applications that directly affect our lives. This has led to extensive research into ensuring the fairness of classifiers. Beyond just fair classification, emerging legislation now mandates that when a classifier delivers a negative decision, it must also offer actionable steps an individual can take to reverse that outcome. This concept is known as algorithmic recourse. Nevertheless, many researchers have expressed concerns about the fairness guarantees within the recourse process itself. In this work, we provide a holistic theoretical characterization of unfairness in algorithmic recourse, formally linking fairness guarantees in recourse and classification, and highlighting limitations of the standard equal cost paradigm. We then introduce a novel fairness framework based on social burden, along with a practical algorithm (MISOB), broadly applicable under real-world conditions. Empirical results on real-world datasets show that MISOB reduces the social burden across all groups without compromising overall classifier accuracy.

Failures in satellite components are costly and challenging to address, often requiring significant human and material resources. Embedding a hybrid AI-based system for fault detection directly in the satellite can greatly reduce this burden by allowing earlier detection. However, such systems must operate with extremely high reliability. To ensure this level of dependability, we employ the formal verification tool Marabou to verify the local robustness of the neural network models used in the AI-based algorithm. This tool allows us to quantify how much a model's input can be perturbed before its output behavior becomes unstable, thereby improving trustworthiness with respect to its performance under uncertainty.

Large Language Model (LLM) agents offer a powerful new paradigm for solving various problems by combining natural language reasoning with the execution of external tools. However, their dynamic and non-transparent behavior introduces critical security risks, particularly in the presence of prompt injection attacks. In this work, we propose a novel insight that treats the agent runtime traces as structured programs with analyzable semantics. Thus, we present AgentArmor, a program analysis framework that converts agent traces into graph intermediate representation-based structured program dependency representations (e.g., CFG, DFG, and PDG) and enforces security policies via a type system. AgentArmor consists of three key components: (1) a graph constructor that reconstructs the agent's runtime traces as graph-based intermediate representations with control and data flow described within; (2) a property registry that attaches security-relevant metadata of interacted tools \& data, and (3) a type system that performs static inference and checking over the intermediate representation. By representing agent behavior as structured programs, AgentArmor enables program analysis for sensitive data flow, trust boundaries, and policy violations. We evaluate AgentArmor on the AgentDojo benchmark, the results show that AgentArmor can reduce the ASR to 3\%, with the utility drop only 1\%.

As AI systems become more capable, deceptive behaviors can undermine evaluation and mislead users at deployment. Recent work has shown that lie detectors can accurately classify deceptive behavior, but they are not typically used in the training pipeline due to concerns around contamination and objective hacking. We examine these concerns by incorporating a lie detector into the labelling step of LLM post-training and evaluating whether the learned policy is genuinely more honest, or instead learns to fool the lie detector while remaining deceptive. Using DolusChat, a novel 65k-example dataset with paired truthful/deceptive responses, we identify three key factors that determine the honesty of learned policies: amount of exploration during preference learning, lie detector accuracy, and KL regularization strength. We find that preference learning with lie detectors and GRPO can lead to policies which evade lie detectors, with deception rates of over 85\%. However, if the lie detector true positive rate (TPR) or KL regularization is sufficiently high, GRPO learns honest policies. In contrast, off-policy algorithms (DPO) consistently lead to deception rates under 25\% for realistic TPRs. Our results illustrate a more complex picture than previously assumed: depending on the context, lie-detector-enhanced training can be a powerful tool for scalable oversight, or a counterproductive method encouraging undetectable misalignment.

The increasing realism of synthetic speech, driven by advancements in text-to-speech models, raises ethical concerns regarding impersonation and disinformation. Audio watermarking offers a promising solution via embedding human-imperceptible watermarks into AI-generated audios.