相关论文: Large-Scale Analysis of Style Injection by Relativ…

Aligning large language models with human preferences is critical for creating reliable and controllable AI systems. A human preference can be visualized as a high-dimensional vector where different directions represent trade-offs between…

Recent advances in large reasoning models have leveraged reinforcement learning with verifiable rewards (RLVR) to improve reasoning capabilities. However, scaling these methods typically requires extensive rollout computation and large…

We show how an off-path (spoofing-only) attacker can perform cross-site scripting (XSS), cross-site request forgery (CSRF) and site spoofing/defacement attacks, without requiring vulnerabilities in either web-browser or server and…

This paper proposes an algorithm that aims to improve generalization for reinforcement learning agents by removing overfitting to confounding features. Our approach consists of a max-min game theoretic objective. A generator transfers the…

Low-Rank Adaptation (LoRA) is widely used for parameter-efficient fine-tuning of large language models, but it is notably ineffective at removing backdoor behaviors from poisoned pretrained models when fine-tuning on clean dataset. Contrary…

Masked diffusion large language models (dLLMs) are emerging as promising alternatives to autoregressive LLMs, offering competitive performance while supporting unique generation capabilities such as inpainting. We explore how inpainting can…

Cross-site scripting (XSS) is one of the major threats menacing the privacy of data and the navigation of trusted web applications. Since its reveal in late 1999 by Microsoft security engineers, several techniques have been developed in the…

Strong physical unclonable function (PUF) is a promising solution for device authentication in resourceconstrained applications but vulnerable to machine learning attacks. In order to resist such attack, many defenses have been proposed in…

Web-use agents are rapidly being deployed to automate complex web tasks with extensive browser capabilities. However, these capabilities create a critical and previously unexplored attack surface. This paper demonstrates how attackers can…

A Webview embeds a full-fledged browser in a mobile application and allows the application to expose a custom interface to JavaScript code. This is a popular technique to build so-called hybrid applications, but it circumvents the usual…

Modern browsers implement different security policies such as the Content Security Policy (CSP), a mechanism designed to mitigate popular web vulnerabilities, and the Same Origin Policy (SOP), a mechanism that governs interactions between…

Agentic large language model systems increasingly automate tasks by retrieving URLs and calling external tools. We show that this workflow gives rise to implicit prompt injection: adversarial instructions embedded in automatically generated…

AI alignment in the shape of Reinforcement Learning from Human Feedback (RLHF) is increasingly treated as a crucial ingredient for high performance large language models. Proximal Policy Optimization (PPO) has been positioned by recent…

Website Fingerprinting (WF) attacks exploit patterns in encrypted traffic to infer the websites visited by users, posing a serious threat to anonymous communication systems. Although recent WF techniques achieve over 90% accuracy in…

Modern society is getting accustomed to the Internet of Things (IoT) and Cyber-Physical Systems (CPS) for a variety of applications that involves security-critical user data and information transfers. In the lower end of the spectrum, these…

RL-based post-training of language models is almost exclusively done using on-policy methods such as PPO. These methods cannot learn from arbitrary sequences such as those produced earlier in training, in earlier runs, by human experts or…

Leaderboards for LRMs have turned evaluation into a competition, incentivizing developers to optimize directly on benchmark suites. A shortcut to achieving higher rankings is to incorporate evaluation benchmarks into the training data,…

Technical Questions and Answers (Q&A) sites, such as Stack Overflow (SO), accumulate a significant variety of information related to software development in posts from users. To ensure the quality of this information, SO encourages its…

Template engines play a pivotal role in modern web application development by enabling the dynamic rendering of content, products, and user interfaces. Today, they are essential for any website that handles dynamic data, from e-commerce to…

Reinforcement learning (RL) is widely used for post-training large language models (LLMs) in code editing, where group-relative methods, such as GRPO, are popular due to their critic-free and normalized advantage estimation. However, in…