EU Regulatory Changes

This publication, titled "Architectural Bias in Face Presentation Attack Detection," is a research paper from arXiv that compares the performance of Vision Transformers and Convolutional Neural Net...

This publication introduces CodeSentinel, a proposed three-layer defense framework designed to detect and mitigate indirect prompt injection attacks in AI systems that interact with code. Indirect ...

This publication, PhantomSkill: Malicious Code Injection in Agent Skill Ecosystems, details a newly identified vulnerability in AI agent systems that rely on third-party skills or plugins. The rese...

This publication, dated June 17, 2026, introduces OpenAnt, a novel framework that uses large language models to automate the discovery of software vulnerabilities. The method combines code decompos...

This paper, published on arXiv, introduces Giskard, a new cryptographic protocol designed to secure large-scale decentralized machine learning systems. It addresses two critical vulnerabilities: By...

This paper, published on arXiv, introduces a novel quantitative framework for assessing the risk of compromise in exceptional access architectures, which are systems that allow law enforcement or o...

This publication introduces a novel framework called Compute-Budgeted Exploitability Evidence Graphs, designed to improve how organizations prioritize software vulnerabilities based on their real-w...

This publication introduces a new technical framework, PYPILINE, designed to detect malicious packages in the Python Package Index (PyPI) by analyzing suspicious API calls and employing an automate...

This publication from arXiv presents a technical proposal called CHERI-D, which introduces a method for improving temporal memory safety in CHERI-based hardware architectures. CHERI is a capability...

This publication introduces a technical framework for lifecycle-aware dynamic analysis of machine learning models, aimed at detecting and mitigating security vulnerabilities during execution. The a...