Llm-Security

A systematic evaluation of cache side-channel attacks on modern Intel microarchitectures, and a new security architecture to mitigate integrity violations in LLM-integrated app systems.

LLM agents are central today, revealing new architectural risks like capability laundering and the stochastic deputy problem, while hardware vulnerabilities expand into GPU memory architectures.

New perspectives on LLM agent security: a census of the Model Context Protocol ecosystem, attacks via gated memories and contextual bias, new attacks on agent architectures, and LLM-guided smart contract fuzzing.

Recent NDSS research covering LLM jailbreak causality, bootloader fuzzing, the impact of compiler inlining on binary analysis, enclave privilege scaling, adversarial evasion of traffic detection, and identity-preserving facial anonymization.

A deep dive into TypeSafe AI’s Jev, exploring its role as a System One model that returns calibrated, typed decisions, and tracing the long history of alternative approaches in the field.