PRISM: Generation-Time Detection and Mitigation of Secret Leakage in Multi-Agent LLM Pipelines

Multi-agent LLM systems introduce a security risk in which sensitive information accessed by one agent can propagate through shared context and reappear in downstream outputs, even without explicit adversarial intent. We formalise this phenomenon as propagation amplification, where leakage risk...

Automated Framework to Evaluate and Harden LLM System Instructions against Encoding Attacks

System Instructions in Large Language Models LLMs are commonly used to enforce safety policies, define agent behavior, and protect sensitive operational context in agentic AI applications. These instructions may contain sensitive information such as API credentials, internal policies, and...

POLARIS: Explainable Artificial Intelligence for Mitigating Power Side-Channel Leakage

Microelectronic systems are widely used in many sensitive applications e.g., manufacturing, energy, defense. These systems increasingly handle sensitive data e.g., encryption key and are vulnerable to diverse threats, such as, power side-channel attacks, which infer sensitive data through dynamic...

CVE-2024-39691

CVE-2024-39691 affects matrix-appservice-irc, a Node.js IRC bridge for Matrix. Before version 2.0.1, the bridge used the Matrix homeserver-provided timestamp (origin_server_ts) to decide if a user could see the event being replied to. A malicious homeserver could fabricate this timestamp, causing...

Unbreakable Enterprise kernel security update

2.6.32-400.34.3 - inet: fix addrlen/msg-msgnamelen assignment in recverror and rxpmtu functions Hannes Frederic Sowa 18247290 CVE-2013-7263 CVE-2013-7265 2.6.32-400.34.2 - exec/ptrace: fix getdumpable incorrect tests Kees Cook 18239033 CVE-2013-2929 CVE-2013-2929 - inet: prevent leakage of...