Multimodal User Authentication Method via Fusion of Keystroke Dynamics and Glove-Based Hand Kinematics

Collection of recent security and AI-safety research covering multimodal biometric authentication, manipulation of medical notes by LLMs, specification infrastructure for AI oversight, a trusting-trust supply-chain attack against NixOS using GNU strip, distributed backdoors in multi-agent LLM systems, malware-detector robustness, visual jailbreaks, adversarial comments against LLM vulnerability detectors, and web-application fuzzing. The most directly actionable findings concern software supply-chain compromise, distributed agent payload execution, jailbreaks, and evasion of AI-based security検