Design and Development of an ML/DL Attack Resistance of RC-Based PUF for IoT Security

This collection presents recent research across ML/AI and IoT security highlighting both new offensive techniques and practical defenses. Key risks: SNEAKDOOR demonstrates highly stealthy backdoor attacks tailored to distribution-matching dataset condensation (high risk to model training pipelines and downstream models); empirical studies show small LMs and multi-agent assistants remain vulnerable to jailbreaks and prompt-injection, with SafeClaw-R and GUARD-SLM proposed as runtime enforcement and token-activation defenses respectively; supply-chain threats against LLM pipelines are addressed‌