Tag: machine learning security
Browse our exclusive articles!
BehaviorGuard: Real-Time Backdoor Defense for DRL
BehaviorGuard offers real-time, trigger-agnostic defense against backdoor attacks in deep reinforcement learning, enhancing AI security and reliability.
MAGE: Protecting LLM Agents from Long-Horizon Threats
Discover how MAGE enhances LLM agent safety by using shadow memory to detect and prevent long-horizon threats with minimal overhead.
AI Red Teaming Revolutionized: From Weeks to Hours
Discover how new AI red teaming agents cut security testing from weeks to hours, boosting defenses with unified frameworks and automated workflows.
Adversarial Flow Matching: Imperceptible Attacks on Autonomous Driving
Discover Adversarial Flow Matching, a novel gray-box attack method that imperceptibly compromises end-to-end autonomous driving models with high efficiency...
BadSNN: Backdoor Attacks on Spiking Neural Networks
Discover BadSNN, a novel backdoor attack targeting Spiking Neural Networks by exploiting adversarial spiking neurons and hyperparameter manipulation.