adversarial-machine-learning

该论文证明，通过将加密的神经网络嵌入宿主模型权重中，攻击者可以隐藏任意能力，即使拥有完整模型权重访问权限，在密码学假设下这些隐藏回路也难以被触发和解释，从而揭示了基于能力引导的模型审计的根本局限性。

The paper introduces SWB-DM, a robust aggregation method for federated learning that combines sliced Wasserstein barycenters with delayed momentum caching to handle Byzantine attacks and partial participation effectively.

This paper introduces Krum-Proxy, a selection-aware backdoor attack that bypasses distance-based robust aggregation methods like Krum in federated learning by optimizing adversarial updates to mimic benign geometry, achieving high attack success while preserving clean accuracy.

This paper introduces SkillJack, the first attack targeting the experience-to-skill pipeline of self-evolving agents, showing that poisoned experiences can be transformed into persistent malicious skills that evade detection and survive deletion of original records.

A survey on quantum adversarial machine learning, covering attacks, defenses, and theoretical underpinnings.