While enabling effective collaboration on complex tasks, LLM-based Multi-Agent Systems (MAS) face critical security challenges due to vulnerabilities at the agent and interaction levels. Most existing MAS security defenses are built upon two core assumptions: semantically-explicit malicious attacks and explicit graph-based modeling of the MAS topology and agent-level interactions. In practice, real-world attacks are becoming more semantically stealthy, while MAS execution is typically asynchronous without the temporal alignment assumed by graph-based propagation models. To address these limitations, we propose AcMAS, an activation-based framework for malicious-behavior detection in MAS. By analyzing internal reasoning states in the activation space of local agents, AcMAS detects even stealthy attacks in a synchronization-robust fashion, without relying on explicit interaction graphs. Moreover, our activation analysis provides critical signals to guide AcMAS in restoring the functionality of compromised agents, rather than the disruptive agent isolation commonly used by the state-of-the-art methods. Comprehensive evaluation demonstrates that AcMAS significantly outperforms graph-based baselines against stealthy attacks, by +0.22 F1 in synchronous settings (0.94 vs. 0.72) and by +0.55 F1 in asynchronous settings (0.93 vs. 0.38), with generalization across diverse open-source LLM backbones, attack intensity, and MAS scale.
GUARD-SLM: Token Activation-Based Defense Against Jailbreak Attacks for Small Language Models
Knight Foundation School of Computing and Information Sciences(奈特基金会计算与信息科学学院);Security, Optimization, and Learning for InterDependent networks laboratory (solid lab)(安全、优化与相互依赖网络学习实验室);Florida International University(佛罗里达国际大学)
Activation-Guided Local Editing for Jailbreaking Attacks
Beihang University(北航);Hangzhou Innovation Institute(杭州创新研究院);HKUST(香港科技大学);South China University of Technology(华南理工大学);Nanyang Technological University(南洋理工大学);Kunming University of Science and Technology(昆明理工大学)
RevPRAG: Revealing Poisoning Attacks in Retrieval-Augmented Generation through LLM Activation Analysis
School of Computer Science, Fudan University(复旦大学计算机科学学院);Department of Computer Science, Worcester Polytechnic Institute(沃思堡理工学院计算机科学系);Institute of Big Data, Fudan University(复旦大学大数据研究院)
Defense Against LLM Backdoors using Critical Neuron Isolation Pruning
Huazhong University of Science and Technology(华中科技大学);Nanjing University of Aeronautics and Astronautics(南京航空航天大学);Nanyang Technological University(南洋理工大学)
Erasing Self-Supervised Learning Backdoor by Cluster Activation Masking
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所 多模态人工智能系统国家重点实验室);University of Chinese Academy of Sciences(中国科学院大学);Xiaomi Inc.(小米公司);State Key Laboratory of CNS/ATM, Beihang University(北京航空航天大学 CNS/ATM 国家重点实验室);College of Computer Science, Inner Mongolia University(内蒙古大学 计算机学院)
On-the-fly Code Activation for Attack Surface Reduction
现代代码重用攻击充分利用了臃肿的软件。攻击者将原本无害的代码中的短指令序列拼凑在一起,以执行恶意操作。消除这些可重用的代码片段(称为小工具)已成为减少攻击面的主要关注点之一。其目的是打破这些小工具链,从而使这种代码重用攻击不可能或不太常见。先前关于攻击面减少的工作通常试图通过对应用程序进行子集化来消除此类攻击,例如,通过用户指定的输入、配置或功能,或者通过关注第三方库来实现对应用程序干扰最小的高小工具减少。在这项工作中,我们提出了一个通用的,整个程序的攻击面减少技术称为OCA,显着减少小工具,并有轻微的性能下降。OCA不需要用户输入,并保持所有功能不变。OCA识别特定的程序点,并通过分析确定在运行时启用/禁用的关键功能集。因此,运行时系统在执行期间控制启用的函数的集合,从而显著减少攻击者可以使用的活动小工具的集合,并且通过扩展,显著地减少活动小工具链的集合。在SPEC CPU 2017上,我们的框架实现了73.2%的小工具减少,平均速度仅为4%。在10个GNU coreutils应用程序上,它实现了87.2%的减少。在nginx服务器上,它实现了80.3%的减少和2%的减速。我们还提供了一个小工具链打破所有应用程序的研究,并表明,我们的框架打破了壳产卵链在所有情况下。
InverTune: Removing Backdoors from Multimodal Contrastive Learning Models via Trigger Inversion and Activation Tuning
1 School of Cyber Science;Engineering, Wuhan University 0.3em 2 School of Computer Science, Wuhan University 0.3em 3 Institute for Math \& AI, Wuhan University
A Novel Defense Against Poisoning Attacks on Federated Learning: LayerCAM Augmented with Autoencoder
CyLab Security and Privacy Institute, Carnegie Mellon University(卡内基梅隆大学CyLab安全与隐私研究所);Real-Time and Embedded Computing Systems Research Centre (CISTER)(实时与嵌入式计算系统研究中心(CISTER));Commonwealth Scientific and Industrial Research Organization (CSIRO)(英联邦科学与工业研究组织(CSIRO));University of Cambridge(剑桥大学);Alan Turing Institute(阿兰·图灵研究所)
Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs
ELLIS Institute Tübingen & MPI for Intelligent Systems(图宾根ELLIS研究所与智能系统马克斯·普朗克研究所);Tübingen AI Center(图宾根人工智能中心);Fraunhofer HHI(弗劳恩霍夫高精尖研究所);ETH Zurich & ETH AI Center(苏黎世联邦理工学院与苏黎世人工智能中心);University of Tübingen(图宾根大学);TU Dublin(都柏林技术大学);TU Berlin & BIFOLD(柏林技术大学与BIFOLD)
Defending the Edge: Representative-Attention Defense against Backdoor Attacks in Federated Learning
Dept. of Computer Science Faculty of Engineering University of Manchester Manchester, United Kingdom;Dept. of Computer Science Faculty of Engineering University of Manchester \& MBZUAI Abu Dhabi, United Arab Emirates;Dept. of Computer Science Faculty of Engineering University of Manchester \& KU Leuven Manchester, United Kingdom