「论文速读」｜大语言模型（LLM）智能体可以自主利用1-day漏洞 本次分享论文：LLM Agents can Autonomously Exploit One-day Vulnerabilities 基本信息 原文作者：Richard Fang, Rohan Bindu, Akul Gupta, Daniel Kang作者单位：无详细信息提供关键词：大语言模型, 网络安全, 1-day漏洞, 利用原文链接：https://arxiv...
Table 13 LLM agents for cybersecurity Full size table LLM Agent is “a system that can use an LLM to reason through a problem, create a plan to solve the problem, and execute the plan with the help of a set of tools (Varshney 2023).” By simulating complex network behaviors and attack...
MultiOnis a startup building AI agents that can autonomously solve real-world tasks. This drives a need for a high level of AI capabilities in autonomous actions, planning, reasoning, acquiring knowledge via search, and balancing exploration and taking action to exploit information on hand. MultiO...
LLM_Agents_can_Autonomously_Exploit_One-day_Vulnerabilities.md Many_shot_Jailbreaking.md Multilingua_Jailbreak_Challenges_in_Large_Language_Models.md PoisonedRAG_Knowledge_Poisoning_Attacks_to_Retrieval-Augmented_Generation_of_Large_Language_Models.md Poisoning_Retrieval_Corpora_by_Injecting_Adversarial...
INJECAGENT: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents 2024.3.24 arXiv Attack Prompt-injection LLM Agents can Autonomously Exploit One-day Vulnerabilities 2024.4.11 arXiv Application Attack Model featuresTitleDatePublishedTag Studying Large Language Model Behaviors ...
INJECAGENT: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents 2024.3.24 arXiv Attack Prompt-injection LLM Agents can Autonomously Exploit One-day Vulnerabilities 2024.4.11 arXiv Application Attack Model featuresTitleDatePublishedTag Studying Large Language Model Behaviors ...
PARDEN, Can You Repeat That? Defending against Jailbreaks via Repetition2024.5.14ICML 2024Method Defense GradSafe: Detecting Unsafe Prompts for LLMs via Safety-Critical Gradient Analysis2024.5.29ACL 2024Method Defense SafeDecoding: Defending against Jailbreak Attacks via Safety-Aware Decoding2024.8.22ACL...