Peking University and WeChat Security Team Release Full Lifecycle Security Survey of General-Purpose Agents
Start here
Peking University's Knowledge Computing Lab and Tencent WeChat's security technology team released a general agent security survey, dividing agent operation into three stages: provenance, orchestration and execution, and pointing out that after execution results are written into long-term memory, skills or artifacts, they form persistent feedback, allowing a single attack to survive across sessions and be amplified; it summarizes four categories of attack surfaces—skill supply chain, user side, long-term memory and external environment—summarizes three main lines of evaluation and three layers of defense, and proposes a security closed loop for self-evolving agents.
Generated by AI from reports · Updated 3 days ago
Attention and developments
Attention, hourly
Now 27, -1% in 24 hours·Peak 31 (Nov 14 16:00)
Progress timeline
- WeChat · Synced (Jiqizhixin)Opening "Pandora's Toolbox": Peking University and WeChat Security Team Build a Full Lifecycle Security Picture of General-Purpose Agents
Peking University's Knowledge Computing Lab (KCL) and Tencent WeChat's security technology team jointly released a survey paper, "Pandora's Toolbox," systematically reviewing, from a lifecycle perspective, the security risks, attack methods, evaluation benchmarks and defense techniques of general-purpose agents. The paper divides agent operation into three stages: provenance, orchestration and execution, and emphasizes that persistent feedback formed when execution results are written into long-term memory, reusable skills or project artifacts allows a single attack to survive across sessions and be amplified; along this framework it summarizes four attack surfaces—skill supply chain, user-side attacks, long-term memory and the external environment—groups evaluations into three main lines: domain-specific evaluation, general evaluation and offline trajectory auditing, divides defenses into three layers: input and context filtering, decision and control integrity, and runtime monitoring and execution constraints, and proposes a continuous security loop of risk generation, risk diagnosis and security evolution for self-evolving agents. The full paper is available at https://openreview.net/forum?
Related discussion
- WeChatDoonsecWhen AI Learns to "Act": Full-Lifecycle Security Attack and Defense and Breakthroughs for General Agents(opens the original post in a new tab)
- Older than 72 hours · Not included in the 72-hour figures
- WebSynced (Jiqizhixin)Opening "Pandora's Toolbox": Peking University and WeChat Security Team Build a Full Lifecycle Security Picture of General-Purpose Agents(opens the original post in a new tab)
- X@bitbiz_bruceThe Knowledge Computing Lab of Peking University and the WeChat Security Technology Team of Tencent jointly released a general agent safety survey, proposing a unified analysis framework covering provenance, orchestration, execution, and persistent feedback. The paper points out that agents can select skills, invoke tools, and read and write long-term memory; safety risks are not only in model outputs but also involve the loading and composition of capabilities, actual actions, and exec…(opens the original post in a new tab)
1 likes·Nov 14 16:41Collect
Attention trend
New sources per day
- Nov 14: 2 new sources (1 outlets, 1 accounts)
- Nov 15: 0 new sources (0 outlets, 0 accounts)
- Nov 16: 0 new sources (0 outlets, 0 accounts)
- Nov 17: 1 new sources (1 outlets, 0 accounts)
- Nov 18: 0 new sources (0 outlets, 0 accounts)
- Nov 19: 0 new sources (0 outlets, 0 accounts)
- Nov 20: 0 new sources (0 outlets, 0 accounts)