WebMCP Tool Surface Poisoning: Runtime Manipulation Attacks on LLM Agents

WebMCP is a newly emerging protocol that enables websites to expose tools directly to AI agents, bypassing traditional user interfaces and introducing new security risks. The dynamic exposure of agent-accessible tools in WebMCP expands the attack surface of web sessions, especially when third-par...

SUSE CVE-2020-15682

When a link to an external protocol was clicked, a prompt was presented that allowed the user to choose what application to open it in. An attacker could induce that prompt to be associated with an origin they didn't control, resulting in a spoofing attack. This was fixed by changing external...