lbeurerkellner

5 ms·

1.
How We Evaluate Model Risk(research.snyk.io)
2 points by lbeurerkellner 28d ago|0 comments
2.
Skill-Inject: Measuring Agent Vulnerability to Skill File Attacks(arxiv.org)
1 points by lbeurerkellner 6mo ago|1 comments
3.
Snyk Agent Scan: Security scanner for AI agents, MCP servers and agent skills(github.com)
2 points by lbeurerkellner 7mo ago|0 comments
4.
MCP-Scan: A Security scanner for AI agents, MCP servers and agent skills(github.com)
3 points by lbeurerkellner 7mo ago|0 comments
5.
Technical Report: Exploring the Emerging Threats of the Agent Skill Ecosystem(github.com)
3 points by lbeurerkellner 7mo ago|0 comments
6.
by lbeurerkellner 10mo ago
Interesting report. Though, I think many of the attack demos cheat a bit, by putting injections more or less directly in the prompt (here via a website at least). I know it is only one more step, but from a privilege perspective, having the
7.
by lbeurerkellner 10mo ago
Everybody should try. It helps a ton to demystify the relatively simple but powerful underpinning of how modern agents work. You can get quite far quite quickly. My toy implementation [1] is <600 LOC and even supports MCP. [1] https:&#x
8.
MCP-Scan: Constrain, log and scan your MCP server for security vulnerabilities(github.com)
2 points by lbeurerkellner 11mo ago|0 comments
9.
Exposing Novel Prompt Injections, "Toxic Flows" in Agentic Systems(invariantlabs.ai)
1 points by lbeurerkellner 1y ago|0 comments
10.
by lbeurerkellner 1y ago
This is way more common with popular MCP server/agent toolsets than you would think. For those interested in some threat modeling exercise, we recently added a feature to mcp-scan that can analyze toolsets for potential lethal trifecta
11.
Hardening AI Agent Security with Toxic Flow Analysis(invariantlabs.ai)
2 points by lbeurerkellner 1y ago|0 comments
12.
by lbeurerkellner 1y ago
This looks really cool, thanks for sharing.
13.
Mcp-scan: constrain, log and scan your active MCP connections(github.com)
1 points by lbeurerkellner 1y ago|0 comments
14.
by lbeurerkellner 1y ago
We have published the full trace, with tool outputs here now: https://explorer.invariantlabs.ai/trace/5f3f3f3c-edd3-4ba7-a... The minesweeper comment was caused by the issue containing explicit instructions in the vers
15.
by lbeurerkellner 1y ago
I agree. It is also interesting to consider how AI security, user eduction/posture and social engineering relate. It is not traditional security in the sense of a code vulnerability, but is is a real vulnerability that can be exploited
16.
by lbeurerkellner 1y ago
One of the authors here. Thanks for posting. If you are interested in learning more about MCP and agent security, check out some of the following resources, that we have created since we started working on this: * The full execution trace o
17.
by lbeurerkellner 1y ago
Yes, any MCP server that is connected to an untrusted source of data, could be abused by an attacker to take over the agent. Here, we just showed an in-server exploit, that does not require more than one server. Also, check out our work on
18.
by lbeurerkellner 1y ago
I agree, one of the issues are tokens with too broad permission sets. However, at the same time, people want general agents which do not have to be unlocked on a repository-by-repository basis. That's why they give them tokens with tho
19.
by lbeurerkellner 1y ago
Be sure to check out the malicious issue + response here: https://github.com/ukend0464/pacman/issues/1 . It's hilarious, the agent is even tail-wiggling about completing the exploit.
20.
mcp-scan: Constrain, log and scan MCP traffic for security vulnerabilities(github.com)
2 points by lbeurerkellner 1y ago|0 comments
21.
MCP Streamable HTTP – Python and TypeScript Examples(github.com)
2 points by lbeurerkellner 1y ago|0 comments
22.
Show HN: Guardrails – A Contextual Security Layer for Agentic AI Systems(invariantlabs.ai)
1 points by lbeurerkellner 1y ago|0 comments
23.
Securing Agents with Rules(explorer.invariantlabs.ai)
2 points by lbeurerkellner 1y ago|0 comments
24.
by lbeurerkellner 1y ago
Original author of some of the initial security notes at Invariant Labs here. Some resources, if you want to learn about MCP security: * Initial Security Note describing Tool Poisoning, Rug Pulls, Tool Shadowing for the first time with diag
25.
MCP-scan: A security scanning tool for MCP servers(github.com)
1 points by lbeurerkellner 1y ago|0 comments
26.
by lbeurerkellner 1y ago
I think what is not to be overlooked in these discussions, is the user base of MCP. In my experience, many less-technical folks started using MCP, and that makes security issues all the more relevant. This audience often lacks intuition aro
27.
AgentDojo: An Environment to Evaluate Prompt Injection Attacks on AI Agents(agentdojo.spylab.ai)
3 points by lbeurerkellner 1y ago|0 comments
28.
by lbeurerkellner 1y ago
We recently updated our post on WhatsApp tool poisoning, to include an attack setup that does not require a malicious MCP server. It suffices to send you an injected message. See: https://invariantlabs.ai/blog/whatsapp-
29.
WhatsApp MCP Exploited: Exfiltrating your message history via MCP(invariantlabs.ai)
3 points by lbeurerkellner 1y ago|0 comments
30.
by lbeurerkellner 1y ago
I think it is important to understand the difference between instruction and implementation level attacks. Yes, running unsafe bash commands in the implementation can be prevented by sandboxing. Instruction level attacks like tool poisoning

More ›