wrhn

Recent submissions

Comment · 6 months ago
We just open-sourced Model Provenance Kit, a toolkit for answering a simple but important question about AI models Where did this model come from, and how similar is it to another model or family of models

github repo : https://github.com/cisco-ai-defense/model-provenance-kit

Comment · 10 months ago
We’re excited to announce a major new capability in MCP Scanner: behavioral source code scanning. This feature goes beyond traditional security checks by analyzing whether an MCP tool’s actual behavior matches what it claims to do. By combining deep static code analysis with AI-powered semantic reasoning, MCP Scanner can now uncover hidden operations, data exfiltration, privilege abuse, and other deceptive behaviors that pattern-based scanners often miss. This helps catch cases where tools look safe on the surface but do something very different under the hood. As MCP tools become a core part of agentic AI systems, trust and supply-chain security matter more than ever. Behavioral code scanning adds a critical pre-deployment layer to help teams confidently adopt MCP tools at scale. This capability is available today in the open source MCP Scanner and integrates easily into CI/CD pipelines, giving you actionable insights before risky tools ever reach production.

https://github.com/cisco-ai-defense/mcp-scanner

Comment · 1 year ago
This is research work from Cisco AI threat research Team

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton