Leveraging AI agents and TEE’s to achieve the best of both worlds: private code with public trust
ATSource creates an environment where code vulnerability scanning happens in a completely secure, verifiable way. The application uses a Trusted Execution Environment (TEE) provided by Marlin Protocol to ensure that code analysis is private and tamper-proof.
When users submit code through our clean, modern UI, it's securely sent to an isolated TEE where Nillion's Secret LLM analyzes it for vulnerabilities like SQL injection, XSS, CSRF, insecure cryptography, and more. The results are returned with a cryptographic signature that verifies the analysis happened within the secure environment and was not modified.
This approach solves the fundamental issue of trusting security analysis - you don't need to trust the provider because the TEE provides cryptographic proof that the analysis was performed securely and accurately. The project bridges the gap between privacy-preserving AI security tools and verifiable computation.
ATSource combines several technologies for a complete, secure code analysis solution:
Frontend: Built with Next.js and TypeScript, featuring a clean React UI with Monaco code editor for syntax highlighting.
Backend: Node.js server running within a Marlin TEE, which communicates with Nillion's Secret LLM to analyze code securely.
TEE Integration: We leverage Marlin's Trusted Execution Environment to perform code analysis inside a secure enclave, providing cryptographic signatures that verify the analysis was conducted securely.
Privacy-Preserving AI: Nillion's Secret LLM allows us to analyze sensitive code without exposing it to third parties, adding an additional layer of privacy and security.
Verifiable Results: Each analysis comes with cryptographic proof that the code was analyzed within the secure environment, ensuring results weren't tampered with.