Show HN: Decypher-env, an RL Env for breaking AES encryption

Category: ai-ml

Tags: reinforcement-learning, aes, cybersecurity, code-generation, openai-gym

Score: 7.3/10 (Innovation: 8, Technical: 8, Documentation: 7, Utility: 6)

A reinforcement learning environment where the policy writes C++ solver code to break reduced-round AES-XTS encryption. It treats code generation as the action space, compiles and executes solvers in a sandbox, and rewards progress on a curriculum of increasing AES rounds, making it a novel benchmark for code-generating RL agents.

Target audience: AI/ML researchers, RL practitioners, cryptographers

Repository: https://github.com/juansebastianl/decypher-env · Python · MIT

View on Hacker News