Crack The Hashword - Using Reinforcement Learning For Faster Password Cracking
Modern cyber defense strategies often include the practice of deploying "red team" professionals who can simulate potential adversary attacks on a target organization's IT infrastructure to proactively identify and mitigate security weaknesses. One common adversary technique is password cracking, which is the process of attempting to recover passwords stored within system data (often in a hashed form). Current state-of-the-art password cracking tools (e.g. John The Ripper) allow cybersecurity professionals to execute these attacks, but are still limited in their capability because they rely on deterministic rule-sets and wordlists. Our team designed and executed an experiment to explore whether a trained reinforcement learning (RL) model might be able to crack more passwords faster than existing password cracking tools.
Interns: Adeniyi Adegbuyi, Saanvi Kakarlapudi, Abdul Moalim, Jhanvi Sanwal, and Tristan Wang