AI-Driven Cyber Operations: Automated Vulnerability Discovery and Weaponization

Both commercial and defense entities have a dire need to find vulnerabilities in code. In the commercial and defensive cyber worlds, such vulnerabilities need to be found to keep attackers out. In offensive cyber, such vulnerabilities need to be found to get in. Autonomous Cyber has found a way to take advantage of recent advances in artificial intelligence (AI) to rapidly find new vulnerabilities in code. In the right hands, this technology will massively contribute to securing Department of the Air Force (DAF) code, the U.S. defense industrial base, and the critical software on which all Americans depend. The Autonomous Cyber approach unites cutting-edge AI with a proven technique that has been used for decades to find the most important vulnerabilities: fuzzing. Fuzzing is a powerful technique for uncovering security vulnerabilities and reliability issues, but attaching a fuzzer to a codebase often poses challenges for developers. Our proposal uses artificial intelligence to automatically attach fuzzing engines to codebases, which allows the codebases to be fuzzed in the absence of the attention of a skilled (and expensive) cybersecurity-savvy developer. By eliminating the integration step, we can build a business that secures code at scale. Crucially, enterprises and the DAF alike will be able to simply download and run our product in a way that has not previously been possible for fuzzers. Autonomous Cyber is an MIT-based startup composed of a former seven-year NSA officer, two Army veterans, and a computer engineer-turned-investor. Our head of research is one of the leading researchers in this space: he gave a talk on fuzzing and neural networks at a leading hacking conference this August. The other three members of the team come from technical backgrounds and have taken on graduate business studies at the MIT Sloan School of Management. We bring a unique blend of deep technical expertise, demonstrated commitment to U.S. national security, and experience commercializing innovative technology both in government and the private sector. Our proposal is backed by a research partnership with arguably the world’s leading AI laboratory, the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. This proposal will fund crucial research into the reliability of AI for cybersecurity: research that Autonomous Cyber will rely upon as it builds its highly impactful product. Autonomous Cyber has put together the right team focusing on the right problem at the right time: we can use cutting-edge AI to supercharge long-proven vulnerability discovery techniques to deliver immense value to both the DAF and the private sector alike.