Video Abstract: AI and Security


June 30, 2017


Taesoo Kim, Dawn Song, Michael Walker


Georgia Institute of Technology, University of California-Berkeley, Microsoft


In the future, every company will be using AI, which means that every company will need a secure infrastructure that addresses AI security concerns. At the same time, the domain of computer security has been revolutionized by AI techniques, including machine learning, planning, and automatic reasoning. What are the opportunities for researchers in both fields—security infrastructure and AI—to learn from each other and continue this fruitful collaboration? This session will cover two main topics. In the first half, we will discuss how AI techniques have changed security, using a case study of the DARPA Cyber Grand Challenge, where teams built systems that can reason about security in real time.  In the second half, we will talk about security issues inherent in AI. How can we ensure the integrity of decisions from the AI that drives a business? How can we defend against adversarial control of training data? Together, we will identify common problems for future research.