Sheds light on the ability to hack AI and the technology industry’s lack of effort to secure vulnerabilities.
We are accelerating towards the automated future. But this new future brings new risks. It is no surprise that after years of development and recent breakthroughs, artificial intelligence is rapidly transforming businesses, consumer electronics, and the national security landscape. But like all digital technologies, AI can fail and be left vulnerable to hacking. The ability to hack AI and the technology industry’s lack of effort to secure it is thought by experts to be the biggest unaddressed technology issue of our time. Hacking Artificial Intelligence sheds light on these hacking risks, explaining them to those who can make a difference.
Today, very few people—including those in influential business and government positions—are aware of the new risks that accompany automated systems. While society hurdles ahead with AI, we are also rushing towards a security and safety nightmare. This book is the first-ever layman’s guide to the new world of hacking AI and introduces the field to thousands of readers who should be aware of these risks. From a security perspective, AI is today where the internet was 30 years ago. It is wide open and can be exploited. Readers from leaders to AI enthusiasts and practitioners alike are shown how AI hacking is a real risk to organizations and are provided with a framework to assess such risks, before problems arise.
Davey Gibian is a technologist and artificial intelligence practitioner. His career has spanned Wall Street, the White House, and active war zones as he has brought cutting-edge data science tools to solve hard problems. He has built two start-ups, Calypso AI and OMG, was a White House Presidential Innovation Fellow for AI and Cybersecurity, and helped scale Palantir Technologies. He holds patents in machine learning and served in the US Air Force. He currently resides in New York City.
Introduction: Hacking facial recognition
Chapter 1: A brief overview of artificial intelligence
Chapter 2: How AI is different from traditional software
Chapter 3: Data bias
Chapter 4: Hacking AI systems
Chapter 5: Evasion Attacks
Chapter 6: Data Poisoning
Chapter 7: Model Inversion (“Privacy”) Attacks
Chapter 8: Obfuscation attacks
Chapter 9: Talking to AI: Model interpretability
Chapter 10: Machine vs. machine
Chapter 11: Will someone hack my AI?
About the Author
An important, timely book that focuses on the practical risks of operationalizing AI. The history of disruptive technology is one of action-counteraction. AI is no different. Davey offers an excellent Threat Model for getting ahead of the AI risk problem. . .before it’s too late! — Lt Gen John (Jack) N. T. Shanahan, USAF (Ret.); inaugural director, Project Maven and inaugural director, Department of Defense Joint AI Center (JAIC)