June 23, 2018 at BSides Asheville (bsidesasheville.com)
Abstract: Machine learning and artificial intelligence are all the rage in cybersecurity (and seemingly every other industry). Despite the hype, machine learning is not a magic bullet for security problems. In fact, the security domain contains many unique challenges that make applying machine learning especially difficult. In this talk, I will discuss challenges including sourcing useful data, labeling data, the adversarial nature of security problems, and validation of machine learning models.