Machine Learning and the Cyberz: Separating Fact from Marketing Fiction

No ratings

Presented at bsideswinnipeg 2017 by

Brandon Enright

The latest marketing fad and hype in our industry is the pushing of machine learning as a magic panacea for all cyber ills. We're told that by utilizing artistically crafted pixie dust and rainbow unicorn algorithms all your cyber woes can cured. Unfortunately, otherwise rational people in our industry seem to give a lot of weight to these marketing claims because 1) machine learning is complicated and 2) we've seen machine learning do some really incredible things like drive cars, beat the best Go players, "dream" crazy images, etc. This talk will illustrate some of the hype we're being bombarded with, pull back the covers on what machine learning is and how it works, present some basic statistics (Bayesian) and show why security is a much harder problem for machine learning than we realize. Like all great cyber security thought leaders, Brandon Enright is a self-aggrandizing media REDACTED. When he isn't leading cyber infosec thoughts, he's leading cyber dreams. Brandon can factor large primes with nothing more than pencil and paper and even recite the digits of Pi in random order in arbitrary bases. Recognized as one of the great legends in machine learning, several of Brandon's machines have learned so much Harvard is working on granting them PhDs. Rumor has it that shortly after Elon Musk met Brandon, Elon bought a small island in the south pacific for shelter and revised his estimate of the coming singularity up a decade. Brandon is why Bruce Schneier has trouble sleeping at night.