Artificial intelligence and machine learning are increasingly being used and trusted by organizations to automate security threat detection. In this talk I will share a few insights on how AI can have the opposite result.

2020 was a year of reckoning when it came to racial bias and inequality. There is a long way to go on the human front, to be sure, but there’s another problem that still hasn’t been addressed to nearly the extent warranted: AI bias. Humans may assume that software and systems are not racial, but that simply isn’t true. Software is written by humans. The systems are running the software. The systems learn from the trainer.

We will explore the consequences of prejudiced, distorted, inaccurate and/or incomplete data on AI and machine learning models, and how it can increase false positives and overlook vulnerabilities that matter and are actively used by bad actors.

I will share a few experiences and lessons learned from building and working with AI models, and present best practices for ensuring data integrity, as well as verifying outputs to detect and correct bias. We will also discuss how the moats and perimeters of the past are becoming less relevant to succeed in 2022, businesses will have to take on the responsibility of detecting the vulnerabilities that will impact them.