Anna Bertiger
Dr. Anna Bertiger uses math to find villains in computer networks as an ML researcher at Sublime Security working on email security. Before Sublime, she worked on post breach security and identity security at Microsoft and was a postdoctoral fellow in the Department of Combinatorics and Optimization at the University of Waterloo.
Session
Email bombs deluge a mailbox with a huge amount of email in a short time, rendering it useless and enabling some form of attack, perhaps in the form of a fake IT help call scam. They are nearly impossible to detect one message at a time, since the extra messages are often legitimate mail in other contexts. This talk will present a mailbox level detection method that uses unsupervised machine learning to detect anomalous mailbox volumes likely to correlate with email bombs for further detection on individual messages to separate the shrapnel from the email bomb from malicious messages and business as usual. In addition to learning about email bombs, attendees will learn how to apply similar anomaly detection to other scenarios where a large volume is likely to correlate with evil behavior.