Over the previous decade, we’ve developed our strategy to translate the concept of red teaming to the newest improvements in know-how, together with AI. The AI Purple Staff is carefully aligned with conventional crimson groups, but additionally has the required AI material experience to hold out complicated technical assaults on AI techniques. To make sure that they’re simulating real looking adversary actions, our group leverages the newest insights from world class Google Menace Intelligence groups like Mandiant and the Threat Analysis Group (TAG), content material abuse crimson teaming in Belief & Security, and analysis into the newest assaults from Google DeepMind.

Frequent forms of crimson group assaults on AI techniques

One of many key obligations of Google’s AI Purple Staff is to take related analysis and adapt it to work in opposition to actual merchandise and options that use AI to study their affect. Workouts can increase findings throughout safety, privateness, and abuse disciplines, relying on the place and the way the know-how is deployed. To determine these alternatives to enhance security, we leverage attackers’ ways, methods and procedures (TTPs) to check a spread of system defenses. In in the present day’s report, there’s a record of TTPs that we contemplate most related and real looking for actual world adversaries and crimson teaming workout routines. They embody immediate assaults, coaching information extraction, backdooring the mannequin, adversarial examples, information poisoning and exfiltration.

Leave a Reply

Your email address will not be published. Required fields are marked *