ADAM: Active Defense Algorithm and Model Sergio Caltagirone University of Idaho
Active Defense “Any action sequence performed by an individual or organization between the time an attack is detected and has completed, in an automated or non-automated fashion, to mitigate a threat against a particular asset.” More than hacking back! Firewall rules, Notifying Authorities, etc. (along with the other stuff)
Goals of ADAM Provide a generalizable, extendable model for any organization Completely model the risk of the threat and AD actions Find best active defense solution for the threat (allow for automation) – maximize benefit, minimize risk Provide legal (and ethical) due diligence Why? Current tools are inefficient and sometimes critically ineffective If you want to respond to an attack, no way to determine which response is best
Active Defense Problems Ethicalness Legal Unintended Consequences Risk Valuation
Solutions Provided by ADAM Ethicalness Incorporates Teleological and Deontological ethical concerns Legal No precedent: minimal force, proportional force, immediate threat Unintended Consequences Statistical measure of confidence in action performing as expected Risk Valuation Provides statistical bounds for potential risk
Future and Upcoming Work Current: (For Fun) Using competitive co- evolution to determine effective active defense strategies Near Future (2-3 mo): Simulate Model for validation Far Future (4-5 mo): Formal validation