AI powered attacks against AI powered protection: classification, scenarios and risk analysis
Olena Veprytska, Vyacheslav Kharchenko
Abstract
The research focuses on the detection, classification, and analysis of artificial intelligence (AI)-based threats/attacks on various types of systems. AI is analyzed as an object of protection (OP), a means of attack (MA) and a means of protection (MP). Duel chain and set of scenarios for triad “MA-MP-OP” considering application AI was developed. As part of the IMECA analysis, the threats, vulnerabilities, effects and criticality (probability x severity) of the attacks are determined. The criticality matrixes of cyber risks of systems (initial, after implementing countermeasures with and without AI means) were built. The possibilities and efficiency of using AI to prevent some mentioned attacks.
Topics & Concepts
CriticalityComputer scienceComputer securityTriad (sociology)Set (abstract data type)Failure mode, effects, and criticality analysisArtificial intelligenceProgramming languageNuclear physicsPhysicsPsychoanalysisPsychologyCybersecurity and Information SystemsAdvanced Data Processing TechniquesNetwork Security and Intrusion Detection