Evaluation of Cyber Deception Using Deep Learning Algorithms

Abstract

A machine learning-based approach is proposed and actualized to measure cyber deceptive defenses with negligible human inclusion. This dodges obstructions related to deceptive examination on humans, amplifying robotized assessment's adequacy before human subject’s research must be attempted. Utilizing ongoing advances in profound learning, the methodology synthesizes realistic, interactive, and adaptive traffic for utilization by target web services. A contextual analysis applies how to assess an interruption identification framework furnished with application layer embedded deceptive reactions to attacks. Results exhibit that blending adaptive web traffic bound with hesitant attacks controlled by outfit learning, online adaptive metric learning, and novel class discovery to recreate able enemies comprises a forceful and challenging test of cyber deceptive defenses.