Radware’s proprietary Intent-based Deep Behavior Analysis (IDBA) performs behavioral analysis at a higher level of abstraction of ‘intent’ unlike commonly used interaction-based behavior analysis. IDBA consists of three stages: intent encoding, intent analysis, and adaptive learning. During adaptive learning, we also apply challenge-response authentication that helps us dynamically improve our machine learning models.
Capturing intent enables IDBA to provide significantly higher levels of accuracy while detecting bots with advanced human-like interaction capabilities. IDBA builds upon Radware Bot Manager's research findings in semi-supervised machine learning and leverages the latest developments in deep learning.