Abstract

This paper introduces TIPS: Threat Actor Informed Prioritization usingSecEncoder, a specialized language model for security. TIPS combines thestrengths of both encoder and decoder language models to detect and prioritizecompromised applications. By integrating threat actor intelligence, TIPSenhances the accuracy and relevance of its detections. Extensive experimentswith a real-world benchmark dataset of applications demonstrate TIPS's highefficacy, achieving an F-1 score of 0.90 in identifying malicious applications.Additionally, in real-world scenarios, TIPS significantly reduces the backlogof investigations for security analysts by 87%, thereby streamlining the threatresponse process and improving overall security posture.

 

Quick Read (beta)