← Back to Blog
ResearchJune 29, 2026By Sanjay Sharma· 5 min read

SERAPH — Detecting advanced malware using pre-trained LLMs and behavioural attribution

SERAPH is a novel framework leveraging pre-trained language models to detect advanced malware through semantic analysis and behavioral attribution.

This post summarizes the architecture, the multi-source feature pipeline, and how SHAP-based attribution makes detections explainable to analysts.