A Python library for Secure and Explainable Machine Learning

Detects concept and model drift in DNS traffic using ML, analyzes attack recall collapse, engages alarm for threshold drop, and compares retraining feasibility in a SOC detection environment.

Evaluates LLM safety failure modes across prompt attacks, context overflow, and RAG poisoning.

Unsupervised anomaly detection model trained on process level endpoint telemetry (BETH dataset) and Isolation Forests to study malicious events detection, false positives, and SOC implementation.