THREATBYTESTM research
Evaluation
Evaluation quantifies predictive accuracy, pre-execution detection, false positives, mitigation effectiveness, and latency, using confusion matrices, ROC and precision-recall curves, behavioural sequence analysis, and explainable SI.
Metrics include predictive accuracy, pre-execution detection rate, false positive rate, mitigation effectiveness, and decision latency under controlled, cyber-range, and community-submitted test conditions.
Quantitative analysis uses confusion matrices, ROC curves, and precision-recall metrics for detection and pre-execution inference. Cognitive layers are assessed with explainable SI (including SHAP and LIME) for transparency of autonomous mitigation. Behavioural sequence and anomaly scoring measure anticipation from incomplete signals. Results are benchmarked against reactive systems (e.g. Snort, Suricata) and existing hybrid ML baselines. Community submissions provide real-world validation of adaptive learning.