Ransomware Detection on Ransomware Detection Dataset Behavioral and Static
99.15AccuracyDDQN + Asym. R2 (4:1)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DDQN + Asym. R2 (4:1)Feature Type=Behavioral (runtime), Cost-Aware Training=Embedded in reward, Statistical Validation=Friedman + Wilcoxon, Security Selection=SOMS (security-first)2026.07 | 99.15 | 99.15 | 0.8 | |
| RF, GB TreesFeature Type=Behavioral (dynamic), Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 99 | 99 | — | |
| BERT/RoBERTaFeature Type=API call sequences, Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 99 | 99 | 1 | |
| API call sequencesFeature Type=No, Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 98.5 | 98.5 | 1.5 | |
| Cost-sensitive Pareto EnsembleFeature Type=Behavioral + static, Cost-Aware Training=Post-hoc only, Statistical Validation=No, Security Selection=No2026.07 | 98.5 | 98.5 | — | |
| XAI deep learningFeature Type=Behavioral (dynamic), Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 98.1 | 98 | — | |
| DRL (symmetric R)Feature Type=Static (PE header), Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 97.6 | 97.6 | — | |
| RF (DNA seq.)Feature Type=Behavioral API calls, Cost-Aware Training=No, Statistical Validation=No, Security Selection=No2026.07 | 97.3 | — | — |