TemporalDrift-ETM

Concept Drift-Aware Malware Evolution Tracking and Adaptive Self-Retraining in Network Flow Classification

Pipeline steps

  1. Upload a .csv of network flows
  2. Preprocess: any column count to 35 features
  3. Model classifies each flow
  4. TMFD drift detection runs per family
  5. If drift found: evolve profiles then self-retrain
  6. Reports are shown in both output windows

CSV format: any NTLFlowLyzer-compatible CSV. Columns not used by the model are ignored. Missing model features are filled with 0.

Drift threshold: TMFD > 0.30

Per-Sample Predictions

TemporalDrift-ETM - BSc Thesis - Dept. of CSE - Rangamati Science and Technology University Supervised by Md Mynoddin, Assistant Professor | Student ID: 2001011025