Architecture

flowchart TB
  subgraph Data
    CIC["CIC-IDS2017 MachineLearningCVE"] --> L["loaders + dedup + sampling"]
    UNSW["UNSW-NB15 official split"] --> L
    L --> S[("Schema: controllable / fixed / derived / relations")]
  end
  S --> P["log1p + standardise"]
  P --> ENS["Ensemble: XGBoost + MLP  OR  benign autoencoder"]
  P --> BL["Baselines: LogReg, RF, IsolationForest"]
  subgraph RedTeam["Red team"]
    PGD["white-box PGD"] --> ADA["adaptive: transfer + black-box search"]
    RS["score-based random search"] --> ADA
    ADA -->|"project every step"| S
    POI["backdoor poisoning, 2 %"]
  end
  ENS --> ADA
  subgraph Harden
    AT["adversarial training"]
    RF2["robust-feature model"]
    AID["adversarial-input alarm"]
    SAN["robust-feature kNN sanitiser"]
  end
  ADA --> AT --> ENS2["hardened ensemble"]
  POI --> SAN
  ENS --> X["TreeSHAP + constraint-valid counterfactuals"]
  ENS2 --> EV["robustness curves, per-family, base-rate precision, drift"]
  AID --> EV
  EV --> R["report.md / report.json / PNG"]
  ENS2 --> API["FastAPI: /score with explanations"]

stage	what FEINT does	code
Constraints	Per-dataset schema: up (only increase: duration, packets, payload), free (TCP window, TTL), integer, fixed (server side, dst port, context), derived (recomputed); relations bytes <= pkts*MTU, mean <= max_len <= min(bytes, MTU). project() + is_valid().	schema.py, ADR 0001
Detect	XGBoost, MLP (analytic input gradients), benign-trained autoencoder, ensemble = max(mean(XGB, MLP), AE); LogReg / RF / IsolationForest baselines	model.py
Evade	White-box PGD; multi-scale score-based random search; adaptive = PGD on target or surrogate, refined black-box, best per flow; curves are monotone (best over all budgets up to eps)	attack.py, ADR 0003
Poison	Copies of attack flows with a trigger in an attacker-controllable feature, labelled benign; sanitiser flags benign-labelled points whose neighbours in robust-feature space are malicious	poison.py
Harden	Adversarial training (any model, any attack), robust-feature model, adversarial-input alarm evaluated against a detector-aware attacker	harden.py
Explain	Exact TreeSHAP via XGBoost; counterfactual = minimal-budget constrained evasion (bisection) + greedy sparsification, reported in raw units	explain.py, ADR 0005
Report	Clean metrics, PR-AUC, precision at 1 % / 0.1 % prevalence, curves, per-family, feature exploitability, drift	metrics.py, pipeline.py, report.py
Serve	feint serve: /score returns verdict, top SHAP features and a counterfactual per flow	api.py

Example counterfactuals produced for real CIC-IDS2017 alerts:

flip to benign by: init_win_fwd 1024 -> 901
flip to benign by: fwd_pkts 9 -> 10; init_win_fwd 29200 -> 30150
no realisable evasion within budget: this alert is robust to the modelled attacker

Module map

flowchart TD
  cli["cli.py"] --> pipeline["pipeline.py (run_study)"]
  cli --> seeds["seeds.py (multi-seed CIs)"]
  cli --> api["api.py (FastAPI)"]
  pipeline --> data["data.py"]
  pipeline --> model["model.py"]
  pipeline --> attack["attack.py"]
  pipeline --> harden["harden.py"]
  pipeline --> explain["explain.py"]
  pipeline --> poison["poison.py"]
  pipeline --> metrics["metrics.py"]
  pipeline --> report["report.py"]
  seeds --> pipeline
  data --> schema["schema.py"]
  attack --> schema