feat: clustering multi-métriques + TCP fingerprinting amélioré
- TCP fingerprinting: 20 signatures OS (p0f-style), scoring multi-signal
TTL/MSS/scale/fenêtre, détection Masscan 97% confiance, réseau path
(Ethernet/PPPoE/VPN/Tunnel), estimation hop-count
- Clustering IPs: K-means++ (Arthur & Vassilvitskii 2007) sur 21 features
TCP stack + anomalie ML + TLS/protocole + navigateur + temporel
PCA-2D par puissance itérative (Hotelling) pour positionnement
- Visualisation redesign: 2 vues lisibles
- Tableau de bord: grille de cartes groupées par niveau de risque
(Bots / Suspects / Légitimes), métriques clés + mini-barres
- Graphe de relations: ReactFlow avec nœuds-cartes en colonnes
par niveau de menace, arêtes colorées par similarité, légende
- Sidebar: RadarChart comportemental + toutes métriques + export CSV
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
SOC Analyst
@ -13,7 +13,7 @@ import os
|
||||
from .config import settings
|
||||
from .database import db
|
||||
from .routes import metrics, detections, variability, attributes, analysis, entities, incidents, audit, reputation, fingerprints
|
||||
from .routes import bruteforce, tcp_spoofing, header_fingerprint, heatmap, botnets, rotation, ml_features, investigation_summary, search
|
||||
from .routes import bruteforce, tcp_spoofing, header_fingerprint, heatmap, botnets, rotation, ml_features, investigation_summary, search, clustering
|
||||
|
||||
# Configuration logging
|
||||
logging.basicConfig(
|
||||
@ -84,6 +84,7 @@ app.include_router(rotation.router)
|
||||
app.include_router(ml_features.router)
|
||||
app.include_router(investigation_summary.router)
|
||||
app.include_router(search.router)
|
||||
app.include_router(clustering.router)
|
||||
|
||||
|
||||
# Route pour servir le frontend
|
||||
|
||||
458
backend/routes/clustering.py
Normal file
458
backend/routes/clustering.py
Normal file
@ -0,0 +1,458 @@
|
||||
"""
|
||||
Clustering d'IPs multi-métriques — backend ReactFlow.
|
||||
|
||||
Features utilisées (21 dimensions) :
|
||||
TCP stack : TTL initial, MSS, scale, fenêtre TCP
|
||||
Comportement : vélocité, POST ratio, fuzzing, assets, accès direct
|
||||
Anomalie ML : score, IP-ID zéro
|
||||
TLS/Protocole: ALPN mismatch, ALPN absent, efficacité H2
|
||||
Navigateur : browser score, headless, ordre headers, UA-CH mismatch
|
||||
Temporel : entropie, diversité JA4, UA rotatif
|
||||
|
||||
Algorithme :
|
||||
1. Échantillonnage stratifié (top détections + top hits)
|
||||
2. Construction + normalisation des vecteurs de features
|
||||
3. K-means++ (Arthur & Vassilvitskii, 2007)
|
||||
4. PCA-2D par power iteration pour les positions ReactFlow
|
||||
5. Nommage automatique par features dominantes du centroïde
|
||||
6. Calcul des arêtes : k-NN dans l'espace des features
|
||||
"""
|
||||
from __future__ import annotations
|
||||
import math
|
||||
import time
|
||||
import hashlib
|
||||
from typing import Optional
|
||||
from fastapi import APIRouter, HTTPException, Query
|
||||
|
||||
from ..database import db
|
||||
from ..services.clustering_engine import (
|
||||
FEATURES, FEATURE_KEYS, FEATURE_NORMS, FEATURE_NAMES, N_FEATURES,
|
||||
build_feature_vector, kmeans_pp, pca_2d,
|
||||
name_cluster, risk_score_from_centroid, _mean_vec,
|
||||
)
|
||||
|
||||
router = APIRouter(prefix="/api/clustering", tags=["clustering"])
|
||||
|
||||
# ─── Cache en mémoire ─────────────────────────────────────────────────────────
|
||||
# Stocke (cluster_id → liste d'IPs) pour le drill-down
|
||||
# + timestamp de dernière mise à jour
|
||||
_cache: dict = {
|
||||
"assignments": {}, # ip+ja4 → cluster_idx
|
||||
"cluster_ips": {}, # cluster_idx → [(ip, ja4)]
|
||||
"params": {}, # k, ts
|
||||
}
|
||||
|
||||
# ─── Couleurs ─────────────────────────────────────────────────────────────────
|
||||
_THREAT_COLOR = {
|
||||
0.92: "#dc2626", # Bot scanner
|
||||
0.70: "#ef4444", # Critique
|
||||
0.45: "#f97316", # Élevé
|
||||
0.25: "#eab308", # Modéré
|
||||
0.00: "#6b7280", # Sain / inconnu
|
||||
}
|
||||
|
||||
def _risk_to_color(risk: float) -> str:
|
||||
for threshold, color in sorted(_THREAT_COLOR.items(), reverse=True):
|
||||
if risk >= threshold:
|
||||
return color
|
||||
return "#6b7280"
|
||||
|
||||
|
||||
# ─── SQL ──────────────────────────────────────────────────────────────────────
|
||||
|
||||
_SQL_FEATURES = """
|
||||
SELECT
|
||||
replaceRegexpAll(toString(t.src_ip), '^::ffff:', '') AS ip,
|
||||
t.ja4,
|
||||
any(t.tcp_ttl_raw) AS ttl,
|
||||
any(t.tcp_win_raw) AS win,
|
||||
any(t.tcp_scale_raw) AS scale,
|
||||
any(t.tcp_mss_raw) AS mss,
|
||||
any(t.first_ua) AS ua,
|
||||
sum(t.hits) AS hits,
|
||||
|
||||
avg(abs(ml.anomaly_score)) AS avg_score,
|
||||
avg(ml.hit_velocity) AS avg_velocity,
|
||||
avg(ml.fuzzing_index) AS avg_fuzzing,
|
||||
avg(ml.is_headless) AS pct_headless,
|
||||
avg(ml.post_ratio) AS avg_post,
|
||||
avg(ml.ip_id_zero_ratio) AS ip_id_zero,
|
||||
avg(ml.temporal_entropy) AS entropy,
|
||||
avg(ml.modern_browser_score) AS browser_score,
|
||||
avg(ml.alpn_http_mismatch) AS alpn_mismatch,
|
||||
avg(ml.is_alpn_missing) AS alpn_missing,
|
||||
avg(ml.multiplexing_efficiency) AS h2_eff,
|
||||
avg(ml.header_order_confidence) AS hdr_conf,
|
||||
avg(ml.ua_ch_mismatch) AS ua_ch_mismatch,
|
||||
avg(ml.asset_ratio) AS asset_ratio,
|
||||
avg(ml.direct_access_ratio) AS direct_ratio,
|
||||
avg(ml.distinct_ja4_count) AS ja4_count,
|
||||
max(ml.is_ua_rotating) AS ua_rotating,
|
||||
|
||||
max(ml.threat_level) AS threat,
|
||||
any(ml.country_code) AS country,
|
||||
any(ml.asn_org) AS asn_org
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h t
|
||||
LEFT JOIN mabase_prod.ml_detected_anomalies ml
|
||||
ON t.src_ip = ml.src_ip AND t.ja4 = ml.ja4
|
||||
AND ml.detected_at >= now() - INTERVAL 24 HOUR
|
||||
WHERE t.window_start >= now() - INTERVAL 24 HOUR
|
||||
AND t.tcp_ttl_raw > 0
|
||||
GROUP BY t.src_ip, t.ja4
|
||||
ORDER BY
|
||||
-- Stratégie : IPs anormales en premier, puis fort trafic
|
||||
-- Cela garantit que les bots Masscan (anomalie=0.97, hits=1-2) sont inclus
|
||||
avg(abs(ml.anomaly_score)) DESC,
|
||||
sum(t.hits) DESC
|
||||
LIMIT %(limit)s
|
||||
"""
|
||||
|
||||
# Noms des colonnes SQL dans l'ordre
|
||||
_SQL_COLS = [
|
||||
"ip", "ja4", "ttl", "win", "scale", "mss", "ua", "hits",
|
||||
"avg_score", "avg_velocity", "avg_fuzzing", "pct_headless", "avg_post",
|
||||
"ip_id_zero", "entropy", "browser_score", "alpn_mismatch", "alpn_missing",
|
||||
"h2_eff", "hdr_conf", "ua_ch_mismatch", "asset_ratio", "direct_ratio",
|
||||
"ja4_count", "ua_rotating", "threat", "country", "asn_org",
|
||||
]
|
||||
|
||||
|
||||
# ─── Endpoints ────────────────────────────────────────────────────────────────
|
||||
|
||||
@router.get("/clusters")
|
||||
async def get_clusters(
|
||||
k: int = Query(14, ge=4, le=30, description="Nombre de clusters"),
|
||||
n_samples: int = Query(3000, ge=500, le=8000, description="Taille de l'échantillon"),
|
||||
):
|
||||
"""
|
||||
Clustering multi-métriques des IPs.
|
||||
|
||||
Retourne les nœuds (clusters) + arêtes pour ReactFlow, avec :
|
||||
- positions 2D issues de PCA sur les 21 features
|
||||
- profil radar des features par cluster (normalisé [0,1])
|
||||
- statistiques détaillées (moyennes brutes des features)
|
||||
- sample d'IPs représentatives
|
||||
"""
|
||||
t0 = time.time()
|
||||
try:
|
||||
result = db.query(_SQL_FEATURES, {"limit": n_samples})
|
||||
except Exception as e:
|
||||
raise HTTPException(status_code=500, detail=f"ClickHouse: {e}")
|
||||
|
||||
# ── Construction des vecteurs de features ─────────────────────────────
|
||||
rows: list[dict] = []
|
||||
for row in result.result_rows:
|
||||
d = {col: row[i] for i, col in enumerate(_SQL_COLS)}
|
||||
rows.append(d)
|
||||
|
||||
if len(rows) < k:
|
||||
raise HTTPException(status_code=400, detail="Pas assez de données pour ce k")
|
||||
|
||||
points = [build_feature_vector(r) for r in rows]
|
||||
|
||||
# ── K-means++ ────────────────────────────────────────────────────────
|
||||
km = kmeans_pp(points, k=k, max_iter=60, seed=42)
|
||||
|
||||
# ── PCA-2D sur les centroïdes ─────────────────────────────────────────
|
||||
# On projette les centroïdes dans l'espace PCA des données
|
||||
# → les positions relatives reflètent la variance des données
|
||||
coords_all = pca_2d(points)
|
||||
# Moyenne des positions PCA par cluster = position 2D du centroïde
|
||||
cluster_xs: list[list[float]] = [[] for _ in range(k)]
|
||||
cluster_ys: list[list[float]] = [[] for _ in range(k)]
|
||||
for i, label in enumerate(km.labels):
|
||||
cluster_xs[label].append(coords_all[i][0])
|
||||
cluster_ys[label].append(coords_all[i][1])
|
||||
|
||||
centroid_2d: list[tuple[float, float]] = []
|
||||
for j in range(k):
|
||||
if cluster_xs[j]:
|
||||
cx = sum(cluster_xs[j]) / len(cluster_xs[j])
|
||||
cy = sum(cluster_ys[j]) / len(cluster_ys[j])
|
||||
else:
|
||||
cx, cy = 0.5, 0.5
|
||||
centroid_2d.append((cx, cy))
|
||||
|
||||
# ── Agrégation des statistiques par cluster ───────────────────────────
|
||||
cluster_rows: list[list[dict]] = [[] for _ in range(k)]
|
||||
cluster_members: list[list[tuple[str, str]]] = [[] for _ in range(k)]
|
||||
for i, label in enumerate(km.labels):
|
||||
cluster_rows[label].append(rows[i])
|
||||
cluster_members[label].append((rows[i]["ip"], rows[i]["ja4"]))
|
||||
|
||||
# Mise à jour du cache pour le drill-down
|
||||
_cache["cluster_ips"] = {j: cluster_members[j] for j in range(k)}
|
||||
_cache["params"] = {"k": k, "ts": t0}
|
||||
|
||||
# ── Construction des nœuds ReactFlow ─────────────────────────────────
|
||||
CANVAS_W, CANVAS_H = 1400, 780
|
||||
|
||||
nodes = []
|
||||
for j in range(k):
|
||||
if not cluster_rows[j]:
|
||||
continue
|
||||
|
||||
# Statistiques brutes moyennées
|
||||
def avg_feat(key: str) -> float:
|
||||
vals = [float(r.get(key) or 0) for r in cluster_rows[j]]
|
||||
return sum(vals) / len(vals) if vals else 0.0
|
||||
|
||||
mean_ttl = avg_feat("ttl")
|
||||
mean_mss = avg_feat("mss")
|
||||
mean_scale = avg_feat("scale")
|
||||
mean_win = avg_feat("win")
|
||||
mean_score = avg_feat("avg_score")
|
||||
mean_vel = avg_feat("avg_velocity")
|
||||
mean_fuzz = avg_feat("avg_fuzzing")
|
||||
mean_hless = avg_feat("pct_headless")
|
||||
mean_post = avg_feat("avg_post")
|
||||
mean_asset = avg_feat("asset_ratio")
|
||||
mean_direct= avg_feat("direct_ratio")
|
||||
mean_alpn = avg_feat("alpn_mismatch")
|
||||
mean_h2 = avg_feat("h2_eff")
|
||||
mean_hconf = avg_feat("hdr_conf")
|
||||
mean_ua_ch = avg_feat("ua_ch_mismatch")
|
||||
mean_entr = avg_feat("entropy")
|
||||
mean_ja4 = avg_feat("ja4_count")
|
||||
mean_ip_id = avg_feat("ip_id_zero")
|
||||
mean_brow = avg_feat("browser_score")
|
||||
mean_uarot = avg_feat("ua_rotating")
|
||||
|
||||
ip_count = len(set(r["ip"] for r in cluster_rows[j]))
|
||||
hit_count = int(sum(float(r.get("hits") or 0) for r in cluster_rows[j]))
|
||||
|
||||
# Pays / ASN / Menace dominants
|
||||
threats = [str(r.get("threat") or "") for r in cluster_rows[j] if r.get("threat")]
|
||||
countries = [str(r.get("country") or "") for r in cluster_rows[j] if r.get("country")]
|
||||
orgs = [str(r.get("asn_org") or "") for r in cluster_rows[j] if r.get("asn_org")]
|
||||
|
||||
def topk(lst: list[str], n: int = 5) -> list[str]:
|
||||
from collections import Counter
|
||||
return [v for v, _ in Counter(lst).most_common(n) if v]
|
||||
|
||||
raw_stats = {
|
||||
"mean_ttl": mean_ttl, "mean_mss": mean_mss,
|
||||
"mean_scale": mean_scale,
|
||||
}
|
||||
label = name_cluster(km.centroids[j], raw_stats)
|
||||
risk = risk_score_from_centroid(km.centroids[j])
|
||||
color = _risk_to_color(risk)
|
||||
|
||||
# Profil radar normalisé (valeurs centroïde [0,1])
|
||||
radar = [
|
||||
{"feature": name, "value": round(km.centroids[j][i], 4)}
|
||||
for i, name in enumerate(FEATURE_NAMES)
|
||||
]
|
||||
|
||||
# Position 2D (PCA normalisée → pixels ReactFlow)
|
||||
px_x = centroid_2d[j][0] * CANVAS_W * 0.85 + 80
|
||||
px_y = (1 - centroid_2d[j][1]) * CANVAS_H * 0.85 + 50 # inverser y (haut=risque)
|
||||
|
||||
# Rayon ∝ √ip_count
|
||||
radius = max(18, min(90, int(math.sqrt(ip_count) * 0.3)))
|
||||
|
||||
# Sample IPs (top 8 par hits)
|
||||
sample_rows = sorted(cluster_rows[j], key=lambda r: float(r.get("hits") or 0), reverse=True)[:8]
|
||||
sample_ips = [r["ip"] for r in sample_rows]
|
||||
sample_ua = str(cluster_rows[j][0].get("ua") or "")
|
||||
|
||||
cluster_id = f"c{j}_k{k}"
|
||||
|
||||
nodes.append({
|
||||
"id": cluster_id,
|
||||
"label": label,
|
||||
"cluster_idx": j,
|
||||
"x": round(px_x, 1),
|
||||
"y": round(px_y, 1),
|
||||
"radius": radius,
|
||||
"color": color,
|
||||
"risk_score": risk,
|
||||
|
||||
# Caractéristiques TCP
|
||||
"mean_ttl": round(mean_ttl, 1),
|
||||
"mean_mss": round(mean_mss, 0),
|
||||
"mean_scale": round(mean_scale, 1),
|
||||
"mean_win": round(mean_win, 0),
|
||||
|
||||
# Comportement HTTP
|
||||
"mean_score": round(mean_score, 4),
|
||||
"mean_velocity": round(mean_vel, 3),
|
||||
"mean_fuzzing": round(mean_fuzz, 3),
|
||||
"mean_headless": round(mean_hless, 3),
|
||||
"mean_post": round(mean_post, 3),
|
||||
"mean_asset": round(mean_asset, 3),
|
||||
"mean_direct": round(mean_direct, 3),
|
||||
|
||||
# TLS / Protocole
|
||||
"mean_alpn_mismatch": round(mean_alpn, 3),
|
||||
"mean_h2_eff": round(mean_h2, 3),
|
||||
"mean_hdr_conf": round(mean_hconf, 3),
|
||||
"mean_ua_ch": round(mean_ua_ch, 3),
|
||||
|
||||
# Temporel
|
||||
"mean_entropy": round(mean_entr, 3),
|
||||
"mean_ja4_diversity": round(mean_ja4, 3),
|
||||
"mean_ip_id_zero": round(mean_ip_id, 3),
|
||||
"mean_browser_score": round(mean_brow, 1),
|
||||
"mean_ua_rotating": round(mean_uarot, 3),
|
||||
|
||||
# Meta
|
||||
"ip_count": ip_count,
|
||||
"hit_count": hit_count,
|
||||
"top_threat": topk(threats, 1)[0] if topk(threats, 1) else "",
|
||||
"top_countries": topk(countries, 5),
|
||||
"top_orgs": topk(orgs, 5),
|
||||
"sample_ips": sample_ips,
|
||||
"sample_ua": sample_ua,
|
||||
|
||||
# Profil radar pour visualisation
|
||||
"radar": radar,
|
||||
})
|
||||
|
||||
# ── Arêtes : k-NN dans l'espace des features ──────────────────────────
|
||||
# Chaque cluster est connecté à ses 2 voisins les plus proches
|
||||
edges = []
|
||||
seen: set[frozenset] = set()
|
||||
centroids = km.centroids
|
||||
|
||||
for i, ni in enumerate(nodes):
|
||||
ci = ni["cluster_idx"]
|
||||
# Distance² aux autres centroïdes
|
||||
dists = [
|
||||
(j, nj["cluster_idx"],
|
||||
sum((centroids[ci][d] - centroids[nj["cluster_idx"]][d]) ** 2
|
||||
for d in range(N_FEATURES)))
|
||||
for j, nj in enumerate(nodes) if j != i
|
||||
]
|
||||
dists.sort(key=lambda x: x[2])
|
||||
# 2 voisins les plus proches
|
||||
for j, cj, dist2 in dists[:2]:
|
||||
key = frozenset([ni["id"], nodes[j]["id"]])
|
||||
if key in seen:
|
||||
continue
|
||||
seen.add(key)
|
||||
similarity = round(1.0 / (1.0 + math.sqrt(dist2)), 3)
|
||||
edges.append({
|
||||
"id": f"e_{ni['id']}_{nodes[j]['id']}",
|
||||
"source": ni["id"],
|
||||
"target": nodes[j]["id"],
|
||||
"similarity": similarity,
|
||||
"weight": round(similarity * 5, 1),
|
||||
})
|
||||
|
||||
# ── Stats globales ────────────────────────────────────────────────────
|
||||
total_ips = sum(n["ip_count"] for n in nodes)
|
||||
total_hits = sum(n["hit_count"] for n in nodes)
|
||||
bot_ips = sum(n["ip_count"] for n in nodes if n["risk_score"] > 0.40 or "🤖" in n["label"])
|
||||
high_risk = sum(n["ip_count"] for n in nodes if n["risk_score"] > 0.20)
|
||||
|
||||
elapsed = round(time.time() - t0, 2)
|
||||
|
||||
return {
|
||||
"nodes": nodes,
|
||||
"edges": edges,
|
||||
"stats": {
|
||||
"total_clusters": len(nodes),
|
||||
"total_ips": total_ips,
|
||||
"total_hits": total_hits,
|
||||
"bot_ips": bot_ips,
|
||||
"high_risk_ips": high_risk,
|
||||
"n_samples": len(rows),
|
||||
"k": k,
|
||||
"elapsed_s": elapsed,
|
||||
},
|
||||
"feature_names": FEATURE_NAMES,
|
||||
}
|
||||
|
||||
|
||||
@router.get("/cluster/{cluster_id}/ips")
|
||||
async def get_cluster_ips(
|
||||
cluster_id: str,
|
||||
limit: int = Query(100, ge=1, le=500),
|
||||
offset: int = Query(0, ge=0),
|
||||
):
|
||||
"""
|
||||
IPs appartenant à un cluster (depuis le cache de la dernière exécution).
|
||||
Si le cache est expiré, retourne une erreur guidant vers /clusters.
|
||||
"""
|
||||
if not _cache.get("cluster_ips"):
|
||||
raise HTTPException(
|
||||
status_code=404,
|
||||
detail="Cache expiré — appelez /api/clustering/clusters d'abord"
|
||||
)
|
||||
|
||||
# Extrait l'index cluster depuis l'id (format: c{idx}_k{k})
|
||||
try:
|
||||
idx = int(cluster_id.split("_")[0][1:])
|
||||
except (ValueError, IndexError):
|
||||
raise HTTPException(status_code=400, detail="cluster_id invalide")
|
||||
|
||||
members = _cache["cluster_ips"].get(idx, [])
|
||||
if not members:
|
||||
return {"ips": [], "total": 0, "cluster_id": cluster_id}
|
||||
|
||||
total = len(members)
|
||||
page_members = members[offset: offset + limit]
|
||||
|
||||
# Requête SQL pour les détails de ces IPs spécifiques
|
||||
ip_list = [m[0] for m in page_members]
|
||||
ja4_list = [m[1] for m in page_members]
|
||||
|
||||
if not ip_list:
|
||||
return {"ips": [], "total": total, "cluster_id": cluster_id}
|
||||
|
||||
# On ne peut pas facilement passer une liste en paramètre ClickHouse —
|
||||
# on la construit directement (valeurs nettoyées)
|
||||
safe_ips = [ip.replace("'", "") for ip in ip_list[:100]]
|
||||
ip_filter = ", ".join(f"'{ip}'" for ip in safe_ips)
|
||||
|
||||
sql = f"""
|
||||
SELECT
|
||||
replaceRegexpAll(toString(t.src_ip), '^::ffff:', '') AS src_ip,
|
||||
t.ja4,
|
||||
any(t.tcp_ttl_raw) AS ttl,
|
||||
any(t.tcp_win_raw) AS win,
|
||||
any(t.tcp_scale_raw) AS scale,
|
||||
any(t.tcp_mss_raw) AS mss,
|
||||
sum(t.hits) AS hits,
|
||||
any(t.first_ua) AS ua,
|
||||
round(avg(abs(ml.anomaly_score)), 3) AS avg_score,
|
||||
max(ml.threat_level) AS threat_level,
|
||||
any(ml.country_code) AS country_code,
|
||||
any(ml.asn_org) AS asn_org,
|
||||
round(avg(ml.fuzzing_index), 2) AS fuzzing,
|
||||
round(avg(ml.hit_velocity), 2) AS velocity
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h t
|
||||
LEFT JOIN mabase_prod.ml_detected_anomalies ml
|
||||
ON t.src_ip = ml.src_ip AND t.ja4 = ml.ja4
|
||||
AND ml.detected_at >= now() - INTERVAL 24 HOUR
|
||||
WHERE t.window_start >= now() - INTERVAL 24 HOUR
|
||||
AND replaceRegexpAll(toString(t.src_ip), '^::ffff:', '') IN ({ip_filter})
|
||||
GROUP BY t.src_ip, t.ja4
|
||||
ORDER BY hits DESC
|
||||
"""
|
||||
try:
|
||||
result = db.query(sql)
|
||||
except Exception as e:
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
|
||||
ips = []
|
||||
for row in result.result_rows:
|
||||
ips.append({
|
||||
"ip": str(row[0]),
|
||||
"ja4": str(row[1] or ""),
|
||||
"tcp_ttl": int(row[2] or 0),
|
||||
"tcp_win": int(row[3] or 0),
|
||||
"tcp_scale": int(row[4] or 0),
|
||||
"tcp_mss": int(row[5] or 0),
|
||||
"hits": int(row[6] or 0),
|
||||
"ua": str(row[7] or ""),
|
||||
"avg_score": float(row[8] or 0),
|
||||
"threat_level": str(row[9] or ""),
|
||||
"country_code": str(row[10] or ""),
|
||||
"asn_org": str(row[11] or ""),
|
||||
"fuzzing": float(row[12] or 0),
|
||||
"velocity": float(row[13] or 0),
|
||||
})
|
||||
|
||||
return {"ips": ips, "total": total, "cluster_id": cluster_id}
|
||||
@ -7,6 +7,7 @@ agg_host_ip_ja4_1h (rotation JA4), view_ip_recurrence, view_ai_features_1h.
|
||||
from fastapi import APIRouter, HTTPException
|
||||
|
||||
from ..database import db
|
||||
from ..services.tcp_fingerprint import fingerprint_os, detect_spoof, declared_os_from_ua
|
||||
|
||||
router = APIRouter(prefix="/api/investigation", tags=["investigation"])
|
||||
|
||||
@ -62,32 +63,45 @@ async def get_ip_full_summary(ip: str):
|
||||
"top_hosts": [str(h) for h in (bf_row[3] or [])] if bf_row else [],
|
||||
}
|
||||
|
||||
# ── 3. TCP spoofing ────────────────────────────────────────────────────
|
||||
# ── 3. TCP spoofing — fingerprinting multi-signal ─────────────────────
|
||||
tcp_sql = """
|
||||
SELECT tcp_ttl, first_ua
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
SELECT
|
||||
any(tcp_ttl_raw) AS ttl,
|
||||
any(tcp_win_raw) AS win,
|
||||
any(tcp_scale_raw) AS scale,
|
||||
any(tcp_mss_raw) AS mss,
|
||||
any(first_ua) AS ua
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE replaceRegexpAll(toString(src_ip), '^::ffff:', '') = %(ip)s
|
||||
AND tcp_ttl > 0
|
||||
AND window_start >= now() - INTERVAL 24 HOUR
|
||||
AND tcp_ttl_raw > 0
|
||||
LIMIT 1
|
||||
"""
|
||||
tcp_res = db.query(tcp_sql, {"ip": clean_ip})
|
||||
tcp_data = {"detected": False, "tcp_ttl": None, "suspected_os": None}
|
||||
if tcp_res.result_rows:
|
||||
ttl = int(tcp_res.result_rows[0][0])
|
||||
if 52 <= ttl <= 65:
|
||||
sus_os = "Linux/Mac"
|
||||
elif 110 <= ttl <= 135:
|
||||
sus_os = "Windows"
|
||||
else:
|
||||
sus_os = "Unknown"
|
||||
ua = str(tcp_res.result_rows[0][1] or "")
|
||||
dec_os = "Windows" if "Windows" in ua else ("macOS" if "Mac OS X" in ua else "Linux/Android" if "Linux" in ua else "Unknown")
|
||||
spoof = sus_os != "Unknown" and dec_os != "Unknown" and sus_os != dec_os
|
||||
r = tcp_res.result_rows[0]
|
||||
ttl = int(r[0] or 0)
|
||||
win = int(r[1] or 0)
|
||||
scale = int(r[2] or 0)
|
||||
mss = int(r[3] or 0)
|
||||
ua = str(r[4] or "")
|
||||
fp = fingerprint_os(ttl, win, scale, mss)
|
||||
dec_os = declared_os_from_ua(ua)
|
||||
spoof_res = detect_spoof(fp, dec_os)
|
||||
tcp_data = {
|
||||
"detected": spoof,
|
||||
"tcp_ttl": ttl,
|
||||
"suspected_os": sus_os,
|
||||
"declared_os": dec_os,
|
||||
"detected": spoof_res.is_spoof,
|
||||
"tcp_ttl": ttl,
|
||||
"tcp_mss": mss,
|
||||
"tcp_win_scale": scale,
|
||||
"initial_ttl": fp.initial_ttl,
|
||||
"hop_count": fp.hop_count,
|
||||
"suspected_os": fp.os_name,
|
||||
"declared_os": dec_os,
|
||||
"confidence": fp.confidence,
|
||||
"network_path": fp.network_path,
|
||||
"is_bot_tool": fp.is_bot_tool,
|
||||
"spoof_reason": spoof_res.reason,
|
||||
}
|
||||
|
||||
# ── 4. JA4 rotation ────────────────────────────────────────────────────
|
||||
@ -146,7 +160,9 @@ async def get_ip_full_summary(ip: str):
|
||||
risk = 0
|
||||
risk += min(50, ml_data["max_score"] * 50)
|
||||
if bf_data["active"]: risk += 20
|
||||
if tcp_data["detected"]: risk += 15
|
||||
if tcp_data["detected"]:
|
||||
if tcp_data.get("is_bot_tool"): risk += 30 # outil de scan connu
|
||||
else: risk += 15 # spoof OS
|
||||
if rot_data["rotating"]: risk += min(15, rot_data["distinct_ja4_count"] * 3)
|
||||
if pers_data["persistent"]: risk += min(10, pers_data["recurrence"] * 2)
|
||||
risk = min(100, round(risk))
|
||||
|
||||
@ -1,130 +1,95 @@
|
||||
"""
|
||||
Endpoints pour la détection du TCP spoofing (TTL / window size anormaux)
|
||||
Endpoints pour la détection du TCP spoofing / fingerprinting OS
|
||||
|
||||
Règle de corrélation :
|
||||
- TTL=0 ou tcp_window_size=0 → données TCP absentes (proxy/LB) → pas de corrélation possible
|
||||
- TTL 55-65 → fingerprint Linux/Mac (initial TTL 64)
|
||||
- TTL 120-135 → fingerprint Windows (initial TTL 128)
|
||||
- TTL 110-120 → fingerprint Windows (initial TTL 128, quelques sauts)
|
||||
- Toute autre valeur → OS indéterminé → pas de flag spoofing
|
||||
- spoof_flag = True UNIQUEMENT si OS fingerprinting TCP possible ET incompatible avec l'UA
|
||||
Approche multi-signal (p0f-style) :
|
||||
- TTL initial estimé → famille OS (Linux/Mac=64, Windows=128, Cisco/BSD=255)
|
||||
- MSS → type de réseau (Ethernet=1460, PPPoE=1452, VPN=1380-1420)
|
||||
- Taille de fenêtre → signature OS précise
|
||||
- Facteur d'échelle → affine la version kernel/stack TCP
|
||||
|
||||
Détection bots : signatures connues (Masscan/ZMap/Mirai) identifiées par combinaison
|
||||
win+scale+mss indépendamment de l'UA.
|
||||
"""
|
||||
from fastapi import APIRouter, HTTPException, Query
|
||||
|
||||
from ..database import db
|
||||
from ..services.tcp_fingerprint import (
|
||||
fingerprint_os,
|
||||
detect_spoof,
|
||||
declared_os_from_ua,
|
||||
)
|
||||
|
||||
router = APIRouter(prefix="/api/tcp-spoofing", tags=["tcp_spoofing"])
|
||||
|
||||
# Plages TTL qui permettent une corrélation fiable
|
||||
_TTL_LINUX = (range(52, 66), "Linux/Mac") # initial 64, 1-12 sauts
|
||||
_TTL_WINDOWS = (range(110, 136), "Windows") # initial 128, 1-18 sauts
|
||||
_TTL_CISCO = (range(240, 256), "Cisco/BSD") # initial 255
|
||||
|
||||
|
||||
def _suspected_os(ttl: int) -> str:
|
||||
"""Retourne l'OS probable à partir du TTL observé.
|
||||
Retourne 'Unknown' si le TTL ne permet pas une corrélation fiable
|
||||
(TTL=0 = pas de données TCP, ou hors plage connue).
|
||||
"""
|
||||
if ttl <= 0:
|
||||
return "Unknown" # Pas de données TCP (proxy/CDN)
|
||||
for rng, name in (_TTL_LINUX, _TTL_WINDOWS, _TTL_CISCO):
|
||||
if ttl in rng:
|
||||
return name
|
||||
return "Unknown"
|
||||
|
||||
|
||||
def _declared_os(ua: str) -> str:
|
||||
ua = ua or ""
|
||||
if "Windows" in ua:
|
||||
return "Windows"
|
||||
if "Mac OS X" in ua:
|
||||
return "macOS"
|
||||
if "Linux" in ua or "Android" in ua:
|
||||
return "Linux/Android"
|
||||
return "Unknown"
|
||||
|
||||
|
||||
def _is_spoof(suspected_os: str, declared_os: str) -> bool:
|
||||
"""Spoof confirmé uniquement si on a un fingerprint TCP fiable ET une incompatibilité d'OS."""
|
||||
if suspected_os == "Unknown" or declared_os == "Unknown":
|
||||
return False # Pas de corrélation possible
|
||||
# Linux/Mac fingerprint TCP mais UA déclare Windows
|
||||
if suspected_os == "Linux/Mac" and declared_os == "Windows":
|
||||
return True
|
||||
# Windows fingerprint TCP mais UA déclare Linux/Android ou macOS
|
||||
if suspected_os == "Windows" and declared_os in ("Linux/Android", "macOS"):
|
||||
return True
|
||||
return False
|
||||
|
||||
|
||||
@router.get("/overview")
|
||||
async def get_tcp_spoofing_overview():
|
||||
"""Statistiques globales : seules les entrées avec données TCP valides sont analysées."""
|
||||
"""Statistiques globales avec fingerprinting multi-signal (TTL + MSS + fenêtre + scale)."""
|
||||
try:
|
||||
sql = """
|
||||
SELECT
|
||||
count() AS total_entries,
|
||||
uniq(src_ip) AS unique_ips,
|
||||
countIf(tcp_ttl = 0) AS no_tcp_data,
|
||||
countIf(tcp_ttl > 0) AS with_tcp_data,
|
||||
countIf(tcp_ttl BETWEEN 52 AND 65) AS linux_fingerprint,
|
||||
countIf(tcp_ttl BETWEEN 110 AND 135) AS windows_fingerprint
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
count() AS total_entries,
|
||||
uniq(src_ip) AS unique_ips,
|
||||
countIf(tcp_ttl_raw = 0) AS no_tcp_data,
|
||||
countIf(tcp_ttl_raw > 0) AS with_tcp_data,
|
||||
countIf(tcp_ttl_raw > 0 AND tcp_ttl_raw <= 64) AS linux_mac_fp,
|
||||
countIf(tcp_ttl_raw > 64 AND tcp_ttl_raw <= 128) AS windows_fp,
|
||||
countIf(tcp_ttl_raw > 128) AS cisco_bsd_fp,
|
||||
countIf(tcp_win_raw = 5808 AND tcp_mss_raw = 1452 AND tcp_scale_raw = 4) AS bot_scanner_fp
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR
|
||||
"""
|
||||
result = db.query(sql)
|
||||
row = result.result_rows[0]
|
||||
total_entries = int(row[0])
|
||||
unique_ips = int(row[1])
|
||||
no_tcp_data = int(row[2])
|
||||
with_tcp_data = int(row[3])
|
||||
linux_fp = int(row[4])
|
||||
windows_fp = int(row[5])
|
||||
|
||||
# Distribution TTL uniquement pour les entrées avec données TCP valides
|
||||
# Distribution TTL (top 15)
|
||||
ttl_sql = """
|
||||
SELECT
|
||||
tcp_ttl,
|
||||
count() AS cnt,
|
||||
uniq(src_ip) AS ips
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
WHERE tcp_ttl > 0
|
||||
GROUP BY tcp_ttl
|
||||
ORDER BY cnt DESC
|
||||
LIMIT 15
|
||||
SELECT tcp_ttl_raw AS ttl, count() AS cnt, uniq(src_ip) AS ips
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_ttl_raw > 0
|
||||
GROUP BY ttl ORDER BY cnt DESC LIMIT 15
|
||||
"""
|
||||
ttl_res = db.query(ttl_sql)
|
||||
ttl_distribution = [
|
||||
{"ttl": int(r[0]), "count": int(r[1]), "ips": int(r[2])}
|
||||
for r in ttl_res.result_rows
|
||||
]
|
||||
|
||||
# Distribution window_size pour entrées avec données TCP
|
||||
# Distribution MSS — nouveau signal clé (top 12)
|
||||
mss_sql = """
|
||||
SELECT tcp_mss_raw AS mss, count() AS cnt, uniq(src_ip) AS ips
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_mss_raw > 0
|
||||
GROUP BY mss ORDER BY cnt DESC LIMIT 12
|
||||
"""
|
||||
mss_res = db.query(mss_sql)
|
||||
|
||||
# Distribution fenêtre (top 10)
|
||||
win_sql = """
|
||||
SELECT
|
||||
tcp_window_size,
|
||||
count() AS cnt
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
WHERE tcp_ttl > 0
|
||||
GROUP BY tcp_window_size
|
||||
ORDER BY cnt DESC
|
||||
LIMIT 10
|
||||
SELECT tcp_win_raw AS win, count() AS cnt
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_ttl_raw > 0
|
||||
GROUP BY win ORDER BY cnt DESC LIMIT 10
|
||||
"""
|
||||
win_res = db.query(win_sql)
|
||||
window_size_distribution = [
|
||||
{"window_size": int(r[0]), "count": int(r[1])}
|
||||
for r in win_res.result_rows
|
||||
]
|
||||
|
||||
return {
|
||||
"total_entries": total_entries,
|
||||
"unique_ips": unique_ips,
|
||||
"no_tcp_data": no_tcp_data,
|
||||
"with_tcp_data": with_tcp_data,
|
||||
"linux_fingerprint": linux_fp,
|
||||
"windows_fingerprint": windows_fp,
|
||||
"ttl_distribution": ttl_distribution,
|
||||
"window_size_distribution": window_size_distribution,
|
||||
"total_entries": int(row[0]),
|
||||
"unique_ips": int(row[1]),
|
||||
"no_tcp_data": int(row[2]),
|
||||
"with_tcp_data": int(row[3]),
|
||||
"linux_mac_fingerprint": int(row[4]),
|
||||
"windows_fingerprint": int(row[5]),
|
||||
"cisco_bsd_fingerprint": int(row[6]),
|
||||
"bot_scanner_fingerprint": int(row[7]),
|
||||
"ttl_distribution": [
|
||||
{"ttl": int(r[0]), "count": int(r[1]), "ips": int(r[2])}
|
||||
for r in ttl_res.result_rows
|
||||
],
|
||||
"mss_distribution": [
|
||||
{"mss": int(r[0]), "count": int(r[1]), "ips": int(r[2])}
|
||||
for r in mss_res.result_rows
|
||||
],
|
||||
"window_size_distribution": [
|
||||
{"window_size": int(r[0]), "count": int(r[1])}
|
||||
for r in win_res.result_rows
|
||||
],
|
||||
}
|
||||
except Exception as e:
|
||||
raise HTTPException(status_code=500, detail=str(e))
|
||||
@ -134,54 +99,75 @@ async def get_tcp_spoofing_overview():
|
||||
async def get_tcp_spoofing_list(
|
||||
limit: int = Query(100, ge=1, le=1000),
|
||||
offset: int = Query(0, ge=0),
|
||||
spoof_only: bool = Query(False, description="Ne retourner que les vrais spoofs (TTL corrélable + OS mismatch)"),
|
||||
spoof_only: bool = Query(False, description="Retourner uniquement les spoofs/bots confirmés"),
|
||||
):
|
||||
"""Liste des entrées avec données TCP valides (tcp_ttl > 0).
|
||||
Entrées sans données TCP (TTL=0) exclues : pas de corrélation possible.
|
||||
Si spoof_only=True, retourne uniquement les entrées avec fingerprint OS identifiable (Linux/Mac TTL 52-65).
|
||||
"""Liste avec fingerprinting multi-signal (TTL + MSS + fenêtre + scale).
|
||||
Inclut les champs enrichis : mss, win_scale, initial_ttl, hop_count, confidence, network_path, is_bot_tool.
|
||||
"""
|
||||
try:
|
||||
# Filtre SQL : seules les entrées avec TTL valide, et si spoof_only les plages corrélables
|
||||
if spoof_only:
|
||||
# Seules les plages de TTL qui permettent une identification OS fiable
|
||||
ttl_filter = "tcp_ttl BETWEEN 52 AND 65 OR tcp_ttl BETWEEN 110 AND 135 OR tcp_ttl BETWEEN 240 AND 255"
|
||||
else:
|
||||
ttl_filter = "tcp_ttl > 0"
|
||||
|
||||
count_sql = f"SELECT count() FROM mabase_prod.view_tcp_spoofing_detected WHERE {ttl_filter}"
|
||||
count_sql = """
|
||||
SELECT count() FROM (
|
||||
SELECT src_ip, ja4
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_ttl_raw > 0
|
||||
GROUP BY src_ip, ja4
|
||||
)
|
||||
"""
|
||||
total = int(db.query(count_sql).result_rows[0][0])
|
||||
|
||||
sql = f"""
|
||||
sql = """
|
||||
SELECT
|
||||
replaceRegexpAll(toString(src_ip), '^::ffff:', '') AS src_ip,
|
||||
ja4, tcp_ttl, tcp_window_size, first_ua
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
WHERE {ttl_filter}
|
||||
ORDER BY tcp_ttl ASC
|
||||
ja4,
|
||||
any(tcp_ttl_raw) AS tcp_ttl,
|
||||
any(tcp_win_raw) AS tcp_window_size,
|
||||
any(tcp_scale_raw) AS tcp_win_scale,
|
||||
any(tcp_mss_raw) AS tcp_mss,
|
||||
any(first_ua) AS first_ua,
|
||||
sum(hits) AS hits
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_ttl_raw > 0
|
||||
GROUP BY src_ip, ja4
|
||||
ORDER BY hits DESC
|
||||
LIMIT %(limit)s OFFSET %(offset)s
|
||||
"""
|
||||
result = db.query(sql, {"limit": limit, "offset": offset})
|
||||
items = []
|
||||
for row in result.result_rows:
|
||||
ip = str(row[0])
|
||||
ja4 = str(row[1] or "")
|
||||
ttl = int(row[2])
|
||||
window_size = int(row[3])
|
||||
ua = str(row[4] or "")
|
||||
sus_os = _suspected_os(ttl)
|
||||
dec_os = _declared_os(ua)
|
||||
spoof_flag = _is_spoof(sus_os, dec_os)
|
||||
if spoof_only and not spoof_flag:
|
||||
ip = str(row[0])
|
||||
ja4 = str(row[1] or "")
|
||||
ttl = int(row[2] or 0)
|
||||
win = int(row[3] or 0)
|
||||
scale = int(row[4] or 0)
|
||||
mss = int(row[5] or 0)
|
||||
ua = str(row[6] or "")
|
||||
hits = int(row[7] or 0)
|
||||
|
||||
fp = fingerprint_os(ttl, win, scale, mss)
|
||||
dec_os = declared_os_from_ua(ua)
|
||||
spoof_res = detect_spoof(fp, dec_os)
|
||||
|
||||
if spoof_only and not spoof_res.is_spoof:
|
||||
continue
|
||||
|
||||
items.append({
|
||||
"ip": ip,
|
||||
"ja4": ja4,
|
||||
"tcp_ttl": ttl,
|
||||
"tcp_window_size": window_size,
|
||||
"tcp_window_size": win,
|
||||
"tcp_win_scale": scale,
|
||||
"tcp_mss": mss,
|
||||
"hits": hits,
|
||||
"first_ua": ua,
|
||||
"suspected_os": sus_os,
|
||||
"suspected_os": fp.os_name,
|
||||
"initial_ttl": fp.initial_ttl,
|
||||
"hop_count": fp.hop_count,
|
||||
"confidence": fp.confidence,
|
||||
"network_path": fp.network_path,
|
||||
"is_bot_tool": fp.is_bot_tool,
|
||||
"declared_os": dec_os,
|
||||
"spoof_flag": spoof_flag,
|
||||
"spoof_flag": spoof_res.is_spoof,
|
||||
"spoof_reason": spoof_res.reason,
|
||||
})
|
||||
return {"items": items, "total": total}
|
||||
except Exception as e:
|
||||
@ -190,29 +176,44 @@ async def get_tcp_spoofing_list(
|
||||
|
||||
@router.get("/matrix")
|
||||
async def get_tcp_spoofing_matrix():
|
||||
"""Matrice suspected_os × declared_os — uniquement entrées avec TTL valide."""
|
||||
"""Matrice OS suspecté × OS déclaré avec fingerprinting multi-signal."""
|
||||
try:
|
||||
sql = """
|
||||
SELECT tcp_ttl, first_ua
|
||||
FROM mabase_prod.view_tcp_spoofing_detected
|
||||
WHERE tcp_ttl > 0
|
||||
SELECT
|
||||
any(tcp_ttl_raw) AS ttl,
|
||||
any(tcp_win_raw) AS win,
|
||||
any(tcp_scale_raw) AS scale,
|
||||
any(tcp_mss_raw) AS mss,
|
||||
any(first_ua) AS ua,
|
||||
count() AS cnt
|
||||
FROM mabase_prod.agg_host_ip_ja4_1h
|
||||
WHERE window_start >= now() - INTERVAL 24 HOUR AND tcp_ttl_raw > 0
|
||||
GROUP BY src_ip, ja4
|
||||
"""
|
||||
result = db.query(sql)
|
||||
counts: dict = {}
|
||||
for row in result.result_rows:
|
||||
ttl = int(row[0])
|
||||
ua = str(row[1] or "")
|
||||
sus_os = _suspected_os(ttl)
|
||||
dec_os = _declared_os(ua)
|
||||
key = (sus_os, dec_os)
|
||||
counts[key] = counts.get(key, 0) + 1
|
||||
ttl = int(row[0] or 0)
|
||||
win = int(row[1] or 0)
|
||||
scale = int(row[2] or 0)
|
||||
mss = int(row[3] or 0)
|
||||
ua = str(row[4] or "")
|
||||
cnt = int(row[5] or 1)
|
||||
|
||||
fp = fingerprint_os(ttl, win, scale, mss)
|
||||
dec_os = declared_os_from_ua(ua)
|
||||
spoof_res = detect_spoof(fp, dec_os)
|
||||
|
||||
key = (fp.os_name, dec_os, spoof_res.is_spoof, fp.is_bot_tool)
|
||||
counts[key] = counts.get(key, 0) + cnt
|
||||
|
||||
matrix = [
|
||||
{
|
||||
"suspected_os": k[0],
|
||||
"declared_os": k[1],
|
||||
"count": v,
|
||||
"is_spoof": _is_spoof(k[0], k[1]),
|
||||
"is_spoof": k[2],
|
||||
"is_bot_tool": k[3],
|
||||
}
|
||||
for k, v in counts.items()
|
||||
]
|
||||
|
||||
328
backend/services/clustering_engine.py
Normal file
328
backend/services/clustering_engine.py
Normal file
@ -0,0 +1,328 @@
|
||||
"""
|
||||
Moteur de clustering K-means++ multi-métriques (pur Python).
|
||||
|
||||
Ref: Arthur & Vassilvitskii (2007) — k-means++: The Advantages of Careful Seeding
|
||||
Hotelling (1933) — PCA par puissance itérative (deflation)
|
||||
|
||||
Features (21 dimensions, normalisées [0,1]) :
|
||||
0 ttl_n : TTL initial normalisé (hops-count estimé)
|
||||
1 mss_n : MSS normalisé → type réseau (Ethernet/PPPoE/VPN)
|
||||
2 scale_n : facteur de mise à l'échelle TCP
|
||||
3 win_n : fenêtre TCP normalisée
|
||||
4 score_n : score anomalie ML (abs)
|
||||
5 velocity_n : vélocité de requêtes (log1p)
|
||||
6 fuzzing_n : index de fuzzing (log1p)
|
||||
7 headless_n : ratio sessions headless
|
||||
8 post_n : ratio POST/total
|
||||
9 ip_id_zero_n : ratio IP-ID=0 (Linux/spoofé)
|
||||
10 entropy_n : entropie temporelle
|
||||
11 browser_n : score navigateur moderne (normalisé max 50)
|
||||
12 alpn_n : mismatch ALPN/protocole
|
||||
13 alpn_absent_n : ratio ALPN absent
|
||||
14 h2_n : efficacité H2 multiplexing (log1p)
|
||||
15 hdr_conf_n : confiance ordre headers
|
||||
16 ua_ch_n : mismatch User-Agent-Client-Hints
|
||||
17 asset_n : ratio assets statiques
|
||||
18 direct_n : ratio accès directs
|
||||
19 ja4_div_n : diversité JA4 (log1p)
|
||||
20 ua_rot_n : UA rotatif (booléen)
|
||||
"""
|
||||
from __future__ import annotations
|
||||
import math
|
||||
import random
|
||||
from dataclasses import dataclass, field
|
||||
|
||||
# ─── Définition des features ──────────────────────────────────────────────────
|
||||
|
||||
# (clé SQL, nom lisible, fonction de normalisation)
|
||||
FEATURES = [
|
||||
# TCP stack
|
||||
("ttl", "TTL Initial", lambda v: min(1.0, (v or 0) / 255.0)),
|
||||
("mss", "MSS Réseau", lambda v: min(1.0, (v or 0) / 1460.0)),
|
||||
("scale", "Scale TCP", lambda v: min(1.0, (v or 0) / 14.0)),
|
||||
("win", "Fenêtre TCP", lambda v: min(1.0, (v or 0) / 65535.0)),
|
||||
# Anomalie ML
|
||||
("avg_score", "Score Anomalie", lambda v: min(1.0, float(v or 0))),
|
||||
("avg_velocity", "Vélocité (rps)", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(100))),
|
||||
("avg_fuzzing", "Fuzzing", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(300))),
|
||||
("pct_headless", "Headless", lambda v: min(1.0, float(v or 0))),
|
||||
("avg_post", "Ratio POST", lambda v: min(1.0, float(v or 0))),
|
||||
# IP-ID
|
||||
("ip_id_zero", "IP-ID Zéro", lambda v: min(1.0, float(v or 0))),
|
||||
# Temporel
|
||||
("entropy", "Entropie Temporelle", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(10))),
|
||||
# Navigateur
|
||||
("browser_score","Score Navigateur", lambda v: min(1.0, float(v or 0) / 50.0)),
|
||||
# TLS / Protocole
|
||||
("alpn_mismatch","ALPN Mismatch", lambda v: min(1.0, float(v or 0))),
|
||||
("alpn_missing", "ALPN Absent", lambda v: min(1.0, float(v or 0))),
|
||||
("h2_eff", "H2 Multiplexing", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(20))),
|
||||
("hdr_conf", "Ordre Headers", lambda v: min(1.0, float(v or 0))),
|
||||
("ua_ch_mismatch","UA-CH Mismatch", lambda v: min(1.0, float(v or 0))),
|
||||
# Comportement HTTP
|
||||
("asset_ratio", "Ratio Assets", lambda v: min(1.0, float(v or 0))),
|
||||
("direct_ratio", "Accès Direct", lambda v: min(1.0, float(v or 0))),
|
||||
# Diversité JA4
|
||||
("ja4_count", "Diversité JA4", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(30))),
|
||||
# UA rotatif
|
||||
("ua_rotating", "UA Rotatif", lambda v: 1.0 if float(v or 0) > 0 else 0.0),
|
||||
]
|
||||
|
||||
FEATURE_KEYS = [f[0] for f in FEATURES]
|
||||
FEATURE_NAMES = [f[1] for f in FEATURES]
|
||||
FEATURE_NORMS = [f[2] for f in FEATURES]
|
||||
N_FEATURES = len(FEATURES)
|
||||
|
||||
|
||||
# ─── Utilitaires vectoriels (pur Python) ──────────────────────────────────────
|
||||
|
||||
def _dist2(a: list[float], b: list[float]) -> float:
|
||||
return sum((x - y) ** 2 for x, y in zip(a, b))
|
||||
|
||||
def _mean_vec(vecs: list[list[float]]) -> list[float]:
|
||||
n = len(vecs)
|
||||
if n == 0:
|
||||
return [0.0] * N_FEATURES
|
||||
return [sum(v[i] for v in vecs) / n for i in range(N_FEATURES)]
|
||||
|
||||
|
||||
# ─── Construction du vecteur de features ─────────────────────────────────────
|
||||
|
||||
def build_feature_vector(row: dict) -> list[float]:
|
||||
"""Normalise un dict de colonnes SQL → vecteur [0,1]^N_FEATURES."""
|
||||
return [fn(row.get(key)) for key, fn in zip(FEATURE_KEYS, FEATURE_NORMS)]
|
||||
|
||||
|
||||
# ─── K-means++ ───────────────────────────────────────────────────────────────
|
||||
|
||||
@dataclass
|
||||
class KMeansResult:
|
||||
centroids: list[list[float]]
|
||||
labels: list[int]
|
||||
inertia: float
|
||||
n_iter: int
|
||||
|
||||
|
||||
def kmeans_pp(
|
||||
points: list[list[float]],
|
||||
k: int,
|
||||
max_iter: int = 60,
|
||||
seed: int = 42,
|
||||
n_init: int = 3,
|
||||
) -> KMeansResult:
|
||||
"""
|
||||
K-means avec initialisation k-means++ (Arthur & Vassilvitskii, 2007).
|
||||
Lance `n_init` fois et retourne le meilleur résultat (inertie minimale).
|
||||
"""
|
||||
rng = random.Random(seed)
|
||||
best: KMeansResult | None = None
|
||||
|
||||
for attempt in range(n_init):
|
||||
# ── Initialisation k-means++ ────────────────────────────────────
|
||||
first_idx = rng.randrange(len(points))
|
||||
centroids = [points[first_idx][:]]
|
||||
|
||||
for _ in range(k - 1):
|
||||
d2 = [min(_dist2(p, c) for c in centroids) for p in points]
|
||||
total = sum(d2)
|
||||
if total == 0:
|
||||
break
|
||||
r = rng.random() * total
|
||||
cumul = 0.0
|
||||
for i, d in enumerate(d2):
|
||||
cumul += d
|
||||
if cumul >= r:
|
||||
centroids.append(points[i][:])
|
||||
break
|
||||
else:
|
||||
centroids.append(points[rng.randrange(len(points))][:])
|
||||
|
||||
# ── Itérations EM ───────────────────────────────────────────────
|
||||
labels: list[int] = [0] * len(points)
|
||||
for iteration in range(max_iter):
|
||||
# E-step : affectation
|
||||
new_labels = [
|
||||
min(range(len(centroids)), key=lambda c: _dist2(p, centroids[c]))
|
||||
for p in points
|
||||
]
|
||||
if new_labels == labels and iteration > 0:
|
||||
break
|
||||
labels = new_labels
|
||||
|
||||
# M-step : mise à jour
|
||||
clusters: list[list[list[float]]] = [[] for _ in range(k)]
|
||||
for i, l in enumerate(labels):
|
||||
clusters[l].append(points[i])
|
||||
for j in range(k):
|
||||
if clusters[j]:
|
||||
centroids[j] = _mean_vec(clusters[j])
|
||||
|
||||
inertia = sum(_dist2(points[i], centroids[labels[i]]) for i in range(len(points)))
|
||||
result = KMeansResult(
|
||||
centroids=centroids,
|
||||
labels=labels,
|
||||
inertia=inertia,
|
||||
n_iter=iteration + 1,
|
||||
)
|
||||
if best is None or inertia < best.inertia:
|
||||
best = result
|
||||
|
||||
return best # type: ignore
|
||||
|
||||
|
||||
# ─── PCA 2D par puissance itérative ──────────────────────────────────────────
|
||||
|
||||
def pca_2d(points: list[list[float]]) -> list[tuple[float, float]]:
|
||||
"""
|
||||
Projection PCA 2D par puissance itérative avec déflation (Hotelling).
|
||||
Retourne les coordonnées (pc1, pc2) normalisées dans [0,1].
|
||||
"""
|
||||
n = len(points)
|
||||
if n == 0:
|
||||
return []
|
||||
|
||||
# Centrage
|
||||
mean = _mean_vec(points)
|
||||
X = [[p[i] - mean[i] for i in range(N_FEATURES)] for p in points]
|
||||
|
||||
def power_iter(X_centered: list[list[float]], n_iter: int = 30) -> list[float]:
|
||||
"""Trouve le premier vecteur propre de X^T X par puissance itérative."""
|
||||
v = [1.0 / math.sqrt(N_FEATURES)] * N_FEATURES
|
||||
for _ in range(n_iter):
|
||||
# Xv = X @ v
|
||||
Xv = [sum(row[j] * v[j] for j in range(N_FEATURES)) for row in X_centered]
|
||||
# Xtxv = X^T @ Xv
|
||||
xtxv = [sum(X_centered[i][j] * Xv[i] for i in range(len(X_centered))) for j in range(N_FEATURES)]
|
||||
norm = math.sqrt(sum(x ** 2 for x in xtxv)) or 1e-10
|
||||
v = [x / norm for x in xtxv]
|
||||
return v
|
||||
|
||||
# PC1
|
||||
v1 = power_iter(X)
|
||||
proj1 = [sum(row[j] * v1[j] for j in range(N_FEATURES)) for row in X]
|
||||
|
||||
# Déflation : retire la composante PC1 de X
|
||||
X2 = [
|
||||
[X[i][j] - proj1[i] * v1[j] for j in range(N_FEATURES)]
|
||||
for i in range(n)
|
||||
]
|
||||
|
||||
# PC2
|
||||
v2 = power_iter(X2)
|
||||
proj2 = [sum(row[j] * v2[j] for j in range(N_FEATURES)) for row in X2]
|
||||
|
||||
# Normalisation [0,1]
|
||||
def _norm01(vals: list[float]) -> list[float]:
|
||||
lo, hi = min(vals), max(vals)
|
||||
rng = hi - lo or 1e-10
|
||||
return [(v - lo) / rng for v in vals]
|
||||
|
||||
p1 = _norm01(proj1)
|
||||
p2 = _norm01(proj2)
|
||||
|
||||
return list(zip(p1, p2))
|
||||
|
||||
|
||||
# ─── Nommage automatique des clusters ────────────────────────────────────────
|
||||
|
||||
def name_cluster(centroid: list[float], raw_stats: dict | None = None) -> str:
|
||||
"""
|
||||
Génère un nom lisible à partir du centroïde normalisé et de statistiques brutes.
|
||||
Priorité : signaux les plus discriminants en premier.
|
||||
"""
|
||||
score = centroid[4] # anomalie ML
|
||||
vel = centroid[5] # vélocité
|
||||
fuzz = centroid[6] # fuzzing (log1p normalisé, >0.35 ≈ fuzzing_index > 100)
|
||||
hless = centroid[7] # headless
|
||||
post = centroid[8] # POST ratio
|
||||
alpn = centroid[12] # ALPN mismatch
|
||||
h2 = centroid[14] # H2 eff
|
||||
ua_ch = centroid[16] # UA-CH mismatch
|
||||
ja4d = centroid[19] # JA4 diversité
|
||||
ua_rot = centroid[20] # UA rotatif
|
||||
|
||||
raw_mss = (raw_stats or {}).get("mean_mss", 0)
|
||||
raw_ttl = (raw_stats or {}).get("mean_ttl", 0) or (centroid[0] * 255)
|
||||
raw_scale = (raw_stats or {}).get("mean_scale", 0)
|
||||
|
||||
# ── Signaux forts (déterministes) ────────────────────────────────────
|
||||
|
||||
# Pattern Masscan : mss≈1452, scale≈4, TTL 48-57
|
||||
if raw_mss and 1440 <= raw_mss <= 1460 and raw_scale and 3 <= raw_scale <= 5 and raw_ttl < 60:
|
||||
return "🤖 Masscan / Scanner IP"
|
||||
|
||||
# Fuzzer agressif (fuzzing_index normalisé > 0.35 ≈ valeur brute > 100)
|
||||
if fuzz > 0.35:
|
||||
return "🤖 Bot Fuzzer / Scanner"
|
||||
|
||||
# UA rotatif + UA-CH mismatch : bot sophistiqué simulant un navigateur
|
||||
if ua_rot > 0.5 and ua_ch > 0.7:
|
||||
return "🤖 Bot UA Rotatif + CH Mismatch"
|
||||
|
||||
# UA-CH mismatch fort seul (navigateur simulé sans headers CH)
|
||||
if ua_ch > 0.8:
|
||||
return "⚠️ Bot UA-CH Incohérent"
|
||||
|
||||
# ── Score ML modéré + signal comportemental ──────────────────────────
|
||||
|
||||
if score > 0.20:
|
||||
if hless > 0.3:
|
||||
return "⚠️ Navigateur Headless Suspect"
|
||||
if vel > 0.25:
|
||||
return "⚠️ Bot Haute Vélocité"
|
||||
if post > 0.4:
|
||||
return "⚠️ Bot POST Automatisé"
|
||||
if alpn > 0.5 or h2 > 0.5:
|
||||
return "⚠️ TLS/H2 Anormal"
|
||||
if ua_ch > 0.4:
|
||||
return "⚠️ Anomalie UA-CH"
|
||||
return "⚠️ Anomalie ML Modérée"
|
||||
|
||||
# ── Signaux faibles ───────────────────────────────────────────────────
|
||||
|
||||
if ua_ch > 0.4:
|
||||
return "🔎 UA-CH Incohérent"
|
||||
|
||||
if ja4d > 0.5:
|
||||
return "🔄 Client Multi-Fingerprint"
|
||||
|
||||
# ── Classification réseau / OS ────────────────────────────────────────
|
||||
|
||||
# MSS bas → VPN ou tunnel
|
||||
if raw_mss and raw_mss < 1360:
|
||||
return "🌐 VPN / Tunnel"
|
||||
|
||||
if raw_ttl < 70:
|
||||
return "🐧 Linux / Mobile"
|
||||
if raw_ttl > 110:
|
||||
return "🪟 Windows"
|
||||
|
||||
return "✅ Trafic Légitime"
|
||||
|
||||
|
||||
def risk_score_from_centroid(centroid: list[float]) -> float:
|
||||
"""Score de risque [0,1] pondéré. Calibré pour les valeurs observées (score ML ~0.3)."""
|
||||
# Normalisation de score ML : x / 0.5 pour étendre la plage utile (0-0.5 → 0-1)
|
||||
score_n = min(1.0, centroid[4] / 0.5)
|
||||
fuzz_n = centroid[6]
|
||||
ua_ch_n = centroid[16]
|
||||
ua_rot_n = centroid[20]
|
||||
vel_n = centroid[5]
|
||||
hless_n = centroid[7]
|
||||
ip_id_n = centroid[9]
|
||||
alpn_n = centroid[12]
|
||||
ja4d_n = centroid[19]
|
||||
post_n = centroid[8]
|
||||
|
||||
return min(1.0,
|
||||
0.25 * score_n +
|
||||
0.20 * ua_ch_n +
|
||||
0.15 * fuzz_n +
|
||||
0.12 * ua_rot_n +
|
||||
0.10 * hless_n +
|
||||
0.07 * vel_n +
|
||||
0.04 * ip_id_n +
|
||||
0.04 * alpn_n +
|
||||
0.03 * ja4d_n +
|
||||
0.03 * post_n
|
||||
)
|
||||
436
backend/services/tcp_fingerprint.py
Normal file
436
backend/services/tcp_fingerprint.py
Normal file
@ -0,0 +1,436 @@
|
||||
"""
|
||||
Service de fingerprinting OS par signature TCP — approche multi-signal inspirée de p0f.
|
||||
|
||||
Signaux utilisés (par ordre de poids) :
|
||||
1. TTL initial estimé (→ famille OS : Linux/Mac=64, Windows=128, Cisco/BSD=255)
|
||||
2. MSS (→ type de réseau : Ethernet=1460, PPPoE=1452, VPN=1380-1420)
|
||||
3. Taille de fenêtre (→ signature OS précise)
|
||||
4. Facteur d'échelle (→ affine la version du kernel/stack TCP)
|
||||
|
||||
Références :
|
||||
- p0f v3 (Michal Zalewski) — passive OS fingerprinting
|
||||
- Nmap OS detection (Gordon Lyon)
|
||||
- "OS Fingerprinting Revisited" (Beverly, 2004)
|
||||
- "Passive OS fingerprinting" (Orebaugh, Ramirez)
|
||||
- Recherche sur Masscan/ZMap : signatures SYN craftées connues
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
from dataclasses import dataclass
|
||||
from typing import Optional
|
||||
|
||||
|
||||
# ─── Constantes ───────────────────────────────────────────────────────────────
|
||||
|
||||
_INITIAL_TTLS = (64, 128, 255)
|
||||
|
||||
# MSS → type de chemin réseau (MTU - 40 octets d'en-têtes IP+TCP)
|
||||
_MSS_PATH: list[tuple[range, str]] = [
|
||||
(range(1461, 9001), "Ethernet/Jumbo"), # jumbo frames (CDN/datacenter)
|
||||
(range(1460, 1461), "Ethernet directe"), # MTU 1500 standard
|
||||
(range(1453, 1460), "Ethernet directe"), # légèrement réduit (padding)
|
||||
(range(1452, 1453), "PPPoE/DSL"), # MTU 1492
|
||||
(range(1436, 1452), "PPPoE/DSL ajusté"), # variations DSL
|
||||
(range(1420, 1436), "VPN léger"), # WireGuard / IPsec transport
|
||||
(range(1380, 1420), "VPN/Tunnel"), # OpenVPN / L2TP
|
||||
(range(1300, 1380), "VPN double ou mobile"),
|
||||
(range(0, 1300), "Lien bas débit / GPRS"),
|
||||
]
|
||||
|
||||
|
||||
# ─── Base de signatures OS ─────────────────────────────────────────────────────
|
||||
#
|
||||
# Format : chaque entrée est un dict avec :
|
||||
# ttl : int — TTL initial attendu (64 | 128 | 255)
|
||||
# win : set[int]|None — tailles de fenêtre attendues (None = ignorer)
|
||||
# scale : set[int]|None — facteurs d'échelle attendus (None = ignorer)
|
||||
# mss : set[int]|None — valeurs MSS attendues (None = ignorer)
|
||||
# name : str — libellé affiché
|
||||
# conf : float — poids de confiance de base (0–1)
|
||||
# bot : bool — outil de scan/bot connu
|
||||
|
||||
_SIGNATURES: list[dict] = [
|
||||
# ══════════════════════════════════════════════════════
|
||||
# OUTILS DE SCAN ET BOTS CONNUS (priorité maximale)
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# Masscan / scanner personnalisé avec stack Linux modifiée (PPPoE MSS=1452)
|
||||
# Pattern très présent dans les données : ~111k requêtes, UA spoofé macOS/Windows
|
||||
{
|
||||
"ttl": 64, "win": {5808}, "scale": {4}, "mss": {1452},
|
||||
"name": "Bot-Scanner/Masscan", "conf": 0.97, "bot": True,
|
||||
},
|
||||
# Masscan TTL=255 (mode direct, pas de hop)
|
||||
{
|
||||
"ttl": 255, "win": {1024}, "scale": {0}, "mss": None,
|
||||
"name": "Bot-ZMap/Masscan", "conf": 0.96, "bot": True,
|
||||
},
|
||||
# Mirai variant (petite fenêtre, pas de scale, TTL Linux)
|
||||
{
|
||||
"ttl": 64, "win": {1024, 2048}, "scale": {0}, "mss": {1460},
|
||||
"name": "Bot-Mirai", "conf": 0.92, "bot": True,
|
||||
},
|
||||
# Mirai variant (petite fenêtre Windows)
|
||||
{
|
||||
"ttl": 128, "win": {1024, 2048}, "scale": {0}, "mss": {1460},
|
||||
"name": "Bot-Mirai/Win", "conf": 0.92, "bot": True,
|
||||
},
|
||||
# Scapy / forge manuelle (fenêtre 8192 exactement + TTL 64 + pas de scale)
|
||||
{
|
||||
"ttl": 64, "win": {8192}, "scale": {0}, "mss": {1460},
|
||||
"name": "Bot-Scapy/Forge", "conf": 0.85, "bot": True,
|
||||
},
|
||||
# Nmap SYN scan (window=1024, MSS=1460, TTL=64 ou 128)
|
||||
{
|
||||
"ttl": 64, "win": {1}, "scale": None, "mss": None,
|
||||
"name": "Bot-ZMap", "conf": 0.95, "bot": True,
|
||||
},
|
||||
|
||||
# ══════════════════════════════════════════════════════
|
||||
# WINDOWS
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# Windows 10 / 11 — signature standard (LAN direct)
|
||||
{
|
||||
"ttl": 128, "win": {64240}, "scale": {8}, "mss": {1460},
|
||||
"name": "Windows 10/11", "conf": 0.93, "bot": False,
|
||||
},
|
||||
# Windows 10/11 — derrière VPN/proxy (MSS réduit)
|
||||
{
|
||||
"ttl": 128, "win": {64240}, "scale": {8}, "mss": {1380, 1400, 1412, 1420, 1440},
|
||||
"name": "Windows 10/11 (VPN)", "conf": 0.90, "bot": False,
|
||||
},
|
||||
# Windows Server 2019/2022 — grande fenêtre
|
||||
{
|
||||
"ttl": 128, "win": {65535, 131072}, "scale": {8, 9}, "mss": {1460},
|
||||
"name": "Windows Server", "conf": 0.88, "bot": False,
|
||||
},
|
||||
# Windows 7/8.1
|
||||
{
|
||||
"ttl": 128, "win": {8192, 65535}, "scale": {4, 8}, "mss": {1460},
|
||||
"name": "Windows 7/8", "conf": 0.83, "bot": False,
|
||||
},
|
||||
# Windows générique (TTL=128, scale=8, tout MSS)
|
||||
{
|
||||
"ttl": 128, "win": None, "scale": {8}, "mss": None,
|
||||
"name": "Windows", "conf": 0.70, "bot": False,
|
||||
},
|
||||
|
||||
# ══════════════════════════════════════════════════════
|
||||
# ANDROID (stack BBRv2 / CUBIC moderne)
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# Android 10+ — scale=9 ou 10, grande fenêtre (BBRv2)
|
||||
{
|
||||
"ttl": 64, "win": {65535, 131072, 42340, 35844}, "scale": {9, 10}, "mss": {1460},
|
||||
"name": "Android 10+", "conf": 0.82, "bot": False,
|
||||
},
|
||||
# Android via proxy TTL=128 (app Facebook, TikTok etc. passant par infra)
|
||||
{
|
||||
"ttl": 128, "win": {62727, 65535}, "scale": {7}, "mss": {1460},
|
||||
"name": "Android/App (proxy)", "conf": 0.75, "bot": False,
|
||||
},
|
||||
# Android derrière VPN (MSS réduit)
|
||||
{
|
||||
"ttl": 64, "win": {65535, 59640, 63940}, "scale": {8, 9, 10}, "mss": {1380, 1390, 1400, 1418, 1420},
|
||||
"name": "Android (VPN/mobile)", "conf": 0.78, "bot": False,
|
||||
},
|
||||
|
||||
# ══════════════════════════════════════════════════════
|
||||
# iOS / macOS
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# iOS 14+ / macOS Monterey+ — scale=6, win=65535 (signature XNU)
|
||||
{
|
||||
"ttl": 64, "win": {65535, 32768}, "scale": {6}, "mss": {1460},
|
||||
"name": "iOS/macOS", "conf": 0.87, "bot": False,
|
||||
},
|
||||
# macOS Sonoma+ / iOS 17+ (scale=9, fenêtre plus grande)
|
||||
{
|
||||
"ttl": 64, "win": {65535, 32768}, "scale": {9}, "mss": {1460},
|
||||
"name": "macOS Sonoma+/iOS 17+", "conf": 0.83, "bot": False,
|
||||
},
|
||||
# macOS derrière VPN (MSS réduit)
|
||||
{
|
||||
"ttl": 64, "win": {65535}, "scale": {6, 9}, "mss": {1380, 1400, 1412, 1436},
|
||||
"name": "iOS/macOS (VPN)", "conf": 0.80, "bot": False,
|
||||
},
|
||||
|
||||
# ══════════════════════════════════════════════════════
|
||||
# LINUX (desktop/serveur)
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# Linux 5.x+ — scale=7, win=64240 ou 65535 (kernel ≥ 4.19)
|
||||
{
|
||||
"ttl": 64, "win": {64240, 65320}, "scale": {7}, "mss": {1460},
|
||||
"name": "Linux 5.x+", "conf": 0.86, "bot": False,
|
||||
},
|
||||
# Linux 4.x / ChromeOS
|
||||
{
|
||||
"ttl": 64, "win": {29200, 65535, 43690, 32120}, "scale": {7}, "mss": {1460},
|
||||
"name": "Linux 4.x/ChromeOS", "conf": 0.83, "bot": False,
|
||||
},
|
||||
# Linux derrière VPN (MSS réduit)
|
||||
{
|
||||
"ttl": 64, "win": {64240, 65535, 42600}, "scale": {7}, "mss": {1380, 1400, 1420, 1436},
|
||||
"name": "Linux (VPN)", "conf": 0.80, "bot": False,
|
||||
},
|
||||
# Linux 2.6.x (ancien — win=5840/14600)
|
||||
{
|
||||
"ttl": 64, "win": {5840, 14600, 16384}, "scale": {4, 5}, "mss": {1460},
|
||||
"name": "Linux 2.6", "conf": 0.78, "bot": False,
|
||||
},
|
||||
|
||||
# ══════════════════════════════════════════════════════
|
||||
# BSD / ÉQUIPEMENTS RÉSEAU / CDN
|
||||
# ══════════════════════════════════════════════════════
|
||||
|
||||
# FreeBSD / OpenBSD (initial TTL=64)
|
||||
{
|
||||
"ttl": 64, "win": {65535}, "scale": {6}, "mss": {512, 1460},
|
||||
"name": "FreeBSD/OpenBSD", "conf": 0.74, "bot": False,
|
||||
},
|
||||
# Cisco IOS / équipements réseau (initial TTL=255, fenêtre petite)
|
||||
{
|
||||
"ttl": 255, "win": {4096, 4128, 8760}, "scale": {0, 1, 2}, "mss": {512, 1460},
|
||||
"name": "Cisco/Réseau", "conf": 0.87, "bot": False,
|
||||
},
|
||||
# CDN / Applebot (TTL=255, jumbo MSS, fenêtre élevée)
|
||||
{
|
||||
"ttl": 255, "win": {26883, 65535, 59640}, "scale": {7, 8}, "mss": {8373, 8365, 1460},
|
||||
"name": "CDN/Applebot (jumbo)", "conf": 0.85, "bot": False,
|
||||
},
|
||||
# BSD/Unix générique (TTL=255)
|
||||
{
|
||||
"ttl": 255, "win": None, "scale": {6, 7, 8}, "mss": {1460},
|
||||
"name": "BSD/Unix", "conf": 0.68, "bot": False,
|
||||
},
|
||||
]
|
||||
|
||||
|
||||
# ─── Data classes ──────────────────────────────────────────────────────────────
|
||||
|
||||
@dataclass
|
||||
class OsFingerprint:
|
||||
os_name: str
|
||||
initial_ttl: int
|
||||
hop_count: int
|
||||
confidence: float
|
||||
is_bot_tool: bool
|
||||
network_path: str
|
||||
|
||||
|
||||
@dataclass
|
||||
class SpoofResult:
|
||||
is_spoof: bool
|
||||
is_bot_tool: bool
|
||||
reason: str
|
||||
|
||||
|
||||
# ─── Fonctions utilitaires ─────────────────────────────────────────────────────
|
||||
|
||||
def _estimate_initial_ttl(observed_ttl: int) -> tuple[int, int]:
|
||||
"""Retourne (initial_ttl, hop_count).
|
||||
Cherche le TTL standard le plus bas >= observed_ttl.
|
||||
Rejette les hop counts > 45 (réseau légitimement long = peu probable).
|
||||
"""
|
||||
if observed_ttl <= 0:
|
||||
return 0, -1
|
||||
for initial in _INITIAL_TTLS:
|
||||
if observed_ttl <= initial:
|
||||
hop = initial - observed_ttl
|
||||
if hop <= 45:
|
||||
return initial, hop
|
||||
return 255, 255 - observed_ttl # TTL > 255 impossible, fallback
|
||||
|
||||
|
||||
def _infer_network_path(mss: int) -> str:
|
||||
"""Retourne le type de chemin réseau probable à partir du MSS."""
|
||||
if mss <= 0:
|
||||
return "Inconnu"
|
||||
for rng, label in _MSS_PATH:
|
||||
if mss in rng:
|
||||
return label
|
||||
return "Inconnu"
|
||||
|
||||
|
||||
def _os_family(os_name: str) -> str:
|
||||
"""Réduit un nom OS détaillé à sa famille pour comparaison avec l'UA."""
|
||||
n = os_name.lower()
|
||||
if "windows" in n:
|
||||
return "Windows"
|
||||
if "android" in n:
|
||||
return "Android"
|
||||
if "ios" in n or "macos" in n or "iphone" in n or "ipad" in n:
|
||||
return "Apple"
|
||||
if "linux" in n or "chromeos" in n:
|
||||
return "Linux"
|
||||
if "bsd" in n or "cisco" in n or "cdn" in n or "réseau" in n:
|
||||
return "Network"
|
||||
if "bot" in n or "scanner" in n or "mirai" in n or "zmap" in n:
|
||||
return "Bot"
|
||||
return "Unknown"
|
||||
|
||||
|
||||
def _ua_os_family(declared_os: str) -> str:
|
||||
"""Réduit l'OS déclaré (UA) à sa famille."""
|
||||
mapping = {
|
||||
"Windows": "Windows",
|
||||
"Android": "Android",
|
||||
"iOS": "Apple",
|
||||
"macOS": "Apple",
|
||||
"Linux": "Linux",
|
||||
"ChromeOS": "Linux",
|
||||
"BSD": "Network",
|
||||
}
|
||||
return mapping.get(declared_os, "Unknown")
|
||||
|
||||
|
||||
# ─── Fonctions publiques ───────────────────────────────────────────────────────
|
||||
|
||||
def declared_os_from_ua(ua: str) -> str:
|
||||
"""Infère l'OS déclaré à partir du User-Agent."""
|
||||
ua = ua or ""
|
||||
ul = ua.lower()
|
||||
if not ul:
|
||||
return "Unknown"
|
||||
if "windows nt" in ul:
|
||||
return "Windows"
|
||||
if "android" in ul:
|
||||
return "Android"
|
||||
if "iphone" in ul or "ipad" in ul or "cpu iphone" in ul or "cpu ipad" in ul:
|
||||
return "iOS"
|
||||
if "mac os x" in ul or "macos" in ul:
|
||||
return "macOS"
|
||||
if "cros" in ul or "chromeos" in ul:
|
||||
return "ChromeOS"
|
||||
if "linux" in ul:
|
||||
return "Linux"
|
||||
if "freebsd" in ul or "openbsd" in ul or "netbsd" in ul:
|
||||
return "BSD"
|
||||
return "Unknown"
|
||||
|
||||
|
||||
def fingerprint_os(ttl: int, win: int, scale: int, mss: int) -> OsFingerprint:
|
||||
"""Fingerprint OS multi-signal avec scoring pondéré.
|
||||
|
||||
Poids des signaux :
|
||||
- TTL initial 40 % (discriminant principal : famille OS)
|
||||
- MSS 30 % (type de réseau ET OS)
|
||||
- Fenêtre TCP 20 % (version/distrib précise)
|
||||
- Scale 10 % (affine la version kernel)
|
||||
"""
|
||||
initial_ttl, hop_count = _estimate_initial_ttl(ttl)
|
||||
network_path = _infer_network_path(mss)
|
||||
|
||||
if initial_ttl == 0:
|
||||
return OsFingerprint(
|
||||
os_name="Unknown", initial_ttl=0, hop_count=-1,
|
||||
confidence=0.0, is_bot_tool=False, network_path=network_path,
|
||||
)
|
||||
|
||||
best_score: float = -1.0
|
||||
best_sig: Optional[dict] = None
|
||||
|
||||
for sig in _SIGNATURES:
|
||||
# Le TTL est un filtre strict — pas de correspondance, on passe
|
||||
if sig["ttl"] != initial_ttl:
|
||||
continue
|
||||
|
||||
score: float = 0.40 # Score de base pour correspondance TTL
|
||||
|
||||
# MSS (poids 0.30)
|
||||
if sig["mss"] is not None:
|
||||
score += 0.30 if mss in sig["mss"] else -0.12
|
||||
|
||||
# Fenêtre (poids 0.20)
|
||||
if sig["win"] is not None:
|
||||
score += 0.20 if win in sig["win"] else -0.08
|
||||
|
||||
# Scale (poids 0.10)
|
||||
if sig["scale"] is not None:
|
||||
score += 0.10 if scale in sig["scale"] else -0.04
|
||||
|
||||
# Pénalité si hop count anormalement élevé (>30 hops)
|
||||
if hop_count > 30:
|
||||
score -= 0.05
|
||||
|
||||
if score > best_score:
|
||||
best_score = score
|
||||
best_sig = sig
|
||||
|
||||
if best_sig and best_score >= 0.38:
|
||||
# Pondérer la confiance finale par le score et le conf de la signature
|
||||
raw_conf = best_score * best_sig["conf"]
|
||||
confidence = round(min(max(raw_conf, 0.0), 1.0), 2)
|
||||
return OsFingerprint(
|
||||
os_name=best_sig["name"],
|
||||
initial_ttl=initial_ttl,
|
||||
hop_count=hop_count,
|
||||
confidence=confidence,
|
||||
is_bot_tool=best_sig["bot"],
|
||||
network_path=network_path,
|
||||
)
|
||||
|
||||
# Repli : classification TTL seule (confiance minimale)
|
||||
fallback = {64: "Linux/macOS", 128: "Windows", 255: "Cisco/BSD"}
|
||||
return OsFingerprint(
|
||||
os_name=fallback.get(initial_ttl, "Unknown"),
|
||||
initial_ttl=initial_ttl,
|
||||
hop_count=hop_count,
|
||||
confidence=round(0.40 * 0.65, 2), # confiance faible
|
||||
is_bot_tool=False,
|
||||
network_path=network_path,
|
||||
)
|
||||
|
||||
|
||||
def detect_spoof(fp: OsFingerprint, declared_os: str) -> SpoofResult:
|
||||
"""Détecte les incohérences OS entre TCP et UA.
|
||||
|
||||
Règles :
|
||||
1. Outil de scan connu → spoof/bot, quelle que soit l'UA
|
||||
2. Confiance < 0.50 → indéterminable
|
||||
3. OS incompatibles → spoof confirmé
|
||||
4. Cohérent → OK
|
||||
"""
|
||||
if fp.is_bot_tool:
|
||||
return SpoofResult(
|
||||
is_spoof=True,
|
||||
is_bot_tool=True,
|
||||
reason=f"Outil de scan détecté ({fp.os_name})",
|
||||
)
|
||||
|
||||
if fp.confidence < 0.50 or fp.os_name == "Unknown" or declared_os == "Unknown":
|
||||
return SpoofResult(
|
||||
is_spoof=False,
|
||||
is_bot_tool=False,
|
||||
reason="Corrélation insuffisante",
|
||||
)
|
||||
|
||||
tcp_family = _os_family(fp.os_name)
|
||||
ua_family = _ua_os_family(declared_os)
|
||||
|
||||
# Les familles Network/Bot sont toujours suspectes si l'UA prétend être un navigateur
|
||||
if tcp_family == "Network" and ua_family not in ("Network", "Unknown"):
|
||||
return SpoofResult(
|
||||
is_spoof=True,
|
||||
is_bot_tool=False,
|
||||
reason=f"Équipement réseau/CDN (TCP) vs {declared_os} (UA)",
|
||||
)
|
||||
|
||||
if tcp_family == "Unknown" or ua_family == "Unknown":
|
||||
return SpoofResult(is_spoof=False, is_bot_tool=False, reason="OS indéterminé")
|
||||
|
||||
# Android passant par un proxy infra (ex: Facebook app → proxy Windows)
|
||||
# → pas forcément un spoof, noté mais non flaggé
|
||||
if declared_os == "Android" and tcp_family == "Windows" and "proxy" in fp.os_name.lower():
|
||||
return SpoofResult(is_spoof=False, is_bot_tool=False, reason="App mobile via proxy infra")
|
||||
|
||||
if tcp_family != ua_family:
|
||||
return SpoofResult(
|
||||
is_spoof=True,
|
||||
is_bot_tool=False,
|
||||
reason=f"TCP→{tcp_family} vs UA→{ua_family}",
|
||||
)
|
||||
|
||||
return SpoofResult(is_spoof=False, is_bot_tool=False, reason="Cohérent")
|
||||
Reference in New Issue
Block a user