code_public/dashboard
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(clustering): intégration Fingerprint HTTP Headers (agg_header_fingerprint_1h)

Browse Source 
Sources des nouvelles features :
- agg_header_fingerprint_1h : Cookie, Referer par src_ip (JOIN sur IPv6)
- ml_detected_anomalies : header_order_shared_count, distinct_header_orders (déjà jointé)

Nouvelles features (indices 27-30) :
  [27] FP Popularité  : popularité du fingerprint headers (log1p/log1p(500k))
       fingerprint rare (bot artisanal) → 0.0 ; très populaire (browser) → 1.0
  [28] FP Rotation    : distinct_header_orders (log1p/log1p(10))
       rotation de fingerprint entre requêtes = comportement bot
  [29] Cookie Présent : présence header Cookie (engagement utilisateur réel)
  [30] Referer Présent: présence header Referer (navigation HTTP normale)

risk_score_from_centroid() : 14 termes, somme=1.0
  + hfp_rare (1-popularité) × 0.06 + hfp_rotating × 0.06
  ML × 0.25 reste dominant

name_cluster() : 2 nouveaux labels
  '🔄 Bot fingerprint tournant' : hfp_rotating>0.6 + anomalie>0.15
  '🕵️ Fingerprint rare suspect' : hfp_popular<0.15 + anomalie>0.20
  '🌐 Navigateur légitime' : fingerprint populaire confirmé

N_FEATURES : 27 → 31

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This commit is contained in:
SOC Analyst
2026-03-19 11:13:37 +01:00
parent 8fb054c8b7
commit 6ff59a36d7
2 changed files with 114 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
backend/routes/clustering.py
View File

@ -95,7 +95,18 @@ SELECT
avg(ml.has_accept_language) AS hdr_accept_lang,
any(vh.hdr_enc) AS hdr_has_encoding,
any(vh.hdr_sec_fetch) AS hdr_has_sec_fetch,
any(vh.hdr_count) AS hdr_count_raw
any(vh.hdr_count) AS hdr_count_raw,
-- Fingerprint HTTP Headers (depuis agg_header_fingerprint_1h + ml_detected_anomalies)
-- header_order_shared_count : nb d'IPs partageant le même fingerprint
-- → faible = fingerprint rare = comportement suspect
avg(ml.header_order_shared_count) AS hfp_shared_count,
-- distinct_header_orders : nb de fingerprints distincts émis par cette IP
-- → élevé = rotation de fingerprint = comportement bot
avg(ml.distinct_header_orders) AS hfp_distinct_orders,
-- Cookie et Referer issus de la table dédiée aux empreintes
any(hfp.hfp_cookie) AS hfp_cookie,
any(hfp.hfp_referer) AS hfp_referer
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
@ -112,6 +123,15 @@ LEFT JOIN (
AND log_date >= today() - 2
GROUP BY src_ip_v6, ja4
) vh ON t.src_ip = vh.src_ip_v6 AND t.ja4 = vh.ja4
LEFT JOIN (
SELECT
src_ip,
avg(has_cookie) AS hfp_cookie,
avg(has_referer) AS hfp_referer
FROM mabase_prod.agg_header_fingerprint_1h
WHERE window_start >= now() - INTERVAL %(hours)s HOUR
GROUP BY src_ip
) hfp ON t.src_ip = hfp.src_ip
WHERE t.window_start >= now() - INTERVAL %(hours)s HOUR
AND t.tcp_ttl_raw > 0
GROUP BY t.src_ip, t.ja4
@ -124,6 +144,7 @@ _SQL_COLS = [
"h2_eff", "hdr_conf", "ua_ch_mismatch", "asset_ratio", "direct_ratio",
"ja4_count", "ua_rotating", "threat", "country", "asn_org",
"hdr_accept_lang", "hdr_has_encoding", "hdr_has_sec_fetch", "hdr_count_raw",
"hfp_shared_count", "hfp_distinct_orders", "hfp_cookie", "hfp_referer",
]

103
backend/services/clustering_engine.py
View File

@ -6,7 +6,7 @@ Ref:
scipy.spatial.ConvexHull — enveloppe convexe (Graham/Qhull)
sklearn-style API — centroids, labels_, inertia_
Features (27 dimensions, normalisées [0,1]) :
Features (31 dimensions, normalisées [0,1]) :
0 ttl_n : TTL initial normalisé
1 mss_n : MSS normalisé → type réseau
2 scale_n : facteur de mise à l'échelle TCP
@ -34,6 +34,12 @@ Features (27 dimensions, normalisées [0,1]) :
24 hdr_encoding_n : présence header Accept-Encoding (0=absent=bot-like)
25 hdr_sec_fetch_n : présence headers Sec-Fetch-* (1=navigateur réel)
26 hdr_count_n : nombre de headers HTTP normalisé (3=bot, 15=browser)
27 hfp_popular_n : popularité du fingerprint headers (log-normalisé)
fingerprint rare = suspect ; très populaire = browser légitime
28 hfp_rotating_n : rotation de fingerprint (distinct_header_orders)
plusieurs fingerprints distincts → bot en rotation
29 hfp_cookie_n : présence header Cookie (engagement utilisateur réel)
30 hfp_referer_n : présence header Referer (navigation HTTP normale)
"""
from __future__ import annotations
@ -155,6 +161,16 @@ FEATURES: list[tuple[str, str, object]] = [
("hdr_has_encoding", "Accept-Encoding", lambda v: 1.0 if float(v or 0) > 0 else 0.0),
("hdr_has_sec_fetch", "Sec-Fetch Headers", lambda v: 1.0 if float(v or 0) > 0 else 0.0),
("hdr_count_raw", "Nb Headers", lambda v: min(1.0, float(v or 0) / 20.0)),
# ── Fingerprint HTTP Headers (agg_header_fingerprint_1h) ──────────────
# header_order_shared_count : nb d'IPs partageant ce fingerprint
# élevé → populaire → browser légitime (normalisé log1p / log1p(500000))
("hfp_shared_count", "FP Popularité", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(500_000))),
# distinct_header_orders : nb de fingerprints distincts pour cette IP
# élevé → rotation de fingerprint → bot (normalisé log1p / log1p(10))
("hfp_distinct_orders", "FP Rotation", lambda v: min(1.0, math.log1p(float(v or 0)) / math.log1p(10))),
# Cookie et Referer : signaux de navigation légitime
("hfp_cookie", "Cookie Présent", lambda v: min(1.0, float(v or 0))),
("hfp_referer", "Referer Présent", lambda v: min(1.0, float(v or 0))),
]
FEATURE_KEYS = [f[0] for f in FEATURES]
@ -334,38 +350,45 @@ def compute_hulls(coords_2d: np.ndarray, labels: np.ndarray,
def name_cluster(centroid: np.ndarray, raw_stats: dict) -> str:
"""Nom lisible basé sur les features dominantes du centroïde [0,1]."""
s = centroid
n = len(s)
ttl_raw = float(raw_stats.get("mean_ttl", 0))
mss_raw = float(raw_stats.get("mean_mss", 0))
country_risk_v = s[21] if len(s) > 21 else 0.0
asn_cloud = s[22] if len(s) > 22 else 0.0
# Features headers (indices 23-26)
accept_lang = s[23] if len(s) > 23 else 1.0
accept_enc = s[24] if len(s) > 24 else 1.0
sec_fetch = s[25] if len(s) > 25 else 0.0
hdr_count = s[26] if len(s) > 26 else 0.5
country_risk_v = s[21] if n > 21 else 0.0
asn_cloud = s[22] if n > 22 else 0.0
accept_lang = s[23] if n > 23 else 1.0
accept_enc = s[24] if n > 24 else 1.0
sec_fetch = s[25] if n > 25 else 0.0
hdr_count = s[26] if n > 26 else 0.5
hfp_popular = s[27] if n > 27 else 0.5
hfp_rotating = s[28] if n > 28 else 0.0
# Scanner pur : aucun header browser, peu de headers
# Scanner pur : aucun header browser, fingerprint rare, peu de headers
if accept_lang < 0.15 and accept_enc < 0.15 and hdr_count < 0.25:
return "🤖 Scanner pur (no headers)"
# Fingerprint tournant ET suspect : bot qui change de profil headers
if hfp_rotating > 0.6 and s[4] > 0.15:
return "🔄 Bot fingerprint tournant"
# Fingerprint très rare et anomalie : bot artisanal unique
if hfp_popular < 0.15 and s[4] > 0.20:
return "🕵️ Fingerprint rare suspect"
# Scanners Masscan
if s[0] > 0.16 and s[0] < 0.25 and mss_raw in range(1440, 1460) and s[2] > 0.25:
return "🤖 Masscan Scanner"
# Bots offensifs agressifs (fuzzing + anomalie + pas de headers browser)
# Bots offensifs agressifs (fuzzing + anomalie)
if s[4] > 0.40 and s[6] > 0.3:
return "🤖 Bot agressif"
# Bot qui simule un navigateur mais sans les vrais headers (ua_ch + absent sec_fetch)
# Bot qui simule un navigateur mais sans les vrais headers
if s[16] > 0.40 and sec_fetch < 0.2 and accept_lang < 0.3:
return "🤖 Bot UA simulé"
# Pays à très haut risque (CN, RU, KP) avec trafic anormal
# Pays à très haut risque avec trafic anormal
if country_risk_v > 0.75 and (s[4] > 0.10 or asn_cloud > 0.5):
return "🌏 Source pays risqué"
# Cloud + UA-CH mismatch = crawler/bot cloud
# Cloud + UA-CH mismatch
if s[16] > 0.50 and asn_cloud > 0.70:
return "☁️ Bot cloud UA-CH"
# UA-CH mismatch seul
if s[16] > 0.60:
return "🤖 UA-CH Mismatch"
# Headless browser avec headers browser réels (Puppeteer, Playwright)
# Headless browser (Puppeteer/Playwright) : a les headers Sec-Fetch mais headless
if s[7] > 0.50 and sec_fetch > 0.5:
return "🤖 Headless Browser"
if s[7] > 0.50:
@ -379,8 +402,9 @@ def name_cluster(centroid: np.ndarray, raw_stats: dict) -> str:
# Pays à risque élevé sans autre signal
if country_risk_v > 0.60:
return "🌏 Trafic suspect (pays)"
# Navigateur légitime : tous les headers présents
if accept_lang > 0.7 and accept_enc > 0.7 and sec_fetch > 0.6 and hdr_count > 0.5:
# Navigateur légitime : tous les signaux positifs y compris fingerprint populaire
if (accept_lang > 0.7 and accept_enc > 0.7 and sec_fetch > 0.5
and hdr_count > 0.5 and hfp_popular > 0.5):
return "🌐 Navigateur légitime"
# OS fingerprinting
if s[3] > 0.85 and ttl_raw > 120:
@ -399,31 +423,34 @@ def name_cluster(centroid: np.ndarray, raw_stats: dict) -> str:
def risk_score_from_centroid(centroid: np.ndarray) -> float:
"""
Score de risque [0,1] depuis le centroïde (espace original [0,1]).
Intègre pays, infrastructure cloud et profil headers HTTP.
Poids calibrés pour sommer à 1.0.
31 features — poids calibrés pour sommer à 1.0.
"""
s = centroid
country_risk_v = s[21] if len(s) > 21 else 0.0
asn_cloud = s[22] if len(s) > 22 else 0.0
# Absence de header = risque → inverser (1 - présence)
no_accept_lang = 1.0 - (s[23] if len(s) > 23 else 1.0)
no_encoding = 1.0 - (s[24] if len(s) > 24 else 1.0)
no_sec_fetch = 1.0 - (s[25] if len(s) > 25 else 0.0)
# Peu de headers → bot : max risque quand hdr_count=0
few_headers = 1.0 - (s[26] if len(s) > 26 else 0.5)
n = len(s)
country_risk_v = s[21] if n > 21 else 0.0
asn_cloud = s[22] if n > 22 else 0.0
no_accept_lang = 1.0 - (s[23] if n > 23 else 1.0)
no_encoding = 1.0 - (s[24] if n > 24 else 1.0)
no_sec_fetch = 1.0 - (s[25] if n > 25 else 0.0)
few_headers = 1.0 - (s[26] if n > 26 else 0.5)
# Fingerprint rare = suspect (faible popularité), fingerprint tournant = bot
hfp_rare = 1.0 - (s[27] if n > 27 else 0.5)
hfp_rotating = s[28] if n > 28 else 0.0
return float(np.clip(
0.28 * s[4] + # score ML anomalie (principal)
0.10 * s[6] + # fuzzing
0.08 * s[16] + # UA-CH mismatch
0.07 * s[7] + # headless
0.06 * s[5] + # vélocité
0.06 * s[9] + # IP-ID zéro
0.10 * country_risk_v+ # risque pays source
0.07 * asn_cloud + # infrastructure cloud/VPN
0.05 * no_accept_lang+ # absence Accept-Language
0.05 * no_encoding + # absence Accept-Encoding
0.25 * s[4] + # score ML anomalie (principal)
0.09 * s[6] + # fuzzing
0.07 * s[16] + # UA-CH mismatch
0.06 * s[7] + # headless
0.05 * s[5] + # vélocité
0.05 * s[9] + # IP-ID zéro
0.09 * country_risk_v+ # risque pays source
0.06 * asn_cloud + # infrastructure cloud/VPN
0.04 * no_accept_lang+ # absence Accept-Language
0.04 * no_encoding + # absence Accept-Encoding
0.04 * no_sec_fetch + # absence Sec-Fetch (pas un vrai navigateur)
0.04 * few_headers, # très peu de headers (scanner/curl)
0.04 * few_headers + # très peu de headers (scanner/curl)
0.06 * hfp_rare + # fingerprint headers rare = suspect
0.06 * hfp_rotating, # rotation de fingerprint = bot
0.0, 1.0
))