fix: resolve dependency alerts and security regressions

.github/dependabot.yml

@@ -1,7 +1,8 @@
 version: 2
 
-# Dependabot auto-creates PRs against `develop` (not `main`).
-# `main` is reserved for releases and is protected — all updates flow through develop.
+# Gli aggiornamenti di versione puntano a `develop`. GitHub forza invece le PR di
+# sicurezza Dependabot sul branch predefinito (`main`): vanno replicate su `develop`,
+# validate e promosse con la normale PR develop -> main; la PR bot viene poi chiusa.
 # Configuration strategy:
 # - production deps: minor/patch updates auto-applied (fasttrack)
 # - dev deps: only minor/patch (no majors — prevents breaking CI changes)

.github/workflows/codeql.yml

@@ -29,16 +29,16 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v7
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
 
       - name: Initialize CodeQL
-        uses: github/codeql-action/init@v4
+        uses: github/codeql-action/init@9b3254085494b55a0f7c4c2b3d2d52cc712ce370 # v4
         with:
           languages: ${{ matrix.language }}
 
       - name: Set up Python (for Python analysis)
         if: matrix.language == 'python'
-        uses: actions/setup-python@v7
+        uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7
         with:
           python-version: '3.13'
           cache: 'pip'
@@ -49,7 +49,7 @@ jobs:
 
       - name: Set up Node.js (for JavaScript analysis)
         if: matrix.language == 'javascript-typescript'
-        uses: actions/setup-node@v7
+        uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7
         with:
           node-version: '20'
           cache: 'npm'
@@ -60,7 +60,6 @@ jobs:
         run: cd frontend && npm ci
 
       - name: Perform CodeQL Analysis
-        uses: github/codeql-action/analyze@v4
+        uses: github/codeql-action/analyze@9b3254085494b55a0f7c4c2b3d2d52cc712ce370 # v4
         with:
           category: "/language:${{matrix.language}}"
-          upload: ${{ github.ref == 'refs/heads/main' }}

.github/workflows/project-automation.yml

@@ -22,7 +22,7 @@ jobs:
 
     steps:
       - name: "Checkout repository (for file detection)"
-        uses: actions/checkout@v7
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
         with:
           fetch-depth: 0
 

.github/workflows/tests.yml

@@ -29,10 +29,10 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v7
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
 
       - name: Set up Python 3.13
-        uses: actions/setup-python@v7
+        uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7
         with:
           python-version: '3.13'
           cache: 'pip'
@@ -56,15 +56,15 @@ jobs:
 
       - name: Upload test results
         if: always()
-        uses: actions/upload-artifact@v7
+        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7
         with:
           name: pytest-results
           path: backend/pytest-results.xml
           retention-days: 7
 
       - name: Upload coverage report
         if: always()
-        uses: actions/upload-artifact@v7
+        uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7
         with:
           name: coverage-report
           path: backend/coverage.xml
@@ -86,10 +86,10 @@ jobs:
 
     steps:
       - name: Checkout repository
-        uses: actions/checkout@v7
+        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
 
       - name: Set up Node.js
-        uses: actions/setup-node@v7
+        uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7
         with:
           node-version: '20'
           cache: 'npm'

CHANGELOG.md

@@ -22,6 +22,52 @@ Features are ordered by implementation priority.
 
 ---
 
+## [0.17.3] — 2026-10-25
+
+### Security
+
+- Frontend dependency fixes: `source-map-js` 1.2.1 → 1.2.2, Vite 8.3.1 →
+  8.3.5, and `follow-redirects` 1.16.0 → 1.16.1 resolve all nine open
+  Dependabot alerts (denial of service, development-server command/path issues,
+  and cross-origin credential-header disclosure).
+- Updated transitive development dependency Zod 4.3.6 → 4.6.7, resolving an
+  additional high-severity memory-exhaustion advisory detected by `npm audit`
+  before GitHub opened a repository alert.
+- The URLScan diagnostic endpoint now requires a loopback client or the admin
+  token, is limited to two requests per minute, and performs its blocking HTTP
+  diagnostics in a worker thread instead of blocking FastAPI's event loop.
+- Redirect-chain analysis now validates the initial URL and every redirect hop,
+  blocks loopback/private/link-local/reserved destinations, limits redirect
+  depth, and removes credentials and query tokens from its result detail.
+- Replaced the ineffective model-file HMAC (whose public, filename-derived key
+  provided no authenticity) with a release-anchored SHA-256 digest checked
+  before `pickle.loads`. Retraining now prints the digest that must be reviewed.
+- Pinned every GitHub Action to a reviewed full commit SHA. CodeQL now uploads
+  SARIF on pull requests instead of reporting a green check without publishing
+  its findings.
+
+### Fixed
+
+- Reputation hostname extraction now uses `urlparse().hostname` and an exact
+  `www.` prefix removal. Hosts such as `web.example.com`, URLs with userinfo or
+  ports, and IPv6 literals are no longer mangled before URLScan, redirect-chain,
+  crt.sh, CIRCL or SecurityTrails checks.
+- Reputation and URL-analysis logs no longer record full URLs, avoiding leakage
+  of userinfo, query tokens and fragments into application logs.
+- Corrected the Dependabot configuration comments: security updates always
+  target GitHub's default branch, so they must be reproduced and validated on
+  permanent `develop` before promotion to `main`.
+
+### Tests
+
+- Added regression coverage for hostname normalization and trusted-model digest
+  validation.
+- Backend suite: 213 passed, 1 skipped without the optional RTFDE package.
+- `npm audit` and `npm audit --omit=dev` report zero vulnerabilities; the Python
+  requirements report no known vulnerabilities with `pip-audit`.
+
+---
+
 ## [0.17.2] — 2026-10-02
 
 ### Fixed

README.md

@@ -98,6 +98,8 @@ Email (.eml / .msg / plain text)
 
 ## 🔧 Version
 
+**v0.17.3** — 🔒 Security and correctness release: resolves the open frontend dependency advisories for Vite, `source-map-js`, `follow-redirects` and Zod; protects the URLScan diagnostic endpoint; blocks SSRF in redirect-chain analysis; anchors the bundled NLP model to a reviewed SHA-256 digest; fixes hostname extraction in reputation checks; and restores CodeQL result uploads on pull requests while pinning GitHub Actions to full commit SHAs. The suite contains 214 tests (213 pass locally and 1 optional RTFDE test is skipped).
+
 **v0.17.2** — 🛠️ Bugfix release for Outlook `.msg` files: recipients, SPF/DKIM/DMARC results, `Message-ID`, the `Received` chain and the other transport-header fields were silently dropped, so header scoring ran without authentication data (re-analyze older `.msg` analyses to recompute them); RTF-only `.msg` bodies (Outlook 97-2003) are now recoverable with the optional RTFDE package. Also: SQLAlchemy 2.1 with the `asyncio` extra, a high-severity dev-dependency advisory (`brace-expansion`) resolved, other dependency updates, and a repaired project-automation workflow. All 199 tests passing ✅ in CI (the one end-to-end test that needs the optional RTFDE package is skipped where it is not installed), zero regressions.
 
 **v0.17.1** — 🐛 Follow-up to the Threat Intelligence release: fixed CERT-AGID bulletins not loading on first visit, added example placeholders and a collapsible mini-guide to the known-campaign form, and added a "Ri-analizza" (Re-analyze) button so an already-analyzed email can pick up a known campaign created or edited afterward. Documented the whole Threat Intelligence panel in `docs/USAGE.md`.
@@ -216,5 +218,5 @@ Distributed with MIT License · [View License](LICENSE)
 
 ---
 
-*Last updated: 2026-09-15*
-*← [Contributing](./CONTRIBUTING.md) | [Docs →](./docs/)*
+*Last updated: 2026-10-25*
+*← [Contributing](./CONTRIBUTING.md) | [Docs →](./docs/)*

backend/api/routes/reputation.py

@@ -24,7 +24,7 @@
 import asyncio
 import ipaddress
 import logging
-from fastapi import APIRouter, HTTPException, Depends, BackgroundTasks
+from fastapi import APIRouter, HTTPException, Depends, BackgroundTasks, Request
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from sqlalchemy.orm.attributes import flag_modified
@@ -35,7 +35,10 @@
     run_slow_checks,
     finalize_fast_only,
     ReputationSummary,
+    normalize_url_host,
 )
+from core.intel.auth import require_trusted_client
+from core.rate_limiting import limiter
 from dataclasses import asdict
 import json
 
@@ -179,8 +182,7 @@ def add_url(raw: str, is_suspicious: bool = False) -> None:
 
         # Estrai hostname dall'URL
         try:
-            from urllib.parse import urlparse
-            hostname = urlparse(url).netloc.split(":")[0].removeprefix("www.")
+            hostname = normalize_url_host(url)
         except Exception:
             hostname = ""
 
@@ -193,7 +195,7 @@ def add_url(raw: str, is_suspicious: bool = False) -> None:
 
         # Altrimenti, escludi URL da CDN/trusted domains
         if _is_trusted_cdn(hostname):
-            _logger.debug(f"[FILTRO] URL {url} da CDN legittima - escluso")
+            _logger.debug("[FILTRO] Host %s da CDN legittima - escluso", hostname)
             return
 
         seen_urls.add(url)
@@ -211,8 +213,7 @@ def add_domain(raw: str) -> None:
         if not raw:
             return
         try:
-            from urllib.parse import urlparse
-            hostname = urlparse(raw).netloc.split(":")[0].removeprefix("www.")
+            hostname = normalize_url_host(raw)
             # Scarta IP diretti
             if hostname and not (hostname.count(".") < 1 or hostname.replace(".", "").isdigit()):
                 if hostname not in seen_domains and not _is_trusted_cdn(hostname):
@@ -370,13 +371,12 @@ def add_url_if_worth_checking(u: dict) -> None:
 
         # Estrai hostname e controlla se trusted CDN (escludi sempre)
         try:
-            from urllib.parse import urlparse
-            hostname = urlparse(url_str).netloc.split(":")[0].removeprefix("www.")
+            hostname = normalize_url_host(url_str)
         except Exception:
             hostname = ""
 
         if _is_trusted_cdn(hostname):
-            _logger.debug(f"[FILTRO SLOW] URL {url_str} da trusted CDN - esclusa")
+            _logger.debug("[FILTRO SLOW] Host %s da trusted CDN - escluso", hostname)
             return
 
         # Determina se è sospetta (per logging)
@@ -394,9 +394,9 @@ def add_url_if_worth_checking(u: dict) -> None:
         seen_urls.add(url_str)
         urls.append(url_str)
         if is_suspicious:
-            _logger.debug(f"[FILTRO SLOW] URL {url_str} (sospetta, risk_score={u.get('risk_score', 0)}) - inclusa")
+            _logger.debug("[FILTRO SLOW] Host %s sospetto (risk_score=%s) incluso", hostname, u.get("risk_score", 0))
         else:
-            _logger.debug(f"[FILTRO SLOW] URL {url_str} (normale, non da CDN) - inclusa per coverage")
+            _logger.debug("[FILTRO SLOW] Host %s non-CDN incluso per coverage", hostname)
 
     # ─────────────────────────────────────────────────────────────────────────
     # IP: Sender IP + primi 2 hop (mittente -> intermediari prossimi)
@@ -451,8 +451,7 @@ def add_url_if_worth_checking(u: dict) -> None:
         url_str = u.get("original_url") or u.get("url", "")
         if url_str:
             try:
-                from urllib.parse import urlparse
-                hostname = urlparse(url_str).netloc.split(":")[0].removeprefix("www.")
+                hostname = normalize_url_host(url_str)
                 is_suspicious = (
                     u.get("is_ip_address") or u.get("is_ip") or
                     u.get("is_shortener") or u.get("is_new_domain") or
@@ -667,8 +666,7 @@ def to_result(r: dict) -> ReputationResult:
 # Phase 2C: Health Check Endpoint per URLScan.io (v0.14.3+)
 # ---------------------------------------------------------------------------
 
-@router.get("/test-urlscan")
-async def test_urlscan_health():
+def _test_urlscan_health_sync() -> dict:
     """
     Testa la connessione a URLScan.io e verifica la configurazione dell'API key.
     Utile per diagnosticare problemi di autenticazione o rate limiting.
@@ -736,4 +734,11 @@ async def test_urlscan_health():
     else:
         result["suggestions"].append("URLSCAN_API_KEY non configurato in .env. Registrati su https://urlscan.io/user/signup e configura la chiave per aumentare il limite a 1000 req/giorno")
 
-    return result
+    return result
+
+
+@router.get("/test-urlscan", dependencies=[Depends(require_trusted_client)])
+@limiter.limit("2/minute")
+async def test_urlscan_health(request: Request):
+    """Esegue la diagnostica URLScan fuori dall'event loop per client fidati."""
+    return await asyncio.to_thread(_test_urlscan_health_sync)

backend/core/analysis/nlp_classifier.py

@@ -19,7 +19,6 @@
 import threading
 import pickle
 import hashlib
-import hmac
 from pathlib import Path
 from dataclasses import dataclass, field
 from typing import Optional
@@ -357,55 +356,28 @@ class NLPResult:
 
 _tabular_model = None
 _tabular_lock = threading.Lock()
+_TABULAR_MODEL_SHA256 = "ee129c78d7ee9347087bf20b24cdd4e58c34f705e87435f0e56465951276304e"
 
 
-def _verify_and_load_pickle(pkl_path: Path):
-    """Carica il pickle solo se l'HMAC-SHA256 è valido (protezione anti-tampering).
-
-    Se il file HMAC non esiste, crea uno nuovo (init-time). Questo consente
-    il deployment iniziale senza dover pre-generare gli HMAC, ma da quel momento
-    in poi il pickle è protetto.
-
-    Se l'HMAC non corrisponde, rifiuta e lancia un'eccezione.
-    """
-    hmac_path = pkl_path.with_suffix('.pkl.hmac')
-
-    # Leggi i dati del pickle
+def _verify_and_load_pickle(pkl_path: Path, expected_sha256: str):
+    """Carica un pickle solo se coincide con il digest della release fidata."""
     pkl_data = pkl_path.read_bytes()
-
-    # Calcola HMAC-SHA256 dei dati
-    # Nota: usiamo una chiave derivata dal path come "salt" — non è una vera chiave
-    # segreta (il filesystem può essere compromesso interamente), ma aggiunge un
-    # ulteriore livello di protezione contro modifiche accidentali o triviali
-    _secret = f"emlyzer-pickle-{pkl_path.name}".encode()
-    computed_hmac = hmac.new(_secret, pkl_data, hashlib.sha256).hexdigest()
-
-    # Se il file HMAC esiste, verifica
-    if hmac_path.exists():
-        stored_hmac = hmac_path.read_text().strip()
-        if not hmac.compare_digest(computed_hmac, stored_hmac):
-            raise RuntimeError(
-                f"Pickle integrity check FAILED for {pkl_path.name}: "
-                f"HMAC mismatch. File may have been tampered with or corrupted. "
-                f"Refusing to load."
-            )
-    else:
-        # Crea il file HMAC la prima volta (init-time)
-        hmac_path.write_text(computed_hmac)
-        logger.info(f"Created HMAC file for {pkl_path.name} (init-time)")
-
-    # HMAC valido, carica il pickle
+    computed_sha256 = hashlib.sha256(pkl_data).hexdigest()
+    if computed_sha256 != expected_sha256:
+        raise RuntimeError(
+            f"Pickle integrity check FAILED for {pkl_path.name}: "
+            "SHA-256 mismatch. Refusing to deserialize an untrusted model."
+        )
     return pickle.loads(pkl_data)
 
 
 def _load_tabular_model():
     """
     Carica il modello Random Forest serializzato (v0.15.1).
 
-    SECURITY NOTE: The pickle file is generated internally by nlp_retrain_tabular_model.py
-    during the training process. It is NOT loaded from untrusted sources, making it safe
-    to deserialize. The model contains only a trained sklearn RandomForestClassifier and
-    StandardScaler with no arbitrary code execution risk.
+    SECURITY NOTE: pickle can execute code while deserializing. The bundled model is
+    loaded only after matching the SHA-256 digest anchored in this module. When the
+    model is intentionally retrained, the digest must be reviewed and updated together.
     """
     global _tabular_model
 
@@ -423,10 +395,7 @@ def _load_tabular_model():
                 logger.debug(f"Tabular model not found at {model_path}")
                 return None
 
-            # Load with HMAC-SHA256 integrity verification (prevents tampering)
-            # Model is generated internally from training, but filesystem can be compromised
-            # HMAC protects against accidental or malicious file modification
-            model_data = _verify_and_load_pickle(model_path)
+            model_data = _verify_and_load_pickle(model_path, _TABULAR_MODEL_SHA256)
 
             _tabular_model = model_data
             logger.info("Tabular NLP model (Random Forest v0.15.1) loaded successfully")