fix: remediate dependency and supply-chain alerts

.clusterfuzzlite/Dockerfile

@@ -1,6 +1,4 @@
-FROM gcr.io/oss-fuzz-base/base-builder-python:v1
-RUN python3 -m pip install --upgrade pip
-RUN python3 -m pip install numpy==2.0.0
+FROM gcr.io/oss-fuzz-base/base-builder-python:v1@sha256:6b2b3a7e4a2da50de47f94925cffbe34747e1aae4f33d7f07ba6c681dd648b23
 COPY . $SRC/SignalGate
 WORKDIR $SRC/SignalGate
 COPY .clusterfuzzlite/build.sh $SRC/build.sh

.clusterfuzzlite/build.sh

@@ -3,11 +3,14 @@
 # ClusterFuzzLite build script.
 # Builds Python fuzzers using OSS-Fuzz helper.
 
-python3 -m pip install --no-cache-dir atheris==2.3.0
+python3 -m pip install \
+  --no-cache-dir \
+  --require-hashes \
+  -r .clusterfuzzlite/requirements.txt
 
-# Install the project (best-effort). If the repo is not installable as a package,
-# fallback to using source imports via PYTHONPATH in fuzzers.
-python3 -m pip install --no-cache-dir -e . || true
+# Make the project root visible to PyInstaller so the packaged fuzzer includes
+# the production sanitizer module instead of relying on a runtime source path.
+export PYTHONPATH="$PWD${PYTHONPATH:+:$PYTHONPATH}"
 
 # Build fuzz targets.
 for fuzzer in fuzz/fuzz_*.py; do

.clusterfuzzlite/requirements.txt

@@ -0,0 +1,8 @@
+atheris==2.3.0 \
+    --hash=sha256:005bde4b5a70e998b7fa097e9aa195972dcc2e04092156a0149cff7aa0de970e \
+    --hash=sha256:6d4e83c8a518e7f7d1c82ee52453255bf6309b7b335657534e83747d4fdfa110 \
+    --hash=sha256:9877b6c2bd5386f9fbbb2989a939c717383d9639f5476411c06fe50fe2fe09a6 \
+    --hash=sha256:b91cb296d60915c3efa4f6db48f09c4678b574cddb7ca98035f1cb9d9fb96f64 \
+    --hash=sha256:c353952ee375bf851527e8b2ea57fdefd7ad16aadfe18143801998075485eccd \
+    --hash=sha256:cf1fdf5fa220a41a2f262b32363fc566549502b2cb0addf4e1baad5531c0e825 \
+    --hash=sha256:e4e43d1ee4760916a84ff73c9c6cf9ac6eee80fc030479bbed43fe0b8e994981

.github/workflows/clusterfuzzlite.yml

@@ -24,9 +24,11 @@ jobs:
         uses: google/clusterfuzzlite/actions/build_fuzzers@884713a6c30a92e5e8544c39945cd7cb630abcd1
         with:
           language: python
+          github-token: ${{ secrets.GITHUB_TOKEN }}
 
       - name: Run fuzzers
         uses: google/clusterfuzzlite/actions/run_fuzzers@884713a6c30a92e5e8544c39945cd7cb630abcd1
         with:
           mode: code-change
           fuzz-seconds: 60
+          github-token: ${{ secrets.GITHUB_TOKEN }}

Dockerfile

@@ -1,21 +1,22 @@
-FROM python:3.11-slim-bookworm AS runtime
+FROM python:3.11-slim-bookworm@sha256:0bee7276f83efd4a1ee05bbbf4281d95ed28e079220a9457f25a93e3f1e3c31b AS runtime
 
 ENV PYTHONDONTWRITEBYTECODE=1 \
     PYTHONUNBUFFERED=1 \
+    PYTHONPATH=/app \
     SIGNALGATE_CONFIG_PATH=/app/config/config.json
 
 RUN addgroup --system signalgate \
     && adduser --system --ingroup signalgate --home /app signalgate
 
 WORKDIR /app
 
-COPY pyproject.toml README.md LICENSE ./
+COPY docker/requirements.txt /app/requirements.txt
+RUN python -m pip install --no-cache-dir --require-hashes -r /app/requirements.txt
+
 COPY signalgate ./signalgate
 COPY docs ./docs
 
-RUN python -m pip install --no-cache-dir --upgrade pip \
-    && python -m pip install --no-cache-dir . \
-    && mkdir -p /app/config /app/data \
+RUN mkdir -p /app/config /app/data \
     && chown -R signalgate:signalgate /app
 
 USER signalgate
@@ -26,4 +27,4 @@ VOLUME ["/app/config", "/app/data"]
 HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
   CMD python -c "import urllib.request; urllib.request.urlopen('http://127.0.0.1:8765/healthz', timeout=2).read()"
 
-CMD ["signalgate"]
+CMD ["python", "-m", "signalgate.cli"]

fuzz/fuzz_sanitize_for_client.py

@@ -7,7 +7,7 @@
 # Ensure local imports work if the package install fails.
 sys.path.insert(0, os.getcwd())
 
-from signalgate.app import _sanitize_for_client  # noqa: E402
+from signalgate.client_sanitize import sanitize_for_client  # noqa: E402
 
 
 def TestOneInput(data: bytes) -> None:
@@ -17,7 +17,7 @@ def TestOneInput(data: bytes) -> None:
     except Exception:
         return
 
-    _sanitize_for_client(obj)
+    sanitize_for_client(obj)
 
 
 def main() -> None:

signalgate/app.py

@@ -15,6 +15,7 @@
 from .budgets import BudgetManager
 from .canary import CanaryConfig, is_canary_user
 from .classifier import KNNTierClassifier
+from .client_sanitize import sanitize_for_client
 from .costing import compute_cost, savings_percent
 from .embeddings import Embedder, build_embedder
 from .errors import SGError, sg_bad_request, sg_payload_too_large, sg_queue_full, sg_unauthorized
@@ -39,55 +40,6 @@
 logger = logging.getLogger("signalgate")
 
 
-_STACKTRACE_MARKERS = (
-    'Traceback (most recent call last):',
-    '  File "',
-)
-
-_BANNED_ERROR_KEYS = {
-    'trace',
-    'traceback',
-    'stack',
-    'stacktrace',
-    'stack_trace',
-    'exc',
-    'exception',
-    'exc_info',
-    'debug_trace',
-    'decision_trace',
-}
-
-
-def _sanitize_for_client(obj: Any) -> Any:
-    """Best-effort scrubber to prevent leaking stack traces or internal debug fields."""
-
-    if obj is None:
-        return None
-
-    if isinstance(obj, str):
-        if any(m in obj for m in _STACKTRACE_MARKERS):
-            return 'redacted'
-        return obj
-
-    if isinstance(obj, (int, float, bool)):
-        return obj
-
-    if isinstance(obj, list):
-        return [_sanitize_for_client(x) for x in obj]
-
-    if isinstance(obj, dict):
-        out: dict[Any, Any] = {}
-        for k, v in obj.items():
-            ks = str(k).lower()
-            if ks in _BANNED_ERROR_KEYS:
-                continue
-            out[k] = _sanitize_for_client(v)
-        return out
-
-    # Fallback for unknown types
-    return str(obj)
-
-
 def _percentile(values: list[int], pct: float) -> int | None:
     if not values:
         return None
@@ -426,7 +378,7 @@ async def _sg_error_handler(_req: Request, exc: SGError):
                 },
             },
         }
-        return JSONResponse(status_code=exc.status_code, content=_sanitize_for_client(body))
+        return JSONResponse(status_code=exc.status_code, content=sanitize_for_client(body))
 
     @app.exception_handler(Exception)
     async def _unhandled_exception_handler(_req: Request, exc: Exception):
@@ -461,7 +413,7 @@ async def _unhandled_exception_handler(_req: Request, exc: Exception):
                 },
             },
         }
-        return JSONResponse(status_code=500, content=_sanitize_for_client(body))
+        return JSONResponse(status_code=500, content=sanitize_for_client(body))
 
     @app.get("/healthz")
     async def healthz():

signalgate/client_sanitize.py

@@ -0,0 +1,49 @@
+"""Lightweight client-error sanitization safe for isolated fuzz packaging."""
+
+from typing import Any
+
+_STACKTRACE_MARKERS = (
+    'Traceback (most recent call last):',
+    '  File "',
+)
+
+_BANNED_ERROR_KEYS = {
+    'trace',
+    'traceback',
+    'stack',
+    'stacktrace',
+    'stack_trace',
+    'exc',
+    'exception',
+    'exc_info',
+    'debug_trace',
+    'decision_trace',
+}
+
+
+def sanitize_for_client(obj: Any) -> Any:
+    """Best-effort scrubber to prevent leaking stack traces or internal debug fields."""
+
+    if obj is None:
+        return None
+
+    if isinstance(obj, str):
+        if any(marker in obj for marker in _STACKTRACE_MARKERS):
+            return 'redacted'
+        return obj
+
+    if isinstance(obj, (int, float, bool)):
+        return obj
+
+    if isinstance(obj, list):
+        return [sanitize_for_client(item) for item in obj]
+
+    if isinstance(obj, dict):
+        out: dict[Any, Any] = {}
+        for key, value in obj.items():
+            if str(key).lower() in _BANNED_ERROR_KEYS:
+                continue
+            out[key] = sanitize_for_client(value)
+        return out
+
+    return str(obj)

tests/unit/test_security_hardening.py

@@ -2,6 +2,7 @@
 
 import pytest
 
+from signalgate.client_sanitize import sanitize_for_client
 from signalgate.errors import SGError
 from signalgate.security import (
     SecurityConfig,
@@ -11,6 +12,23 @@
 )
 
 
+def test_client_sanitizer_removes_nested_debug_data() -> None:
+    sanitized = sanitize_for_client(
+        {
+            "message": "safe",
+            "details": [
+                {"stack_trace": "private", "reason": "still safe"},
+                'Traceback (most recent call last): private',
+            ],
+        }
+    )
+
+    assert sanitized == {
+        "message": "safe",
+        "details": [{"reason": "still safe"}, "redacted"],
+    }
+
+
 def test_tokens_equal_accepts_only_exact_match() -> None:
     assert tokens_equal("secret-token", "secret-token")
     assert not tokens_equal("secret-token", "secret-token-x")