fix(security): prevent secret_fields from leaking into spend logs

litellm/proxy/litellm_pre_call_utils.py

@@ -1469,7 +1469,12 @@ async def add_litellm_data_to_request(  # noqa: PLR0915
     # been parsed and stripped. Consumers (standard_logging_payload, lago,
     # spend_tracking_utils, streaming_iterator) read `body` to audit the
     # request; taking the snapshot here ensures they see cleaned metadata.
-    data["proxy_server_request"]["body"] = copy.copy(data)
+    #
+    # Exclude secret_fields (which contains raw_headers with Authorization
+    # tokens) from the snapshot — they must never be persisted in spend logs
+    # or any other audit trail.
+    _body_snapshot = {k: v for k, v in data.items() if k != "secret_fields"}
+    data["proxy_server_request"]["body"] = _body_snapshot
 
     # Snapshot the (now-cleaned) requester-supplied metadata for downstream
     # consumers. Taking the deepcopy AFTER the strip prevents attacker-

litellm/proxy/spend_tracking/spend_tracking_utils.py

@@ -610,6 +610,9 @@ def _get_messages_for_spend_logs_payload(
     return "{}"
 
 
+_SENSITIVE_REQUEST_BODY_KEYS = frozenset({"secret_fields"})
+
+
 def _sanitize_request_body_for_spend_logs_payload(
     request_body: dict,
     visited: Optional[set] = None,
@@ -618,6 +621,9 @@ def _sanitize_request_body_for_spend_logs_payload(
     """
     Recursively sanitize request body to prevent logging large base64 strings or other large values.
     Truncates strings longer than MAX_STRING_LENGTH_PROMPT_IN_DB characters and handles nested dictionaries.
+
+    Also strips keys listed in _SENSITIVE_REQUEST_BODY_KEYS (e.g. secret_fields
+    which contains raw HTTP headers including Authorization tokens).
     """
     from litellm.constants import (
         LITELLM_TRUNCATED_PAYLOAD_FIELD,
@@ -676,7 +682,11 @@ def _sanitize_value(value: Any) -> Any:
             return value
         return value
 
-    return {k: _sanitize_value(v) for k, v in request_body.items()}
+    return {
+        k: _sanitize_value(v)
+        for k, v in request_body.items()
+        if k not in _SENSITIVE_REQUEST_BODY_KEYS
+    }
 
 
 def _convert_to_json_serializable_dict(

tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py

@@ -1523,3 +1523,67 @@ def test_master_key_hash_is_rejected(self):
         master = "sk-master-key-123"
         hashed = hash_token(master)
         assert _is_master_key(api_key=hashed, _master_key=master) is False
+
+
+def test_sanitize_request_body_strips_secret_fields():
+    """
+    secret_fields contains raw HTTP headers (including Authorization Bearer
+    tokens). _sanitize_request_body_for_spend_logs_payload must strip it so
+    that sensitive credentials are never persisted in the spend-logs DB.
+    """
+    request_body = {
+        "model": "gpt-4",
+        "messages": [{"role": "user", "content": "hi"}],
+        "secret_fields": {
+            "raw_headers": {
+                "authorization": "Bearer sk-GjX--WwRQmiX2cvASbKf5Q",
+                "content-type": "application/json",
+                "host": "litellm.example.com",
+            }
+        },
+    }
+    sanitized = _sanitize_request_body_for_spend_logs_payload(request_body)
+
+    assert (
+        "secret_fields" not in sanitized
+    ), "secret_fields must be stripped from the sanitized request body"
+    assert sanitized["model"] == "gpt-4"
+    assert sanitized["messages"] == [{"role": "user", "content": "hi"}]
+
+
+@patch(
+    "litellm.proxy.spend_tracking.spend_tracking_utils._should_store_prompts_and_responses_in_spend_logs"
+)
+def test_proxy_server_request_payload_excludes_secret_fields(mock_should_store):
+    """
+    End-to-end test: when the proxy_server_request body contains
+    secret_fields (as it did before the body-snapshot fix), the spend-log
+    serialization must still strip them via the sanitizer fallback.
+    """
+    mock_should_store.return_value = True
+
+    litellm_params = {
+        "proxy_server_request": {
+            "body": {
+                "model": "gpt-4",
+                "messages": [{"role": "user", "content": "hello"}],
+                "secret_fields": {
+                    "raw_headers": {
+                        "authorization": "Bearer sk-super-secret-token",
+                        "host": "litellm.example.com",
+                    }
+                },
+            }
+        }
+    }
+
+    result = _get_proxy_server_request_for_spend_logs_payload(
+        metadata={}, litellm_params=litellm_params, kwargs={}
+    )
+    parsed = json.loads(result)
+
+    assert (
+        "secret_fields" not in parsed
+    ), "secret_fields must never appear in the spend-log proxy_server_request column"
+    assert parsed["model"] == "gpt-4"
+    assert parsed["messages"] == [{"role": "user", "content": "hello"}]

tests/test_litellm/proxy/test_litellm_pre_call_utils.py

@@ -449,6 +449,71 @@ async def test_add_litellm_data_to_request_proxy_server_request_body_is_post_str
     )
 
 
+@pytest.mark.asyncio
+async def test_add_litellm_data_to_request_body_snapshot_excludes_secret_fields():
+    """Security: proxy_server_request['body'] must never contain secret_fields
+    because that dict holds raw HTTP headers including Authorization Bearer
+    tokens. The body snapshot is persisted in spend logs and other audit trails,
+    so leaking secret_fields there exposes user credentials.
+
+    secret_fields must still be available on the live ``data`` dict for
+    downstream consumers (MCP, Responses API) that legitimately need raw headers.
+    """
+    from litellm.proxy.litellm_pre_call_utils import add_litellm_data_to_request
+
+    request_mock = MagicMock(spec=Request)
+    request_mock.url.path = "/v1/chat/completions"
+    request_mock.url = MagicMock()
+    request_mock.url.__str__.return_value = "http://localhost/v1/chat/completions"
+    request_mock.method = "POST"
+    request_mock.query_params = {}
+    request_mock.headers = {
+        "Content-Type": "application/json",
+        "Authorization": "Bearer sk-super-secret-token",
+    }
+    request_mock.client = MagicMock()
+    request_mock.client.host = "127.0.0.1"
+
+    data = {
+        "model": "gpt-3.5-turbo",
+        "messages": [{"role": "user", "content": "hello"}],
+    }
+
+    user_api_key_dict = UserAPIKeyAuth(
+        api_key="hashed-key",
+        user_id="test-user",
+        metadata={},
+        team_metadata={},
+        spend=0.0,
+        max_budget=100.0,
+        model_max_budget={},
+        team_spend=0.0,
+        team_max_budget=200.0,
+    )
+
+    updated = await add_litellm_data_to_request(
+        data=data,
+        request=request_mock,
+        user_api_key_dict=user_api_key_dict,
+        proxy_config=MagicMock(),
+        general_settings={},
+        version="test-version",
+    )
+
+    # secret_fields must exist on the live data dict
+    assert (
+        "secret_fields" in updated
+    ), "secret_fields must still be present on the live data dict"
+    assert "raw_headers" in updated["secret_fields"]
+
+    # But the body snapshot must NOT contain secret_fields
+    snapshot_body = updated["proxy_server_request"]["body"]
+    assert "secret_fields" not in snapshot_body, (
+        "secret_fields must be excluded from proxy_server_request['body'] "
+        "to prevent Authorization tokens from leaking into spend logs"
+    )
+
+
 @pytest.mark.asyncio
 async def test_add_litellm_data_to_request_strips_string_encoded_admin_injection():
     """Regression: metadata arriving as a JSON string (multipart/form-data or