feat(api): remove provider credentials for PDF report (#11845)

api/CHANGELOG.md

@@ -2,6 +2,14 @@
 
 All notable changes to the **Prowler API** are documented in this file.
 
+## [1.34.0] (Prowler UNRELEASED)
+
+### 🚀 Added
+
+- Compliance PDF reports no longer require provider credentials: findings are enriched from the provider metadata stored in the database, so reports generate even after the provider secret is deleted or its credentials become invalid [(#11845)](https://github.com/prowler-cloud/prowler/pull/11845)
+
+---
+
 ## [1.33.1] (Prowler v5.32.1)
 
 ### 🐞 Fixed

api/src/backend/tasks/jobs/report.py

@@ -11,7 +11,6 @@ from uuid import UUID
 from api.db_router import READ_REPLICA_ALIAS, MainRouter
 from api.db_utils import rls_transaction
 from api.models import Provider, Scan, ScanSummary, StateChoices, ThreatScoreSnapshot
-from api.utils import initialize_prowler_provider
 from celery.utils.log import get_task_logger
 from config.django.base import DJANGO_TMP_OUTPUT_DIRECTORY
 from prowler.lib.check.compliance_models import (
@@ -27,6 +26,7 @@ from tasks.jobs.reports import (
     ENSReportGenerator,
     NIS2ReportGenerator,
     ThreatScoreReportGenerator,
+    build_provider_metadata,
 )
 from tasks.jobs.threatscore import compute_threatscore_metrics
 from tasks.jobs.threatscore_utils import (
@@ -841,24 +841,12 @@ def generate_compliance_reports(
         tenant_id, scan_id
     )
 
-    # Initialize the Prowler provider once for the whole report batch. Each
-    # generator used to re-init this in _load_compliance_data, paying the
-    # boto3/Azure-SDK construction cost 5 times per scan. The instance is
-    # only used by FindingOutput.transform_api_finding to enrich findings,
-    # so a single shared instance is correct.
-    logger.info("Initializing prowler_provider once for all reports (scan %s)", scan_id)
-    try:
-        with rls_transaction(tenant_id, using=READ_REPLICA_ALIAS):
-            prowler_provider = initialize_prowler_provider(provider_obj)
-    except Exception as init_error:
-        # If init fails the generators will fall back to lazy init in
-        # _load_compliance_data; we just log and continue.
-        logger.warning(
-            "Could not pre-initialize prowler_provider for scan %s: %s",
-            scan_id,
-            init_error,
-        )
-        prowler_provider = None
+    # Build a credential-free provider metadata stub once for the whole
+    # report batch. FindingOutput.transform_api_finding only reads static
+    # attributes (type plus a few identity fields), so reports never decrypt
+    # the ProviderSecret nor construct a cloud SDK session — generation keeps
+    # working after credentials are deleted or invalidated (PROWLER-2145).
+    prowler_provider = build_provider_metadata(provider_obj)
 
     # Create shared findings cache up front so the eviction closure below
     # can reference it. Defined BEFORE the closure to avoid the UnboundLocalError

api/src/backend/tasks/jobs/reports/__init__.py

@@ -98,6 +98,7 @@ from .config import (
 from .csa import CSAReportGenerator
 from .ens import ENSReportGenerator
 from .nis2 import NIS2ReportGenerator
+from .provider_metadata import build_provider_metadata
 from .threatscore import ThreatScoreReportGenerator
 
 __all__ = [
@@ -105,6 +106,7 @@ __all__ = [
     "BaseComplianceReportGenerator",
     "ComplianceData",
     "RequirementData",
+    "build_provider_metadata",
     "create_pdf_styles",
     "get_requirement_metadata",
     # Framework-specific generators

api/src/backend/tasks/jobs/reports/base.py

@@ -11,7 +11,6 @@ from typing import Any
 from api.db_router import READ_REPLICA_ALIAS
 from api.db_utils import rls_transaction
 from api.models import Provider, StatusChoices
-from api.utils import initialize_prowler_provider
 from celery.utils.log import get_task_logger
 from prowler.lib.check.compliance_models import (
     Compliance,
@@ -52,6 +51,7 @@ from .config import (
     PADDING_SMALL,
     FrameworkConfig,
 )
+from .provider_metadata import build_provider_metadata
 
 logger = get_task_logger(__name__)
 
@@ -178,7 +178,8 @@ class ComplianceData:
         attributes_by_requirement_id: Mapping of requirement IDs to their attributes
         findings_by_check_id: Mapping of check IDs to their findings
         provider_obj: Provider model object
-        prowler_provider: Initialized Prowler provider
+        prowler_provider: Credential-free provider metadata stub (see
+            ``build_provider_metadata``)
     """
 
     tenant_id: str
@@ -439,10 +440,10 @@ class BaseComplianceReportGenerator(ABC):
             provider_obj: Optional pre-fetched Provider object
             requirement_statistics: Optional pre-aggregated statistics
             findings_cache: Optional pre-loaded findings cache
-            prowler_provider: Optional pre-initialized Prowler provider. When
-                generating multiple reports for the same scan the master
-                function initializes this once and passes it in to avoid
-                re-running boto3/Azure-SDK setup per framework.
+            prowler_provider: Optional provider metadata stub (see
+                ``build_provider_metadata``). When generating multiple
+                reports for the same scan the master function builds it
+                once and passes it in.
             **kwargs: Additional framework-specific arguments
         """
         framework = self.config.display_name
@@ -896,9 +897,9 @@ class BaseComplianceReportGenerator(ABC):
             provider_obj: Optional pre-fetched Provider
             requirement_statistics: Optional pre-aggregated statistics
             findings_cache: Optional pre-loaded findings
-            prowler_provider: Optional pre-initialized Prowler provider. When
-                the master function initializes it once and passes it in,
-                we skip the per-report ``initialize_prowler_provider`` call.
+            prowler_provider: Optional provider metadata stub. When the
+                master function builds it once and passes it in, we skip
+                the per-report ``build_provider_metadata`` call.
 
         Returns:
             Aggregated ComplianceData object
@@ -909,7 +910,7 @@ class BaseComplianceReportGenerator(ABC):
                 provider_obj = Provider.objects.get(id=provider_id)
 
             if prowler_provider is None:
-                prowler_provider = initialize_prowler_provider(provider_obj)
+                prowler_provider = build_provider_metadata(provider_obj)
             provider_type = provider_obj.provider
 
             # Load compliance framework — fall back to the universal loader

api/src/backend/tasks/jobs/reports/provider_metadata.py

@@ -0,0 +1,124 @@
+from types import SimpleNamespace
+
+from prowler.providers.github.models import GithubIdentityInfo
+
+
+def build_provider_metadata(provider) -> SimpleNamespace:
+    """Build a credential-free stand-in for the Prowler SDK provider.
+
+    ``FindingOutput.transform_api_finding`` only reads static attributes
+    from the provider (``type`` plus a few identity/metadata fields used to
+    label accounts), so compliance reports never need the decrypted
+    ``ProviderSecret`` nor a live cloud SDK session. This builds an object
+    exposing exactly those attributes from the ``Provider`` DB row, which
+    keeps report generation working when the provider secret has been
+    deleted or its credentials are no longer valid (PROWLER-2145).
+
+    Args:
+        provider: The API ``Provider`` model instance (only ``provider``,
+            ``uid`` and ``alias`` are read).
+
+    Returns:
+        A ``SimpleNamespace`` mimicking the SDK provider attributes consumed
+        by ``FindingOutput.transform_api_finding`` / ``generate_output``.
+    """
+    provider_type = provider.provider
+    uid = provider.uid
+    display_name = provider.alias or uid
+
+    # Defaults cover every attribute read unconditionally in
+    # FindingOutput.generate_output (``provider.auth_method`` is accessed
+    # directly for several provider types); identity lookups go through
+    # get_nested_attribute/getattr, which tolerate missing attributes.
+    stub = SimpleNamespace(
+        type=provider_type,
+        auth_method="",
+        identity=SimpleNamespace(),
+    )
+
+    if provider_type == "aws":
+        stub.identity = SimpleNamespace(account=uid)
+    elif provider_type == "azure":
+        stub.identity = SimpleNamespace(
+            identity_type="",
+            identity_id="",
+            tenant_ids=[""],
+            tenant_domain="",
+            subscriptions={uid: display_name},
+        )
+    elif provider_type == "gcp":
+        stub.identity = SimpleNamespace(profile="")
+        stub.projects = {
+            uid: SimpleNamespace(
+                id=uid,
+                name=display_name,
+                labels={},
+                organization=None,
+            )
+        }
+    elif provider_type == "kubernetes":
+        stub.identity = SimpleNamespace(context=uid, cluster=uid)
+    elif provider_type == "m365":
+        stub.identity = SimpleNamespace(
+            identity_type="",
+            identity_id="",
+            tenant_domain=uid,
+            tenant_id="",
+        )
+    elif provider_type == "github":
+        # generate_output assigns account fields only inside
+        # isinstance(identity, Github*IdentityInfo) branches, so the stub
+        # must carry a real GithubIdentityInfo instance.
+        stub.identity = GithubIdentityInfo(
+            account_id=uid,
+            account_name=display_name,
+            account_url="",
+        )
+    elif provider_type == "mongodbatlas":
+        stub.identity = SimpleNamespace(
+            organization_id=uid,
+            organization_name=display_name,
+        )
+    elif provider_type == "iac":
+        stub.provider_uid = uid
+    elif provider_type == "oraclecloud":
+        stub.identity = SimpleNamespace(
+            tenancy_id=uid,
+            tenancy_name=display_name,
+        )
+    elif provider_type == "alibabacloud":
+        stub.identity = SimpleNamespace(
+            identity_arn="",
+            account_id=uid,
+            account_name=display_name,
+        )
+    elif provider_type == "cloudflare":
+        stub.identity = SimpleNamespace(
+            audited_accounts=[uid],
+            accounts=[],
+        )
+    elif provider_type == "openstack":
+        stub.identity = SimpleNamespace(
+            username="",
+            project_id=uid,
+            project_name=display_name,
+        )
+    elif provider_type == "googleworkspace":
+        stub.identity = SimpleNamespace(
+            delegated_user="",
+            customer_id=uid,
+            domain=display_name,
+        )
+    elif provider_type == "vercel":
+        stub.identity = SimpleNamespace(
+            team=None,
+            user_id=uid,
+            username=display_name,
+        )
+    elif provider_type == "okta":
+        stub.identity = SimpleNamespace(
+            org_domain=uid,
+            client_id="",
+        )
+
+    return stub

api/src/backend/tasks/jobs/threatscore_utils.py

@@ -178,7 +178,9 @@ def _load_findings_for_requirement_checks(
         tenant_id (str): The tenant ID for Row-Level Security context.
         scan_id (str): The ID of the scan to retrieve findings for.
         check_ids (list[str]): List of check IDs to load findings for.
-        prowler_provider: The initialized Prowler provider instance.
+        prowler_provider: Credential-free provider metadata stub (see
+            ``tasks.jobs.reports.build_provider_metadata``) consumed by
+            ``FindingOutput.transform_api_finding``.
         findings_cache (dict, optional): Cache of already loaded findings.
             If provided, checks are first looked up in cache before querying database.
         total_counts_out (dict, optional): If provided, populated with

api/src/backend/tasks/tests/test_reports.py

@@ -1044,10 +1044,10 @@ class TestStaleCleanupProtectionHelpers:
 class TestGenerateThreatscoreReportFunction:
     """Test suite for generate_threatscore_report function."""
 
-    @patch("tasks.jobs.reports.base.initialize_prowler_provider")
+    @patch("tasks.jobs.reports.base.build_provider_metadata")
     def test_generate_threatscore_report_exception_handling(
         self,
-        mock_initialize_provider,
+        mock_build_provider_metadata,
         tenants_fixture,
         scans_fixture,
         providers_fixture,
@@ -1057,7 +1057,7 @@ class TestGenerateThreatscoreReportFunction:
         scan = scans_fixture[0]
         provider = providers_fixture[0]
 
-        mock_initialize_provider.side_effect = Exception("Test exception")
+        mock_build_provider_metadata.side_effect = Exception("Test exception")
 
         with pytest.raises(Exception) as exc_info:
             generate_threatscore_report(
@@ -1167,7 +1167,6 @@ class TestGenerateComplianceReportsOptimized:
         assert result["cis"] == {"upload": False, "path": ""}
         mock_cis.assert_not_called()
 
-    @patch("api.utils.initialize_prowler_provider")
     @patch("tasks.jobs.report.rmtree")
     @patch("tasks.jobs.report._upload_to_s3")
     @patch("tasks.jobs.report.generate_cis_report")
@@ -1194,7 +1193,6 @@ class TestGenerateComplianceReportsOptimized:
         mock_cis,
         mock_upload_to_s3,
         mock_rmtree,
-        mock_init_provider,
     ):
         """After each framework finishes, exclusive entries are evicted.
 
@@ -1223,7 +1221,6 @@ class TestGenerateComplianceReportsOptimized:
         mock_aggregate_stats.return_value = {}
         mock_generate_output_dir.return_value = "/tmp/tenant/scan/x/prowler-out"
         mock_upload_to_s3.return_value = "s3://bucket/tenant/scan/x/report.pdf"
-        mock_init_provider.return_value = Mock(name="prowler_provider")
 
         # Seed the cache as if both frameworks had already loaded their
         # findings. We mutate it indirectly: each generator wrapper is a
@@ -1266,7 +1263,7 @@ class TestGenerateComplianceReportsOptimized:
             "shared must remain in cache because ENS still needs it"
         )
 
-    @patch("tasks.jobs.report.initialize_prowler_provider")
+    @patch("tasks.jobs.report.build_provider_metadata")
     @patch("tasks.jobs.report.rmtree")
     @patch("tasks.jobs.report._upload_to_s3")
     @patch("tasks.jobs.report.generate_cis_report")
@@ -1279,7 +1276,7 @@ class TestGenerateComplianceReportsOptimized:
     @patch("tasks.jobs.report.Compliance.get_bulk")
     @patch("tasks.jobs.report.Provider.objects.get")
     @patch("tasks.jobs.report.ScanSummary.objects.filter")
-    def test_prowler_provider_initialized_once(
+    def test_provider_metadata_built_once(
         self,
         mock_scan_summary_filter,
         mock_provider_get,
@@ -1293,11 +1290,11 @@ class TestGenerateComplianceReportsOptimized:
         mock_cis,
         mock_upload_to_s3,
         mock_rmtree,
-        mock_init_provider,
+        mock_build_metadata,
     ):
-        """``initialize_prowler_provider`` must be called exactly once for
-        the whole batch (PROWLER-1733). Previously each generator re-init'd
-        the SDK provider in ``_load_compliance_data`` → 5 inits per scan.
+        """``build_provider_metadata`` must be called exactly once for the
+        whole batch and its result shared across all 5 reports
+        (PROWLER-1733 / PROWLER-2145).
         """
         mock_scan_summary_filter.return_value.exists.return_value = True
         mock_provider_get.return_value = Mock(uid="provider-uid", provider="aws")
@@ -1306,7 +1303,7 @@ class TestGenerateComplianceReportsOptimized:
         mock_aggregate_stats.return_value = {}
         mock_generate_output_dir.return_value = "/tmp/tenant/scan/x/prowler-out"
         mock_upload_to_s3.return_value = "s3://bucket/tenant/scan/x/report.pdf"
-        mock_init_provider.return_value = Mock(name="prowler_provider")
+        mock_build_metadata.return_value = Mock(name="prowler_provider")
 
         generate_compliance_reports(
             tenant_id=str(uuid.uuid4()),
@@ -1325,14 +1322,14 @@ class TestGenerateComplianceReportsOptimized:
         mock_nis2.assert_called_once()
         mock_csa.assert_called_once()
         mock_cis.assert_called_once()
-        # …but the SDK provider was initialized only once.
-        assert mock_init_provider.call_count == 1, (
-            f"expected 1 init, got {mock_init_provider.call_count} "
+        # …but the provider metadata stub was built only once.
+        assert mock_build_metadata.call_count == 1, (
+            f"expected 1 build, got {mock_build_metadata.call_count} "
             f"(prowler_provider must be shared across reports)"
         )
 
         # The shared instance must reach every wrapper as kwargs.
-        shared = mock_init_provider.return_value
+        shared = mock_build_metadata.return_value
         for mock_wrapper in (
             mock_threatscore,
             mock_ens,

api/src/backend/tasks/tests/test_reports_provider_metadata.py

@@ -0,0 +1,193 @@
+"""Tests for the credential-free provider metadata stub (PROWLER-2145).
+
+Every provider object used here is a plain ``SimpleNamespace`` WITHOUT a
+``secret`` attribute: any code path trying to read ``provider.secret`` (the
+coupling these tests guard against) would raise ``AttributeError`` and fail
+the test. No database is required.
+"""
+
+from types import SimpleNamespace
+
+import pytest
+from api.models import Provider
+from prowler.lib.outputs.finding import Finding as FindingOutput
+from prowler.providers.github.models import GithubIdentityInfo
+from tasks.jobs.reports import build_provider_metadata
+
+PROVIDER_UID = "provider-uid-123"
+PROVIDER_ALIAS = "my-provider-alias"
+
+
+def _provider_row(provider_type: str, alias: str | None = PROVIDER_ALIAS):
+    """Mimic the Provider DB row attributes read by build_provider_metadata."""
+    return SimpleNamespace(provider=provider_type, uid=PROVIDER_UID, alias=alias)
+
+
+class TestBuildProviderMetadata:
+    @pytest.mark.parametrize("provider_type", Provider.ProviderChoices.values)
+    def test_every_provider_type_gets_safe_defaults(self, provider_type):
+        stub = build_provider_metadata(_provider_row(provider_type))
+
+        assert stub.type == provider_type
+        assert isinstance(stub.auth_method, str)
+        assert hasattr(stub, "identity")
+
+    def test_aws_identity_account_is_uid(self):
+        stub = build_provider_metadata(_provider_row("aws"))
+        assert stub.identity.account == PROVIDER_UID
+
+    def test_azure_identity_covers_generate_output_accesses(self):
+        stub = build_provider_metadata(_provider_row("azure"))
+
+        # generate_output indexes tenant_ids[0] and reads these directly.
+        assert stub.identity.tenant_ids
+        assert stub.identity.identity_type == ""
+        assert stub.identity.identity_id == ""
+        assert stub.identity.subscriptions == {PROVIDER_UID: PROVIDER_ALIAS}
+
+    def test_gcp_projects_keyed_by_uid(self):
+        stub = build_provider_metadata(_provider_row("gcp"))
+
+        project = stub.projects[PROVIDER_UID]
+        assert project.id == PROVIDER_UID
+        assert project.name == PROVIDER_ALIAS
+        assert project.labels == {}
+        # generate_output calls getattr(project, "organization") without a
+        # default, so the attribute must exist (None skips the org branch).
+        assert project.organization is None
+
+    def test_kubernetes_identity_context_and_cluster(self):
+        stub = build_provider_metadata(_provider_row("kubernetes"))
+        assert stub.identity.context == PROVIDER_UID
+        assert stub.identity.cluster == PROVIDER_UID
+
+    def test_github_identity_is_real_identity_info(self):
+        # generate_output only assigns account fields inside
+        # isinstance(identity, Github*IdentityInfo) branches.
+        stub = build_provider_metadata(_provider_row("github"))
+        assert isinstance(stub.identity, GithubIdentityInfo)
+        assert stub.identity.account_id == PROVIDER_UID
+        assert stub.identity.account_name == PROVIDER_ALIAS
+
+    def test_iac_provider_uid(self):
+        stub = build_provider_metadata(_provider_row("iac"))
+        assert stub.provider_uid == PROVIDER_UID
+
+    def test_alias_falls_back_to_uid(self):
+        stub = build_provider_metadata(_provider_row("azure", alias=None))
+        assert stub.identity.subscriptions == {PROVIDER_UID: PROVIDER_UID}
+
+
+def _check_metadata_dict(provider_type: str, check_id: str) -> dict:
+    return {
+        "provider": provider_type,
+        "checkid": check_id,
+        "checktitle": "Test check title",
+        "checktype": [],
+        # CheckMetadata validates ServiceName == check_id.split("_")[0]
+        "servicename": check_id.split("_")[0],
+        "subservicename": "",
+        "severity": "high",
+        "resourcetype": "resource-type",
+        "description": "",
+        "risk": "",
+        "relatedurl": "",
+        "remediation": {
+            "recommendation": {"text": "", "url": ""},
+            "code": {"nativeiac": "", "terraform": "", "cli": "", "other": ""},
+        },
+        "resourceidtemplate": "",
+        "categories": [],
+        "dependson": [],
+        "relatedto": [],
+        "notes": "",
+    }
+
+
+class _FakeFinding:
+    """Attribute-faithful Finding stand-in.
+
+    A plain object instead of ``Mock``: only the attributes the Django model
+    exposes exist, so any new provider-attribute read in generate_output
+    (e.g. cloudflare's ``getattr(finding, "account_id", ...)``) hits the
+    same missing-attribute path it would hit in production instead of being
+    masked by Mock auto-created attributes.
+    """
+
+
+def _finding_model(provider_type: str, check_id: str, region: str):
+    """Mimic the Django Finding row attributes read by transform_api_finding."""
+    resource = SimpleNamespace(
+        uid="resource-uid",
+        name="resource-name",
+        metadata="{}",
+        details="",
+        region=region,
+        tags=SimpleNamespace(all=lambda: []),
+    )
+    finding = _FakeFinding()
+    finding.resources = SimpleNamespace(first=lambda: resource)
+    finding.check_metadata = _check_metadata_dict(provider_type, check_id)
+    finding.status = "FAIL"
+    finding.status_extended = "failed for testing"
+    finding.muted = False
+    return finding
+
+
+_FINDING_REGION = "region-x"
+
+# Expected (account_uid, region) of the transformed finding per provider
+# type, with resource.region = _FINDING_REGION. Keyed by every
+# Provider.ProviderChoices value so that adding a new provider type without
+# extending build_provider_metadata (and this table) fails the test below
+# instead of breaking PDF generation at runtime.
+_EXPECTED_TRANSFORM = {
+    "aws": (PROVIDER_UID, _FINDING_REGION),
+    "azure": (PROVIDER_UID, _FINDING_REGION),
+    "gcp": (PROVIDER_UID, _FINDING_REGION),
+    # transform_api_finding strips the "namespace: " prefix and
+    # generate_output re-adds it.
+    "kubernetes": (PROVIDER_UID, f"namespace: {_FINDING_REGION}"),
+    "m365": (PROVIDER_UID, _FINDING_REGION),
+    # For GitHub the owner comes from resource.region.
+    "github": (_FINDING_REGION, _FINDING_REGION),
+    "mongodbatlas": (PROVIDER_UID, _FINDING_REGION),
+    "iac": (PROVIDER_UID, _FINDING_REGION),
+    "oraclecloud": (PROVIDER_UID, _FINDING_REGION),
+    "alibabacloud": (PROVIDER_UID, _FINDING_REGION),
+    # Cloudflare uses the zone name (falls back to resource.name) as region.
+    "cloudflare": (PROVIDER_UID, "resource-name"),
+    "openstack": (PROVIDER_UID, _FINDING_REGION),
+    "image": ("image", _FINDING_REGION),
+    "googleworkspace": (PROVIDER_UID, _FINDING_REGION),
+    "vercel": (PROVIDER_UID, "global"),
+    "okta": (PROVIDER_UID, "global"),
+}
+
+
+class TestTransformApiFindingWithMetadataStub:
+    """transform_api_finding must work end-to-end with the stub — i.e.
+    without a credentialed SDK provider — for EVERY API provider type."""
+
+    @pytest.mark.parametrize("provider_type", Provider.ProviderChoices.values)
+    def test_transform_with_stub(self, provider_type):
+        assert provider_type in _EXPECTED_TRANSFORM, (
+            f"New provider type {provider_type!r}: add a branch to "
+            f"build_provider_metadata covering the attributes read by "
+            f"FindingOutput.generate_output, then add its expected "
+            f"(account_uid, region) here."
+        )
+        expected_account_uid, expected_region = _EXPECTED_TRANSFORM[provider_type]
+
+        stub = build_provider_metadata(_provider_row(provider_type))
+        check_id = f"{provider_type}_test_check"
+        finding_model = _finding_model(provider_type, check_id, _FINDING_REGION)
+
+        output = FindingOutput.transform_api_finding(finding_model, stub)
+
+        assert output.check_id == check_id
+        assert output.status == "FAIL"
+        assert output.account_uid == expected_account_uid
+        assert output.region == expected_region
+        assert output.resource_name
+        assert output.resource_uid