Testing

.github/workflows/semgrep.yml

@@ -0,0 +1,26 @@
+on:
+  workflow_dispatch: {}
+  pull_request: {}
+  push:
+    branches:
+    - main
+    - master
+    paths:
+    - .github/workflows/semgrep.yml
+  schedule:
+  # random HH:MM to avoid a load spike on GitHub Actions at 00:00
+  - cron: 6 9 * * *
+name: Semgrep
+jobs:
+  semgrep:
+    name: semgrep/ci
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    env:
+      SEMGREP_APP_TOKEN: ${{ secrets.SEMGREP_APP_TOKEN }}
+    container:
+      image: semgrep/semgrep
+    steps:
+    - uses: actions/checkout@v6
+    - run: semgrep ci

.semgrepignore

@@ -0,0 +1 @@
+test.py

path-traversal-chained-test.py

@@ -0,0 +1,34 @@
+import os
+
+UPLOAD_DIR = "/var/uploads"
+SAFE_DIR = "/var/safe"
+
+# ruleid: path-traversal-open-os-path-join
+# SHOULD MATCH — user_filename flows directly into os.path.join then open()
+def bad_file_read(user_filename):
+    path = os.path.join(UPLOAD_DIR, user_filename)
+    with open(path, 'r') as f:
+        return f.read()
+
+# ruleid: path-traversal-open-os-path-join
+# SHOULD MATCH — inline: os.path.join passed directly into open() in one expression
+def bad_inline(user_filename):
+    with open(os.path.join(UPLOAD_DIR, user_filename), 'rb') as f:
+        return f.read()
+
+# ok: path-traversal-open-os-path-join
+# SHOULD NOT MATCH — hardcoded filename, no user input
+def safe_hardcoded():
+    path = os.path.join(UPLOAD_DIR, "config.json")
+    with open(path, 'r') as f:
+        return f.read()
+
+# ok: path-traversal-open-os-path-join
+# SHOULD NOT MATCH — path is validated before open()
+def safe_validated(user_filename):
+    joined = os.path.join(UPLOAD_DIR, user_filename)
+    safe = os.path.abspath(joined)
+    if not safe.startswith(os.path.abspath(UPLOAD_DIR)):
+        raise ValueError("Path traversal detected")
+    with open(safe, 'r') as f:
+        return f.read()

path-traversal-chained.yaml

@@ -0,0 +1,38 @@
+rules:
+  - id: path-traversal-open-os-path-join-inline
+    pattern: open(os.path.join($BASE, $USER_INPUT), ...)
+    pattern-not: open(os.path.join("...", "..."), ...)
+    message: >
+      Path traversal: os.path.join() result passed directly to open() without
+      path validation. An attacker controlling the second argument can escape
+      the base directory using '../' sequences. Fix: resolve the path with
+      os.path.abspath() and assert it starts with os.path.abspath($BASE).
+    languages: [python]
+    severity: ERROR
+    metadata:
+      cwe: "CWE-22: Improper Limitation of a Pathname to a Restricted Directory"
+      owasp: "A01:2021 - Broken Access Control"
+      confidence: HIGH
+      category: security
+
+  - id: path-traversal-open-os-path-join-two-step
+    patterns:
+      - pattern: |
+          $PATH = os.path.join($BASE, $USER_INPUT)
+          open($PATH, ...)
+      - pattern-not: |
+          $PATH = os.path.join("...", "...")
+          open($PATH, ...)
+    message: >
+      Path traversal: a path built with os.path.join() is passed to open()
+      without validation in between. An attacker controlling the second
+      argument of os.path.join() can escape the base directory. Fix: call
+      os.path.abspath() on $PATH and verify it starts with
+      os.path.abspath($BASE) before calling open().
+    languages: [python]
+    severity: ERROR
+    metadata:
+      cwe: "CWE-22: Improper Limitation of a Pathname to a Restricted Directory"
+      owasp: "A01:2021 - Broken Access Control"
+      confidence: HIGH
+      category: security

requirements.txt

@@ -1,4 +1,6 @@
+ansible-core==2.16.0
 certifi==2021.10.8
+chardet==3.0.4
 charset-normalizer==2.0.12
 click==8.0.1
 Flask==2.0.1

test-vuln.py

@@ -0,0 +1,22 @@
+import sh
+
+long = os.environ.get("LONG", "")
+# ruleid: string-concat
+sh.ls("-a" + long)
+
+# ok: string-concat
+sh.ls("-al")
+
+# ok: string-concat
+sh.semgrep("--config", "https://semgrep.dev/p/r2c-CI")
+
+confurl = os.environ.get("SEMGREP_CONFIG_URL", "")
+# ruleid: string-concat
+sh.semgrep("--config {}".format(confurl))
+
+# ruleid: string-concat
+sh.semgrep(f"--config {confurl}")
+
+# ok: string-concat
+args = ["--config", confurl]
+sh.semgrep(*args)