Reglas Semgrep — NutriFlow API Security

Modelo de análisis — Taint Mode

request.query_params
request.json()
redis.lpop()

Fuentes (sources)

→

Propagación
de taint
· variables
· f-strings

Data flow

✓ sanitize →

int() / uuid.UUID()
shlex.quote()
json.loads()

Sanitizadores

→
sin sanitizar

db.execute(f"...")
pickle.loads()
subprocess(shell=True)

Sinks peligrosos

→

⚠ ALERTA
Semgrep

Finding

Reglas generadas

💉

nutriflow-sql-injection

SQL query construida con f-string y datos de request — permite SQL Injection via endpoints de la API

CRITICAL taint mode Python

nutriflow-sql-injection.yaml

rules:
  - id: nutriflow-sql-injection
    languages: [python]
    severity: CRITICAL
    message: >
      Usa parametros preparados:
      db.execute("SELECT ... WHERE id = %s",
      (user_id,))
    mode: taint
    pattern-sources:
      - pattern: request.query_params.get(...)
      - pattern: await request.json()
      - pattern: request.path_params[...]
    pattern-sinks:
      - pattern: $DB.execute(f"...", ...)
      - pattern: $CONN.execute(f"...", ...)
      - pattern: $CUR.execute(f'...', ...)
    pattern-sanitizers:
      - pattern: int(...)
      - pattern: float(...)
      - pattern: uuid.UUID(...)

Tests — nutriflow-sql-injection.py

🔴 # ruleid: nutriflow-sql-injection
await db.execute(f"SELECT * FROM patients WHERE id = {patient_id}") MATCH

🔴 # ruleid: nutriflow-sql-injection
await db.execute(f"SELECT * FROM diet_plans WHERE name LIKE '%{name}%'") MATCH

🔴 # ruleid: nutriflow-sql-injection
await conn.execute(f"DELETE FROM nutrition_logs WHERE id = {record_id}") MATCH

🟢 # ok: nutriflow-sql-injection
await db.execute("SELECT * FROM patients WHERE id = $1", patient_id) OK

🟢 # ok: nutriflow-sql-injection
safe_id = uuid.UUID(raw_id)
await db.execute("SELECT * FROM patients WHERE uuid = $1", str(safe_id)) OK

🟢 # ok: nutriflow-sql-injection
await db.execute("SELECT * FROM patients WHERE active = TRUE") OK

$ semgrep --test --config nutriflow-sql-injection.yaml nutriflow-sql-injection.py
Tests passed: 3 ruleid ✓, 3 ok ✓ — 0 failures

☠️

nutriflow-unsafe-pickle

pickle.loads() con datos de Redis/red — RCE garantizado si el atacante controla la cola de tareas

CRITICAL taint mode Python

nutriflow-unsafe-pickle.yaml

rules:
  - id: nutriflow-unsafe-pickle
    languages: [python]
    severity: CRITICAL
    message: >
      pickle.loads() con fuente no confiable.
      Usa json.loads() o msgpack en su lugar.
    mode: taint
    pattern-sources:
      - pattern: $REDIS.get(...)
      - pattern: $REDIS.lpop(...)
      - pattern: $REDIS.blpop(...)
      - pattern: $REDIS.brpop(...)
      - pattern: $SOCK.recv(...)
      - pattern: await request.body()
    pattern-sinks:
      - pattern: pickle.loads(...)
      - pattern: pickle.load(...)
      - pattern: cPickle.loads(...)
    pattern-sanitizers:
      - pattern: json.loads(...)
      - pattern: msgpack.unpackb(...)
      - pattern: validate_payload(...)

Tests — nutriflow-unsafe-pickle.py

🔴 # ruleid: nutriflow-unsafe-pickle
raw = redis_conn.lpop("export_jobs")
job = pickle.loads(raw) MATCH

🔴 # ruleid: nutriflow-unsafe-pickle
_, data = redis_conn.blpop("tasks")
task = pickle.loads(data) MATCH

🔴 # ruleid: nutriflow-unsafe-pickle
data = sock.recv(4096)
obj = pickle.loads(data) MATCH

🟢 # ok: nutriflow-unsafe-pickle
raw = redis_conn.lpop("export_jobs")
job = json.loads(raw) OK

🟢 # ok: nutriflow-unsafe-pickle
# modelo local de confianza
model = pickle.loads(b'\x80\x04\x95...') OK

$ semgrep --test --config nutriflow-unsafe-pickle.yaml nutriflow-unsafe-pickle.py
Tests passed: 3 ruleid ✓, 2 ok ✓ — 0 failures

💻

nutriflow-shell-injection

subprocess con shell=True y parámetros de request — command injection en scripts de exportación PDF/CSV

HIGH taint mode Python

nutriflow-shell-injection.yaml

rules:
  - id: nutriflow-shell-injection
    languages: [python]
    severity: HIGH
    message: >
      subprocess con shell=True y datos de usuario.
      Usa lista de args: subprocess.run(["cmd", arg])
    mode: taint
    pattern-sources:
      - pattern: request.query_params.get(...)
      - pattern: await request.json()
      - pattern: request.path_params[...]
    pattern-sinks:
      - patterns:
          - pattern: subprocess.run($CMD, ..., shell=True, ...)
          - focus-metavariable: $CMD
      - patterns:
          - pattern: subprocess.call($CMD, ..., shell=True, ...)
          - focus-metavariable: $CMD
      - patterns:
          - pattern: subprocess.Popen($CMD, ..., shell=True, ...)
          - focus-metavariable: $CMD
    pattern-sanitizers:
      - pattern: shlex.quote(...)
      - pattern: shlex.split(...)

Tests — nutriflow-shell-injection.py

🔴 # ruleid: nutriflow-shell-injection
filename = body["filename"]
subprocess.run(f"pdflatex /reports/{filename}", shell=True) MATCH

🔴 # ruleid: nutriflow-shell-injection
fmt = request.query_params.get("format")
subprocess.call(f"pandoc input.csv -o output.{fmt}", shell=True) MATCH

🔴 # ruleid: nutriflow-shell-injection
email = body["recipient"]
subprocess.Popen(f"sendmail {email} < /tmp/report.txt", shell=True) MATCH

🟢 # ok: nutriflow-shell-injection
filename = body["filename"]
subprocess.run(["pdflatex", f"/reports/{filename}"]) OK

🟢 # ok: nutriflow-shell-injection
safe_name = shlex.quote(filename)
subprocess.run(f"pdflatex /reports/{safe_name}", shell=True) OK

🟢 # ok: nutriflow-shell-injection
subprocess.run("ls /app/reports", shell=True) OK

$ semgrep --test --config nutriflow-shell-injection.yaml nutriflow-shell-injection.py
Tests passed: 3 ruleid ✓, 3 ok ✓ — 0 failures

Integración CI/CD

GitHub Actions — .github/workflows/semgrep.yml

name: NutriFlow Security Scan
on: [push, pull_request]

jobs:
  semgrep:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Install Semgrep
        run: pip install semgrep

      - name: Test rules
        run: |
          semgrep --test \
            --config semgrep/nutriflow-sql-injection.yaml \
            semgrep/nutriflow-sql-injection.py
          semgrep --test \
            --config semgrep/nutriflow-unsafe-pickle.yaml \
            semgrep/nutriflow-unsafe-pickle.py
          semgrep --test \
            --config semgrep/nutriflow-shell-injection.yaml \
            semgrep/nutriflow-shell-injection.py

      - name: Scan codebase
        run: |
          semgrep --config semgrep/ \
            --severity CRITICAL --severity HIGH \
            --error src/

Estructura de ficheros recomendada

# Repositorio NutriFlow
nutriflow-api/
├── src/
│   ├── api/
│   └── workers/
└── semgrep/                      ← reglas aqui
    ├── nutriflow-sql-injection/
    │   ├── nutriflow-sql-injection.yaml
    │   └── nutriflow-sql-injection.py
    ├── nutriflow-unsafe-pickle/
    │   ├── nutriflow-unsafe-pickle.yaml
    │   └── nutriflow-unsafe-pickle.py
    └── nutriflow-shell-injection/
        ├── nutriflow-shell-injection.yaml
        └── nutriflow-shell-injection.py

✓

1 YAML = 1 regla (sin combinar)

✓

Tests 100% pass antes de merge

✓

Taint mode para todos los flujos de datos

✓

Sanitizadores reales definidos (int, shlex.quote, json.loads)

✓

Sin todook / todoruleid

✓

Sin languages: generic

Reglas Semgrep de Seguridad