scanning-containers-with-trivy-in-cicd（CICD 中使用 Trivy 扫描容器）

Scanning Containers with Trivy in CI/CD

When to Use

• When building Docker container images in CI/CD and needing automated vulnerability scanning before registry push
• When establishing quality gates that prevent images with critical or high CVEs from reaching production
• When compliance requirements mandate vulnerability scanning of all container images before deployment
• When scanning IaC files (Dockerfiles, Kubernetes manifests) alongside container image scanning
• When needing a single tool to scan OS packages, language-specific dependencies, and misconfigurations

Do not use for runtime container security monitoring (use Falco), for scanning running containers in production (use runtime agents), or when only scanning application source code without containerization (use SAST tools).

Prerequisites

• Trivy CLI installed (v0.50+) or access to aquasecurity/trivy-action GitHub Action
• Docker daemon available in CI/CD for building and scanning images
• Container registry credentials for pulling base images and pushing scanned images
• Trivy vulnerability database accessible (downloaded automatically or cached)

Workflow

Step 1: Configure Trivy Scanning in GitHub Actions

Set up a GitHub Actions workflow that builds a Docker image and scans it with Trivy before pushing to a container registry.

# .github/workflows/container-security.yml
name: Container Security Scan

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]
    paths:
      - 'Dockerfile'
      - 'docker-compose*.yml'
      - 'src/**'
      - 'requirements*.txt'
      - 'package*.json'

jobs:
  build-and-scan:
    runs-on: ubuntu-latest
    permissions:
      security-events: write
      contents: read

    steps:
      - uses: actions/checkout@v4

      - name: Build Docker image
        run: docker build -t app:${{ github.sha }} .

      - name: Run Trivy vulnerability scanner
        uses: aquasecurity/trivy-action@0.28.0
        with:
          image-ref: 'app:${{ github.sha }}'
          format: 'sarif'
          output: 'trivy-results.sarif'
          severity: 'CRITICAL,HIGH'
          exit-code: '1'
          ignore-unfixed: true

      - name: Upload Trivy scan results
        uses: github/codeql-action/upload-sarif@v3
        if: always()
        with:
          sarif_file: 'trivy-results.sarif'
          category: 'trivy-container'

      - name: Run Trivy misconfiguration scanner
        uses: aquasecurity/trivy-action@0.28.0
        with:
          scan-type: 'config'
          scan-ref: '.'
          format: 'table'
          exit-code: '1'
          severity: 'CRITICAL,HIGH'

Step 2: Scan Dockerfiles for Misconfigurations

Trivy detects common Dockerfile security issues such as running as root, using latest tags, and exposing unnecessary ports.

# Scan Dockerfile for misconfigurations
trivy config --severity HIGH,CRITICAL ./Dockerfile

# Scan with custom policy directory
trivy config --policy ./security-policies --severity MEDIUM,HIGH,CRITICAL .

# Example secure Dockerfile practices Trivy checks for:
# - USER instruction present (not running as root)
# - HEALTHCHECK instruction defined
# - Base image uses specific tag, not :latest
# - No secrets in ENV or ARG instructions
# - COPY preferred over ADD

Step 3: Integrate with GitLab CI/CD

# .gitlab-ci.yml
stages:
  - build
  - scan
  - push

variables:
  TRIVY_CACHE_DIR: .trivycache/

build:
  stage: build
  script:
    - docker build -t $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA .
    - docker save $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA -o image.tar
  artifacts:
    paths:
      - image.tar

trivy-scan:
  stage: scan
  image:
    name: aquasec/trivy:latest
    entrypoint: [""]
  cache:
    paths:
      - .trivycache/
  script:
    - trivy image
        --input image.tar
        --exit-code 1
        --severity CRITICAL,HIGH
        --ignore-unfixed
        --format json
        --output trivy-report.json
    - trivy image
        --input image.tar
        --severity CRITICAL,HIGH,MEDIUM
        --format table
  artifacts:
    reports:
      container_scanning: trivy-report.json
    paths:
      - trivy-report.json
  allow_failure: false

push:
  stage: push
  needs: [trivy-scan]
  script:
    - docker load -i image.tar
    - docker push $CI_REGISTRY_IMAGE:$CI_COMMIT_SHA

Step 4: Configure Trivy Ignore and Exception Handling

Manage false positives and accepted risks through Trivy's ignore file and VEX statements.

# .trivyignore.yaml
vulnerabilities:
  - id: CVE-2023-44487    # HTTP/2 rapid reset - mitigated at load balancer
    statement: "Mitigated by WAF rate limiting at ingress layer"
    expires: 2026-06-01

  - id: CVE-2024-21626    # runc container escape - patched in base image update
    statement: "Tracked in JIRA-SEC-1234, base image update scheduled"
    expires: 2026-03-15

misconfigurations:
  - id: DS002             # User not set - required for init containers
    paths:
      - "docker/init-container/Dockerfile"
    statement: "Init container requires root for volume permission setup"

Step 5: Implement Database Caching and Offline Scanning

Cache the Trivy vulnerability database in CI/CD to reduce scan times and enable air-gapped environments.

# GitHub Actions with database caching
- name: Cache Trivy DB
  uses: actions/cache@v4
  with:
    path: /tmp/trivy-db
    key: trivy-db-${{ hashFiles('.github/workflows/container-security.yml') }}
    restore-keys: trivy-db-

- name: Run Trivy with cached DB
  uses: aquasecurity/trivy-action@0.28.0
  with:
    image-ref: 'app:${{ github.sha }}'
    cache-dir: /tmp/trivy-db
    format: 'json'
    output: 'trivy-results.json'
    severity: 'CRITICAL,HIGH'
    exit-code: '1'

# Air-gapped: Download DB manually and mount
trivy image --download-db-only --cache-dir /path/to/cache
# Transfer cache to air-gapped system
trivy image --skip-db-update --cache-dir /path/to/cache myimage:tag

Step 6: Generate SBOM and Scan for License Compliance

Use Trivy to generate Software Bill of Materials alongside vulnerability scanning.

# Generate SBOM in CycloneDX format
trivy image --format cyclonedx --output sbom.cdx.json app:latest

# Generate SBOM in SPDX format
trivy image --format spdx-json --output sbom.spdx.json app:latest

# Scan SBOM for vulnerabilities (decouple generation from scanning)
trivy sbom sbom.cdx.json --severity CRITICAL,HIGH

# Scan with license detection
trivy image --scanners vuln,license --severity HIGH,CRITICAL app:latest

Key Concepts

Tools & Systems

• Trivy: Open-source vulnerability scanner by Aqua Security supporting images, filesystems, repos, and IaC
• trivy-action: Official GitHub Action for running Trivy scans in GitHub Actions workflows
• Trivy Operator: Kubernetes operator that continuously scans cluster workloads with Trivy
• Grype: Alternative image scanner by Anchore for comparison and validation of scan results
• Harbor: Container registry with built-in Trivy integration for automatic image scanning on push

Common Scenarios

Scenario: Multi-Stage Build with Separate Scan and Push

Context: A team builds multi-stage Docker images and needs to scan the final production image before pushing to ECR, while also scanning the build stage for supply chain risks.

Approach:

1. Build the Docker image with --target production for the final stage
2. Run Trivy with --severity CRITICAL,HIGH --exit-code 1 --ignore-unfixed to block on exploitable issues
3. Generate an SBOM in CycloneDX format and store as a build artifact
4. Upload SARIF results to GitHub Security tab for visibility
5. Only push to ECR if the Trivy scan exits with code 0
6. Tag the pushed image with the scan timestamp and Trivy DB version for audit traceability

Pitfalls: Scanning only the final stage misses vulnerable packages that were present in build stages and may have influenced the build. Run trivy fs on the build context separately. Caching the Trivy DB too aggressively (weekly) means newly published CVEs take days to appear in scans.

Output Format

Trivy Container Scan Report
=============================
Image: app:a1b2c3d4
Base Image: python:3.12-slim-bookworm
Scan Date: 2026-02-23
DB Version: 2026-02-23T00:15:00Z

VULNERABILITY SUMMARY:
  Total: 47
  Critical: 2
  High: 5
  Medium: 18
  Low: 22
  Unfixed: 8 (excluded from gate)

CRITICAL FINDINGS:
  CVE-2025-12345  libssl3    3.0.11-1  3.0.13-1  OpenSSL buffer overflow
  CVE-2025-67890  curl       7.88.1-10 7.88.1-12 curl HSTS bypass

HIGH FINDINGS:
  CVE-2025-11111  zlib1g     1.2.13    1.2.13.1  zlib heap buffer overflow
  CVE-2025-22222  python3.12 3.12.1    3.12.3    CPython path traversal
  CVE-2025-33333  requests   2.31.0    2.32.0    requests SSRF in redirects

MISCONFIGURATION:
  DS002  [HIGH]   Dockerfile: USER instruction not set (running as root)
  DS026  [MEDIUM] Dockerfile: No HEALTHCHECK defined

QUALITY GATE: FAILED (2 Critical, 5 High findings)