container-scanning – skill Bảo mật cho agent AI

Container Scanning

Scan container images for vulnerabilities and security misconfigurations.

When to Use This Skill

Use this skill when:

• Building container images
• Implementing container security gates
• Scanning registry images
• Meeting compliance requirements
• Hardening container deployments

Prerequisites

• Container runtime (Docker, Podman)
• Container images to scan
• Scanning tool installation

Tool Comparison

| Tool | License | Speed | Features |

|------|---------|-------|----------|

| Trivy | OSS | Fast | Comprehensive, IaC |

| Grype | OSS | Fast | Accurate, SBOM |

| Clair | OSS | Medium | Registry integration |

| Snyk Container | Commercial | Fast | Fix suggestions |

| Docker Scout | Commercial | Fast | GitHub integration |

Trivy

Installation

# Linux
# Tool prerequisite: use a verified preinstalled binary; no network installer runs here.

# macOS
brew install trivy

# Docker
docker pull aquasec/trivy

Image Scanning

# Scan local image
trivy image myapp:latest

# Scan remote image
trivy image nginx:1.25

# JSON output
trivy image --format json -o results.json myapp:latest

# Filter by severity
trivy image --severity HIGH,CRITICAL myapp:latest

# Optional filtered view; retain an unfiltered report of unfixed risk too.
trivy image --ignore-unfixed myapp:latest

# Exit code on vulnerability
trivy image --exit-code 1 --severity CRITICAL myapp:latest

Filesystem Scanning

# Scan project directory
trivy fs /path/to/project

# Scan Dockerfile
trivy config Dockerfile

# Scan Kubernetes manifests
trivy config k8s/

Configuration

# trivy.yaml
timeout: 10m
severity:
  - HIGH
  - CRITICAL
ignore-unfixed: true
exit-code: 1

vulnerability:
  type:
    - os
    - library

scan:
  file-patterns:
    - "Dockerfile"
    - "*.yaml"

CI Integration

NON-EXECUTABLE CI DESIGN: before adoption, use least-privilege jobs, pin every action to a reviewed full SHA, and provision verified fixed-version tools in a digest-pinned runner image. Never expose credentials during tool provisioning.
# GitHub Actions
name: Container Security

on:
  push:
    branches: [main]
  pull_request:

jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

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

      - name: Run Trivy
        step: Run the preinstalled reviewed Trivy binary
        with:
          image-ref: 'myapp:${{ github.sha }}'
          format: 'sarif'
          output: 'trivy-results.sarif'
          severity: 'CRITICAL,HIGH'
          exit-code: '1'

      - name: Upload results
        uses: github/codeql-action/upload-sarif@v3
        with:
          sarif_file: 'trivy-results.sarif'

Grype

Installation

# Linux/macOS
# Tool prerequisite: use a verified preinstalled binary; no network installer runs here.

# Homebrew
brew install grype

Usage

# Scan image
grype myapp:latest

# Scan from SBOM
grype sbom:./sbom.json

# JSON output
grype myapp:latest -o json > results.json

# Filter severity
grype myapp:latest --fail-on high

# Scan directory
grype dir:/path/to/project

Configuration

# .grype.yaml
check-for-app-update: false
fail-on-severity: high
output: "json"
scope: "Squashed"

ignore:
  - vulnerability: CVE-2023-12345
    reason: "False positive"
    expires: "2024-12-31"

Docker Scout

Usage

# Enable Docker Scout
docker scout quickview myapp:latest

# Full CVE report
docker scout cves myapp:latest

# Compare images
docker scout compare myapp:v1 myapp:v2

# Recommendations
docker scout recommendations myapp:latest

CI Integration

NON-EXECUTABLE CI DESIGN: before adoption, use least-privilege jobs, pin every action to a reviewed full SHA, and provision verified fixed-version tools in a digest-pinned runner image. Never expose credentials during tool provisioning.
- name: Docker Scout
  uses: docker/scout-action@v1
  with:
    command: cves
    image: ${{ env.IMAGE_NAME }}
    sarif-file: scout-results.sarif
    summary: true

Registry Integration

Amazon ECR

# Enable scan on push
aws ecr put-image-scanning-configuration \
  --repository-name myapp \
  --image-scanning-configuration scanOnPush=true

# Get scan findings
aws ecr describe-image-scan-findings \
  --repository-name myapp \
  --image-id imageTag=latest

# Start manual scan
aws ecr start-image-scan \
  --repository-name myapp \
  --image-id imageTag=latest

Azure ACR

# Enable Defender for Containers
az security pricing create \
  --name Containers \
  --tier Standard

# Read Defender vulnerability findings through its supported findings API/portal.
# az acr repository show returns image metadata, not a vulnerability report.

Google Artifact Registry

# Enable vulnerability scanning
gcloud artifacts repositories update myrepo \
  --location=us-central1 \
  --enable-vulnerability-scanning

# View vulnerabilities
gcloud artifacts docker images describe \
  us-central1-docker.pkg.dev/project/myrepo/myapp:latest \
  --show-package-vulnerability

Admission example boundaries

The Gatekeeper/Kyverno snippets are incomplete/version-specific sketches, not enforceable proof of scan policy. Pin actual CRDs/Rego syntax, supply required Constraint/trusted attestor configuration, and test accept/reject cases before cluster admission. Repository prefix checks require a normalized boundary (for example an approved registry/path plus /), not an attacker-lookalike prefix; cover init and ephemeral containers. A verifyImages predicate needs a verified signer/issuer and actual predicate schema, not an invented scanner variable. The scan-policy.yaml below is conceptual, with no bundled consumer. Record image digest, scanner/DB versions, date, exclusions and residual unfixed findings; a clean scan is not certification.

Admission Controllers

OPA Gatekeeper

apiVersion: templates.gatekeeper.sh/v1beta1
kind: ConstraintTemplate
metadata:
  name: k8sallowedrepos
spec:
  crd:
    spec:
      names:
        kind: K8sAllowedRepos
      validation:
        openAPIV3Schema:
          properties:
            repos:
              type: array
              items:
                type: string
  targets:
    - target: admission.k8s.gatekeeper.sh
      rego: |
        package k8sallowedrepos
        
        violation[{"msg": msg}] {
          container := input.review.object.spec.containers[_]
          satisfied := [good | repo = input.parameters.repos[_]; good = startswith(container.image, repo)]
          not any(satisfied)
          msg := sprintf("container <%v> has an invalid image repo <%v>", [container.name, container.image])
        }

Kyverno

apiVersion: kyverno.io/v1
kind: ClusterPolicy
metadata:
  name: require-image-scan
spec:
  validationFailureAction: enforce
  rules:
    - name: check-vulnerabilities
      match:
        resources:
          kinds:
            - Pod
      verifyImages:
        - image: "*"
          attestations:
            - predicateType: cosign.sigstore.dev/attestation/vuln/v1
              conditions:
                - all:
                    - key: "{{ scanner.result.summary.criticalCount }}"
                      operator: Equals
                      value: "0"

Scanning Policies

Policy Definition

# scan-policy.yaml
policies:
  - name: critical-vulnerabilities
    description: Block images with critical CVEs
    severity: CRITICAL
    action: block
    
  - name: high-vulnerabilities
    description: Warn on high severity CVEs
    severity: HIGH
    action: warn
    max_count: 5
    
  - name: age-policy
    description: Block images older than 30 days
    max_age_days: 30
    action: block
    
  - name: base-image
    description: Only allow approved base images
    allowed_bases:
      - alpine:3.18
      - ubuntu:22.04
      - python:3.11-slim

Common Issues

Issue: False Positives

Problem: Scanner reports non-exploitable vulnerabilities

Solution: Use ignore files, validate with context

Issue: Slow Scans

Problem: Scanning takes too long

Solution: Use caching, scan incrementally, optimize image layers

Issue: Unfixed Vulnerabilities

Problem: No patch available for CVE

Solution: Update base image, implement compensating controls

Best Practices

• Scan in CI/CD pipeline
• Use minimal base images (Alpine, distroless)
• Update base images regularly
• Implement admission control
• Track vulnerabilities over time
• Set severity thresholds
• Document accepted risks
• Use multi-stage builds

Related Skills

• docker-management (docker-management) - Container basics
• container-hardening (container-hardening) - Security hardening
• kubernetes-hardening (kubernetes-hardening) - K8s security

Limitations

• Apply guidance only within authorized scope; test de

… (nội dung đầy đủ có trong gói tải về)