Scanning Container Images With Grype

Scanning Container Images with Grype

Overview

Grype is an open-source vulnerability scanner from Anchore that inspects container images, filesystems, and SBOMs for known CVEs. It leverages Syft-generated SBOMs to match packages against multiple vulnerability databases including NVD, GitHub Advisories, and OS-specific feeds.

Prerequisites

• Docker or Podman installed
• Grype CLI installed (curl -sSfL https://raw.githubusercontent.com/anchore/grype/main/install.sh | sh -s -- -b /usr/local/bin)
• Syft CLI (optional, for SBOM generation)
• Network access to pull vulnerability databases

Core Commands

Install Grype

# Install via script
curl -sSfL https://raw.githubusercontent.com/anchore/grype/main/install.sh | sh -s -- -b /usr/local/bin

# Verify installation
grype version

# Install via Homebrew (macOS/Linux)
brew install grype

Scan Container Images

# Scan a Docker Hub image
grype nginx:latest

# Scan from Docker daemon
grype docker:myapp:1.0

# Scan a local archive
grype docker-archive:image.tar

# Scan an OCI directory
grype oci-dir:path/to/oci/

# Scan a Singularity image
grype sif:image.sif

# Scan a local directory / filesystem
grype dir:/path/to/project

Output Formats

# Default table output
grype alpine:3.18

# JSON output for pipeline processing
grype alpine:3.18 -o json > results.json

# CycloneDX SBOM output
grype alpine:3.18 -o cyclonedx

# SARIF output for GitHub Security tab
grype alpine:3.18 -o sarif > grype.sarif

# Template-based custom output
grype alpine:3.18 -o template -t /path/to/template.tmpl

Filtering and Thresholds

# Fail if vulnerabilities meet or exceed a severity
grype nginx:latest --fail-on critical

# Show only fixed vulnerabilities
grype nginx:latest --only-fixed

# Show only non-fixed vulnerabilities
grype nginx:latest --only-notfixed

# Filter by severity
grype nginx:latest --only-fixed -o json | jq '[.matches[] | select(.vulnerability.severity == "High")]'

# Explain a specific CVE
grype nginx:latest --explain --id CVE-2024-1234

Working with SBOMs

# Generate SBOM with Syft then scan
syft nginx:latest -o spdx-json > nginx-sbom.json
grype sbom:nginx-sbom.json

# Scan CycloneDX SBOM
grype sbom:bom.json

Configuration File (.grype.yaml)

# .grype.yaml
check-for-app-update: false
fail-on-severity: "high"
output: "json"
scope: "squashed"  # or "all-layers"
quiet: false

ignore:
  - vulnerability: CVE-2023-12345
    reason: "False positive - not exploitable in our context"
  - vulnerability: CVE-2023-67890
    fix-state: unknown

db:
  auto-update: true
  cache-dir: "/tmp/grype-db"
  max-allowed-built-age: 120h  # 5 days

match:
  java:
    using-cpes: true
  python:
    using-cpes: true
  javascript:
    using-cpes: false

CI/CD Integration

# GitHub Actions
- name: Scan image with Grype
  uses: anchore/scan-action@v4
  with:
    image: "myregistry/myapp:${{ github.sha }}"
    fail-build: true
    severity-cutoff: high
    output-format: sarif
  id: scan

- name: Upload SARIF
  uses: github/codeql-action/upload-sarif@v3
  with:
    sarif_file: ${{ steps.scan.outputs.sarif }}

# GitLab CI
container_scan:
  stage: test
  image: anchore/grype:latest
  script:
    - grype ${CI_REGISTRY_IMAGE}:${CI_COMMIT_SHA} --fail-on high -o json > grype-report.json
  artifacts:
    reports:
      container_scanning: grype-report.json

Database Management

# Check database status
grype db status

# Manually update vulnerability database
grype db update

# Delete cached database
grype db delete

# List supported database providers
grype db list

Key Vulnerability Sources

Best Practices

1. Pin image tags - Always scan specific digests, not latest
2. Fail on severity - Set --fail-on high or critical in CI gates
3. Use SBOMs - Generate SBOMs with Syft for reproducible scanning
4. Suppress false positives - Use .grype.yaml ignore rules with documented reasons
5. Scan all layers - Use --scope all-layers to catch vulnerabilities in intermediate layers
6. Automate database updates - Keep the vulnerability database current in CI runners
7. Compare scans - Track vulnerability count over time for regression detection