feat: add dedicated Security page and policy-pack foundation (#1362)

* feat: add dedicated Security page and policy-pack foundation

Bring vulnerability scanning, scan history, suppressions, Compose risks,
secrets, policy packs, and scanner setup into one node-scoped Security
command center instead of scattering them across Resources and Settings.

- New top-level Security view with Overview, Images, Compose risks,
  Secrets, Policies, Suppressions, History, and Scanner setup tabs
  (status masthead + signal rail; controlled tabs with deep-link support).
- Backend: GET /security/overview rollup and GET /security/policy-packs
  static catalog (auth-only, Community). DatabaseService gains an uncapped
  scan-status count and a node-eligible block-policy count, and
  getImageScanSummaries now projects secret and misconfig counts.
- Reuse existing surfaces: the scan-history sheet, the control-governed
  suppression and acknowledgement panels, and the scan-detail sheet (now
  with an initial-tab prop so it opens on the matching finding type).
- Extract a shared SeverityBadge (from Resources) and a TrivyManager
  (from Settings) so both surfaces render identical controls.
- Resources "Scan history" now links into the Security page History tab.
- Docs for the new Security surface and tests for the new endpoints,
  helpers, nav wiring, and tabs.

* refactor: consolidate scanner and policy management onto the Security page

Remove the Settings "Vulnerability Scanning" section now that the Security
page covers the same ground, with every option preserved:

- Scanner install / update / uninstall / auto-update live on the Scanner setup
  tab (TrivyManager).
- Scan policies, the honor-suppressions toggle, and the replica
  managed-by-control / demote controls move into a new ScanPolicyManager on the
  Policies tab (paid; Community sees only the policy-pack catalog).
- CVE suppressions and acknowledgements remain on the Suppressions tab.

Wiring removed: the registry section and the now-empty Security settings group,
the SectionId, the SettingsSectionContent case and the isPaid prop it was the
sole consumer of, and SecuritySection itself. The dashboard configuration-status
"Vulnerability scanning" row now navigates to the Security page Policies tab.

Docs that pointed at "Settings -> Security -> Vulnerability Scanning" are swept
to the relevant Security page tabs.

* fix: harden Security page scanner refresh, policy-load errors, and secret-only badges

Address independent-review findings on the Security page:

- Scanner setup now refreshes Trivy state when the active node changes, so the
  displayed scanner status matches the node TrivyManager's actions target (both
  follow x-node-id). Previously, switching nodes on the tab left stale state.
- ScanPolicyManager surfaces an explicit error state on a failed policy fetch
  instead of falling through to a false "No scan policies configured".
- The shared SeverityBadge and the Images findings column no longer label a scan
  "clean" when it has secrets or misconfigurations but no CVE severity
  (highest_severity is derived from vulnerabilities only); they show a "Findings"
  state and the secret/misconfig counts instead.
- The Overview enforcement note points to the Policies tab, not the removed
  Settings section.
- The History tab auto-opens the scan-history sheet only on a deep-link (mount
  with the History tab active), not on every manual tab selection.

Adds tests for the badge secret/misconfig state and the policy-load error state.
This commit is contained in:
Anso
2026-06-12 10:41:39 -04:00
committed by GitHub
parent 77f1611971
commit 2a4955f56d
51 changed files with 2559 additions and 509 deletions
@@ -0,0 +1,118 @@
/**
* Unit coverage for the two DatabaseService helpers added for the Security
* overview:
* - countScansByStatus: an UNCAPPED count (getVulnerabilityScans applies a
* per-image history cap that would undercount failed scans).
* - countEligibleBlockPolicies: counts enabled block-on-deploy policies that
* apply to a node (fleet-wide or this node), built on getScanPoliciesForUi
* so a replica never counts a sibling-identity policy.
*/
import { describe, it, expect, beforeAll, afterAll, beforeEach } from 'vitest';
import { setupTestDb, cleanupTestDb } from './helpers/setupTestDb';
import type { ScanPolicy } from '../services/DatabaseService';
let tmpDir: string;
let DatabaseService: typeof import('../services/DatabaseService').DatabaseService;
beforeAll(async () => {
tmpDir = await setupTestDb();
({ DatabaseService } = await import('../services/DatabaseService'));
});
afterAll(() => cleanupTestDb(tmpDir));
function db() {
return DatabaseService.getInstance();
}
function reset(): void {
const raw = (db() as unknown as { db: { prepare: (s: string) => { run: () => void } } }).db;
raw.prepare('DELETE FROM vulnerability_scans').run();
raw.prepare('DELETE FROM scan_policies').run();
}
function seedFailed(imageRef: string): void {
db().createVulnerabilityScan({
node_id: 1,
image_ref: imageRef,
image_digest: `sha256:${imageRef}-${Math.random().toString(16).slice(2)}`,
scanned_at: 1,
total_vulnerabilities: 0,
critical_count: 0,
high_count: 0,
medium_count: 0,
low_count: 0,
unknown_count: 0,
fixable_count: 0,
secret_count: 0,
misconfig_count: 0,
scanners_used: 'vuln',
highest_severity: null,
os_info: null,
trivy_version: null,
scan_duration_ms: null,
triggered_by: 'manual',
status: 'failed',
error: 'boom',
stack_context: null,
});
}
function seedPolicy(overrides: Partial<Omit<ScanPolicy, 'id' | 'created_at' | 'updated_at'>>): void {
db().createScanPolicy({
name: overrides.name ?? 'p',
node_id: overrides.node_id ?? null,
node_identity: overrides.node_identity ?? '',
stack_pattern: overrides.stack_pattern ?? null,
max_severity: overrides.max_severity ?? 'CRITICAL',
block_on_deploy: overrides.block_on_deploy ?? 1,
enabled: overrides.enabled ?? 1,
replicated_from_control: overrides.replicated_from_control ?? 0,
});
}
beforeEach(() => reset());
describe('countScansByStatus', () => {
it('counts failed scans uncapped, even beyond the per-image history cap', () => {
// The grouped history view caps rows per image_ref (default 50). All 55 of
// these are the same image, so a capped path would undercount.
for (let i = 0; i < 55; i++) seedFailed('same-image:1');
expect(db().countScansByStatus(1, 'failed')).toBe(55);
});
it('is node-scoped', () => {
seedFailed('a:1');
db().createVulnerabilityScan({
node_id: 2, image_ref: 'b:1', image_digest: 'sha256:b', scanned_at: 1,
total_vulnerabilities: 0, critical_count: 0, high_count: 0, medium_count: 0, low_count: 0,
unknown_count: 0, fixable_count: 0, secret_count: 0, misconfig_count: 0, scanners_used: 'vuln',
highest_severity: null, os_info: null, trivy_version: null, scan_duration_ms: null,
triggered_by: 'manual', status: 'failed', error: 'x', stack_context: null,
});
expect(db().countScansByStatus(1, 'failed')).toBe(1);
});
});
describe('countEligibleBlockPolicies (control)', () => {
it('counts fleet-wide and this-node block policies, excludes other nodes / disabled / non-blocking', () => {
seedPolicy({ name: 'fleet-wide', node_id: null }); // counted
seedPolicy({ name: 'this-node', node_id: 1 }); // counted
seedPolicy({ name: 'other-node', node_id: 2 }); // excluded (different node)
seedPolicy({ name: 'disabled', node_id: 1, enabled: 0 }); // excluded (disabled)
seedPolicy({ name: 'no-block', node_id: 1, block_on_deploy: 0 }); // excluded (not blocking)
expect(db().countEligibleBlockPolicies(1, 'control', '')).toBe(2);
});
});
describe('countEligibleBlockPolicies (replica)', () => {
it('filters a replicated policy scoped to a sibling identity, keeps fleet-wide', () => {
// Fleet-wide replicated row (empty identity) applies on every replica.
seedPolicy({ name: 'fleet-wide', node_id: null, replicated_from_control: 1, node_identity: '' });
// Sibling-scoped replicated row must not be counted on this replica.
seedPolicy({ name: 'sibling', node_id: null, replicated_from_control: 1, node_identity: 'sibling-id' });
expect(db().countEligibleBlockPolicies(1, 'replica', 'self-id')).toBe(1);
});
});
@@ -0,0 +1,184 @@
/**
* GET /api/security/overview -> node-scoped posture rollup (Community, auth-only)
* GET /api/security/policy-packs -> static catalog (Community, auth-only, identical per tier)
*/
import { describe, it, expect, beforeAll, afterAll, beforeEach, vi } from 'vitest';
import request from 'supertest';
import bcrypt from 'bcrypt';
import { setupTestDb, cleanupTestDb, loginAsTestAdmin } from './helpers/setupTestDb';
let tmpDir: string;
let app: import('express').Express;
let adminCookie: string;
let viewerCookie: string;
let DatabaseService: typeof import('../services/DatabaseService').DatabaseService;
let LicenseService: typeof import('../services/LicenseService').LicenseService;
let TrivyService: typeof import('../services/TrivyService').default;
const DAY = 24 * 60 * 60 * 1000;
beforeAll(async () => {
tmpDir = await setupTestDb();
({ DatabaseService } = await import('../services/DatabaseService'));
({ LicenseService } = await import('../services/LicenseService'));
TrivyService = (await import('../services/TrivyService')).default;
vi.spyOn(LicenseService.getInstance(), 'getTier').mockReturnValue('community');
// Deterministic scanner readout for the overview's scanner block.
const svc = TrivyService.getInstance();
vi.spyOn(svc, 'isTrivyAvailable').mockReturnValue(true);
vi.spyOn(svc, 'getVersion').mockReturnValue('0.52.0');
vi.spyOn(svc, 'getSource').mockReturnValue('managed');
({ app } = await import('../index'));
adminCookie = await loginAsTestAdmin(app);
const viewerHash = await bcrypt.hash('ovviewer1', 1);
DatabaseService.getInstance().addUser({ username: 'ov-viewer', password_hash: viewerHash, role: 'viewer' });
const res = await request(app).post('/api/auth/login').send({ username: 'ov-viewer', password: 'ovviewer1' });
const cookies = res.headers['set-cookie'] as string | string[];
viewerCookie = Array.isArray(cookies) ? cookies[0] : cookies;
});
afterAll(() => cleanupTestDb(tmpDir));
function db() {
return DatabaseService.getInstance();
}
function seedScan(o: {
node_id?: number;
image_ref: string;
scanned_at: number;
status?: 'completed' | 'failed';
critical?: number;
high?: number;
fixable?: number;
secret?: number;
misconfig?: number;
}): void {
db().createVulnerabilityScan({
node_id: o.node_id ?? 1,
image_ref: o.image_ref,
image_digest: `sha256:${o.image_ref}-${Math.random().toString(16).slice(2)}`,
scanned_at: o.scanned_at,
total_vulnerabilities: (o.critical ?? 0) + (o.high ?? 0),
critical_count: o.critical ?? 0,
high_count: o.high ?? 0,
medium_count: 0,
low_count: 0,
unknown_count: 0,
fixable_count: o.fixable ?? 0,
secret_count: o.secret ?? 0,
misconfig_count: o.misconfig ?? 0,
scanners_used: 'vuln',
highest_severity: (o.critical ?? 0) > 0 ? 'CRITICAL' : null,
os_info: null,
trivy_version: null,
scan_duration_ms: null,
triggered_by: 'manual',
status: o.status ?? 'completed',
error: o.status === 'failed' ? 'boom' : null,
stack_context: o.image_ref.startsWith('stack:') ? o.image_ref.slice(6) : null,
});
}
function resetSecurity(): void {
const raw = (db() as unknown as { db: { prepare: (s: string) => { run: () => void } } }).db;
raw.prepare('DELETE FROM vulnerability_scans').run();
raw.prepare('DELETE FROM scan_policies').run();
}
describe('GET /api/security/overview', () => {
beforeEach(() => {
resetSecurity();
db().updateGlobalSetting('deploy_block_honor_suppressions', '1');
});
it('aggregates node-scoped counts with the documented shape', async () => {
const now = Date.now();
seedScan({ image_ref: 'imgA:1', scanned_at: now - 1000, critical: 2, high: 1, fixable: 3, secret: 1 });
seedScan({ image_ref: 'imgB:1', scanned_at: now - 8 * DAY }); // stale
seedScan({ image_ref: 'stack:web', scanned_at: now - 2000, misconfig: 2 });
// Failed scans (same image) beyond a single row prove the uncapped count.
for (let i = 0; i < 4; i++) seedScan({ image_ref: 'imgA:1', scanned_at: now, status: 'failed' });
// Other node's data must be excluded.
seedScan({ node_id: 2, image_ref: 'other:1', scanned_at: now, critical: 99 });
// One fleet-wide and one this-node block policy count; an other-node one does not.
db().createScanPolicy({ name: 'fw', node_id: null, node_identity: '', stack_pattern: null, max_severity: 'CRITICAL', block_on_deploy: 1, enabled: 1, replicated_from_control: 0 });
db().createScanPolicy({ name: 'n1', node_id: 1, node_identity: '', stack_pattern: null, max_severity: 'CRITICAL', block_on_deploy: 1, enabled: 1, replicated_from_control: 0 });
db().createScanPolicy({ name: 'n2', node_id: 2, node_identity: '', stack_pattern: null, max_severity: 'CRITICAL', block_on_deploy: 1, enabled: 1, replicated_from_control: 0 });
const res = await request(app).get('/api/security/overview').set('Cookie', adminCookie);
expect(res.status).toBe(200);
expect(res.body).toMatchObject({
scannedImages: 2, // imgA + imgB, stack:web excluded
critical: 2, // node-2's 99 excluded
high: 1,
fixable: 3,
secrets: 1,
misconfigs: 2,
staleScans: 1, // imgB only
failedScans: 4, // uncapped
});
expect(res.body.lastSuccessfulScanAt).toBeGreaterThan(0);
expect(res.body.scanner).toMatchObject({ available: true, source: 'managed', version: '0.52.0' });
expect(res.body.deployEnforcement).toMatchObject({
honorSuppressionsOnDeploy: true,
eligibleBlockPolicies: 2,
});
});
it('is reachable by a Community viewer (read-only, auth-only)', async () => {
const res = await request(app).get('/api/security/overview').set('Cookie', viewerCookie);
expect(res.status).toBe(200);
});
it('requires authentication', async () => {
const res = await request(app).get('/api/security/overview');
expect(res.status).toBe(401);
});
});
describe('GET /api/security/policy-packs', () => {
it('returns the 5 default packs with fully-formed rules (auth-only)', async () => {
const res = await request(app).get('/api/security/policy-packs').set('Cookie', adminCookie);
expect(res.status).toBe(200);
expect(Array.isArray(res.body)).toBe(true);
expect(res.body).toHaveLength(5);
for (const pack of res.body) {
expect(pack).toMatchObject({
id: expect.any(String),
name: expect.any(String),
tagline: expect.any(String),
tierCopy: expect.any(String),
});
expect(Array.isArray(pack.rules)).toBe(true);
expect(pack.rules.length).toBeGreaterThan(0);
for (const rule of pack.rules) {
expect(rule).toMatchObject({
id: expect.any(String),
name: expect.any(String),
severity: expect.stringMatching(/^(CRITICAL|HIGH|MEDIUM|LOW)$/),
whatItChecks: expect.any(String),
why: expect.any(String),
howToFix: expect.any(String),
enforcement: expect.stringMatching(/^(warning|enforceable)$/),
});
}
}
});
it('returns 401 unauthenticated', async () => {
const res = await request(app).get('/api/security/policy-packs');
expect(res.status).toBe(401);
});
it('returns an identical catalog regardless of tier', async () => {
vi.spyOn(LicenseService.getInstance(), 'getTier').mockReturnValue('community');
const community = await request(app).get('/api/security/policy-packs').set('Cookie', adminCookie);
vi.spyOn(LicenseService.getInstance(), 'getTier').mockReturnValue('paid');
const paid = await request(app).get('/api/security/policy-packs').set('Cookie', adminCookie);
vi.spyOn(LicenseService.getInstance(), 'getTier').mockReturnValue('community');
expect(paid.body).toEqual(community.body);
});
});
+100
View File
@@ -17,6 +17,7 @@ import { isDebugEnabled } from '../utils/debug';
import { blockIfReplica } from '../middleware/fleetSyncGuards';
import { validateStackPatternForRedos } from './fleet';
import { FINDING_SEVERITIES, POLICY_SEVERITIES } from '../utils/severity';
import { DEFAULT_POLICY_PACKS } from '../services/policy-packs';
const CVE_ID_RE = /^(CVE-\d{4}-\d{4,}|GHSA-[\w-]{14,})$/;
// Trivy emits misconfig rule ids in two shapes that Sencho persists verbatim:
@@ -113,6 +114,28 @@ function shapeScanForResponse(scan: VulnerabilityScan): Omit<VulnerabilityScan,
return { ...rest, policy_evaluation: parsePolicyEvaluation(policy_evaluation) };
}
// A completed scan whose latest run is older than this is considered "stale" in
// the Security overview. Named so the route and its tests share one value.
export const STALE_SCAN_THRESHOLD_MS = 7 * 24 * 60 * 60 * 1000; // 7 days
// Shape of the /overview response. Mirrors the frontend `SecurityOverview` type
// (frontend/src/types/security.ts); annotating the response below makes a
// renamed or dropped field a compile error here instead of an undefined read in
// the UI. Keep the two in sync.
interface SecurityOverviewResponse {
scannedImages: number;
critical: number;
high: number;
fixable: number;
secrets: number;
misconfigs: number;
staleScans: number;
failedScans: number;
lastSuccessfulScanAt: number | null;
scanner: { available: boolean; version: string | null; source: 'managed' | 'host' | 'none'; autoUpdate: boolean };
deployEnforcement: { honorSuppressionsOnDeploy: boolean; eligibleBlockPolicies: number };
}
export const securityRouter = Router();
securityRouter.get('/trivy-status', authMiddleware, (_req: Request, res: Response) => {
@@ -437,6 +460,83 @@ securityRouter.get('/image-summaries', authMiddleware, (req: Request, res: Respo
}
});
// Node-scoped security posture rollup for the Security page Overview. Read-only,
// auth-only (Community). Counts derive from the latest-completed-scan-per-image
// summaries plus two precise helpers; the deploy-enforcement block is this
// node's read-only posture, not policy management.
securityRouter.get('/overview', authMiddleware, (req: Request, res: Response): void => {
try {
const db = DatabaseService.getInstance();
const summaries = Object.values(db.getImageScanSummaries(req.nodeId));
const settings = db.getGlobalSettings();
const svc = TrivyService.getInstance();
const now = Date.now();
let scannedImages = 0;
let critical = 0;
let high = 0;
let fixable = 0;
let secrets = 0;
let misconfigs = 0;
let staleScans = 0;
let lastSuccessfulScanAt: number | null = null;
for (const s of summaries) {
// Severity, secret, and misconfig totals are summed across every summary
// (real images and stack/config scans alike). Only scannedImages excludes
// the stack/config rows (stored under a "stack:" image_ref), since those
// are stacks, not images.
if (!s.image_ref.startsWith('stack:')) scannedImages += 1;
critical += s.critical;
high += s.high;
fixable += s.fixable;
secrets += s.secret_count;
misconfigs += s.misconfig_count;
if (now - s.scanned_at > STALE_SCAN_THRESHOLD_MS) staleScans += 1;
if (lastSuccessfulScanAt === null || s.scanned_at > lastSuccessfulScanAt) {
lastSuccessfulScanAt = s.scanned_at;
}
}
const overview: SecurityOverviewResponse = {
scannedImages,
critical,
high,
fixable,
secrets,
misconfigs,
staleScans,
failedScans: db.countScansByStatus(req.nodeId, 'failed'),
lastSuccessfulScanAt,
scanner: {
available: svc.isTrivyAvailable(),
version: svc.getVersion(),
source: svc.getSource(),
autoUpdate: settings.trivy_auto_update === '1',
},
deployEnforcement: {
honorSuppressionsOnDeploy: settings.deploy_block_honor_suppressions === '1',
eligibleBlockPolicies: db.countEligibleBlockPolicies(
req.nodeId,
FleetSyncService.getRole(),
FleetSyncService.getSelfIdentity(),
),
},
};
res.json(overview);
} catch (error) {
console.error('[Security] Failed to build overview:', error);
res.status(500).json({ error: 'Failed to build security overview' });
}
});
// Static, read-only policy-pack catalog. Auth-only (Community), no DB, no
// enforcement. The frontend fetches this with localOnly so the global catalog
// is available regardless of which node is active.
securityRouter.get('/policy-packs', authMiddleware, (_req: Request, res: Response): void => {
res.json(DEFAULT_POLICY_PACKS);
});
securityRouter.post('/sbom', authMiddleware, async (req: Request, res: Response): Promise<void> => {
if (!requireAdmin(req, res)) return;
const svc = TrivyService.getInstance();
+44 -1
View File
@@ -703,6 +703,8 @@ export interface ScanSummary {
low: number;
unknown: number;
fixable: number;
secret_count: number;
misconfig_count: number;
scanned_at: number;
scan_id: number;
}
@@ -4451,7 +4453,7 @@ export class DatabaseService {
.prepare(
`SELECT vs.image_ref, vs.id as scan_id, vs.highest_severity, vs.total_vulnerabilities,
vs.critical_count, vs.high_count, vs.medium_count, vs.low_count,
vs.unknown_count, vs.fixable_count, vs.scanned_at
vs.unknown_count, vs.fixable_count, vs.secret_count, vs.misconfig_count, vs.scanned_at
FROM vulnerability_scans vs
INNER JOIN (
SELECT image_ref, MAX(scanned_at) AS max_scanned
@@ -4472,6 +4474,8 @@ export class DatabaseService {
low_count: number;
unknown_count: number;
fixable_count: number;
secret_count: number;
misconfig_count: number;
scanned_at: number;
}>;
const out: Record<string, ScanSummary> = {};
@@ -4486,6 +4490,8 @@ export class DatabaseService {
low: r.low_count,
unknown: r.unknown_count,
fixable: r.fixable_count,
secret_count: r.secret_count,
misconfig_count: r.misconfig_count,
scanned_at: r.scanned_at,
scan_id: r.scan_id,
};
@@ -4493,6 +4499,43 @@ export class DatabaseService {
return out;
}
/**
* Uncapped count of scans in a given status for a node. Unlike
* `getVulnerabilityScans`, this never applies the per-image history cap, so
* the Security overview reports the true number of (for example) failed
* scans rather than a capped grouped total.
*/
public countScansByStatus(nodeId: number, status: VulnScanStatus): number {
return (
this.db
.prepare(
'SELECT COUNT(*) AS cnt FROM vulnerability_scans WHERE node_id = ? AND status = ?',
)
.get(nodeId, status) as { cnt: number }
).cnt;
}
/**
* Count of enabled block-on-deploy policies that are eligible to apply to
* this node: fleet-wide (node_id IS NULL) or scoped to this node. Built on
* `getScanPoliciesForUi` so a replica never counts policies scoped to a
* sibling node's identity. Stack-pattern applicability is not evaluated
* (there is no concrete stack name at overview scope), so this is an
* approximate "is this node enforcing" indicator, not a per-stack guarantee.
*/
public countEligibleBlockPolicies(
nodeId: number,
role: 'control' | 'replica',
selfIdentity: string,
): number {
return this.getScanPoliciesForUi(role, selfIdentity).filter(
(p) =>
p.enabled === 1 &&
p.block_on_deploy === 1 &&
(p.node_id === null || p.node_id === nodeId),
).length;
}
// --- Scan Policies ---
public getScanPolicies(): ScanPolicy[] {
+255
View File
@@ -0,0 +1,255 @@
import type { VulnSeverity } from './DatabaseService';
/**
* Static, read-only policy-pack catalog.
*
* Policy packs are curated bundles of security expectations for a deployment
* posture (homelab, production, public edge, ...). This module is the single
* source of truth for the catalog: it has no database, no I/O, and no
* enforcement wiring. The Security page renders it as an educational reference
* so operators can understand what good looks like for their environment.
*
* Each rule carries a stable id, a severity, plain-language explanations
* (what it checks, why it matters, how to fix), and an `enforcement` marker.
* `warning` rules are advisory; `enforceable` rules are the ones a future
* enforcement phase can promote into a block-on-deploy scan policy. The marker
* is metadata only here: nothing in this module blocks a deploy.
*/
export type PolicyRuleEnforcement = 'warning' | 'enforceable';
export interface PolicyPackRule {
/** Stable identifier for the rule within the catalog. */
id: string;
name: string;
severity: Exclude<VulnSeverity, 'UNKNOWN'>;
/** What the rule inspects in a Compose stack or image. */
whatItChecks: string;
/** Why the check matters for security or reliability. */
why: string;
/** Concrete remediation guidance. */
howToFix: string;
/** Advisory (`warning`) or promotable to a block-on-deploy policy (`enforceable`). */
enforcement: PolicyRuleEnforcement;
}
export interface PolicyPack {
/** Stable identifier for the pack. */
id: string;
name: string;
/** One-line description of the posture the pack targets. */
tagline: string;
/** Descriptive copy about how the pack is meant to be used. */
tierCopy: string;
rules: PolicyPackRule[];
}
/**
* Canonical rule definitions, keyed by id. Packs reference these so the
* what/why/how copy stays consistent everywhere a rule appears; each pack
* sets its own `enforcement` level for the rule.
*/
type RuleDefinition = Omit<PolicyPackRule, 'enforcement'>;
const RULE_CATALOG = {
'no-privileged': {
id: 'no-privileged',
name: 'No privileged containers',
severity: 'CRITICAL',
whatItChecks: 'Services that set privileged: true.',
why: 'A privileged container can access all host devices and effectively escape isolation, so a single compromised service can take over the host.',
howToFix: 'Remove privileged: true and grant only the specific capabilities or device mounts the workload actually needs.',
},
'no-docker-socket': {
id: 'no-docker-socket',
name: 'No Docker socket mounts',
severity: 'CRITICAL',
whatItChecks: 'Bind mounts of /var/run/docker.sock into a container.',
why: 'Access to the Docker socket is equivalent to root on the host: the container can start new privileged containers or read every other stack.',
howToFix: 'Drop the socket mount. Where a tool genuinely needs Docker access, use a scoped proxy with a read-only, filtered API surface.',
},
'no-host-network': {
id: 'no-host-network',
name: 'No host networking',
severity: 'HIGH',
whatItChecks: 'Services using network_mode: host.',
why: 'Host networking removes network namespace isolation, exposes every container port on the host directly, and bypasses Compose network segmentation.',
howToFix: 'Use a bridge network and publish only the ports you need with explicit port mappings.',
},
'no-broad-bind-mounts': {
id: 'no-broad-bind-mounts',
name: 'No broad host bind mounts',
severity: 'HIGH',
whatItChecks: 'Bind mounts of sensitive host paths such as /, /etc, /var, or the host home directory.',
why: 'Mounting broad host paths lets a container read or modify host configuration and other services data, widening the blast radius of a compromise.',
howToFix: 'Mount only the specific subdirectory the service needs, and prefer named volumes for persistent data.',
},
'healthcheck-defined': {
id: 'healthcheck-defined',
name: 'Healthcheck defined',
severity: 'MEDIUM',
whatItChecks: 'Services that declare no healthcheck.',
why: 'Without a healthcheck the orchestrator cannot tell a hung container from a healthy one, so failures go unnoticed and dependent services start too early.',
howToFix: 'Add a healthcheck with a command that reflects real readiness, plus sensible interval, timeout, and retries.',
},
'no-public-db-ports': {
id: 'no-public-db-ports',
name: 'No public database ports',
severity: 'HIGH',
whatItChecks: 'Database services that publish their port to the host (for example 5432, 3306, 27017, 6379).',
why: 'Publishing a database port exposes it to anything that can reach the host, a common path to data theft on internet-adjacent machines.',
howToFix: 'Remove the host port mapping and let other services reach the database over the internal Compose network instead.',
},
'run-as-non-root': {
id: 'run-as-non-root',
name: 'Run as non-root',
severity: 'MEDIUM',
whatItChecks: 'Containers that run as root when a non-root user is available.',
why: 'Running as root raises the impact of a container breakout and of any write to a mounted host path.',
howToFix: 'Set a non-root user with the user directive, or use an image that ships a dedicated runtime user.',
},
'pin-image-tag': {
id: 'pin-image-tag',
name: 'Pin image tags',
severity: 'LOW',
whatItChecks: 'Images referenced by the latest tag or with no tag at all.',
why: 'A floating tag makes deploys non-reproducible: the same Compose file can pull different code on different days, including a regressed or compromised build.',
howToFix: 'Pin a specific version tag, and pin a digest for the strongest guarantee.',
},
'restart-policy': {
id: 'restart-policy',
name: 'Restart policy set',
severity: 'LOW',
whatItChecks: 'Services with no restart policy.',
why: 'Without a restart policy a crashed service stays down until someone notices, which turns a transient fault into an outage.',
howToFix: 'Set restart: unless-stopped (or on-failure) so the service recovers from crashes and host reboots.',
},
'no-plaintext-secrets': {
id: 'no-plaintext-secrets',
name: 'No plaintext secrets',
severity: 'HIGH',
whatItChecks: 'Credentials, tokens, or keys detected in Compose files, env values, or image layers.',
why: 'Secrets committed alongside a stack leak through backups, version control, and image registries, and are trivial to extract from a pulled image.',
howToFix: 'Move secrets into an .env file kept out of version control, or a secrets manager, and reference them by variable.',
},
'resource-limits': {
id: 'resource-limits',
name: 'Resource limits set',
severity: 'LOW',
whatItChecks: 'Services with no memory or CPU limits.',
why: 'An unbounded service can exhaust host memory or CPU and starve every other stack on the node.',
howToFix: 'Set memory and CPU limits sized to the workload so one service cannot monopolize the host.',
},
'pin-digest': {
id: 'pin-digest',
name: 'Pin image digest',
severity: 'LOW',
whatItChecks: 'Images that are not pinned to a content digest.',
why: 'A tag can be repointed at a different image after you have reviewed it; a digest is immutable and guarantees you run exactly what you vetted.',
howToFix: 'Reference the image by digest (image@sha256:...) for workloads that need supply-chain certainty.',
},
'read-only-rootfs': {
id: 'read-only-rootfs',
name: 'Read-only root filesystem',
severity: 'MEDIUM',
whatItChecks: 'Containers whose root filesystem is writable.',
why: 'A writable root filesystem lets an attacker drop tools or persist a foothold inside the container.',
howToFix: 'Set read_only: true and mount tmpfs or named volumes for the few paths that must be writable.',
},
'drop-capabilities': {
id: 'drop-capabilities',
name: 'Drop unnecessary capabilities',
severity: 'MEDIUM',
whatItChecks: 'Containers that keep the default Linux capability set instead of dropping what they do not use.',
why: 'Extra capabilities give a compromised process more ways to affect the host than the workload actually requires.',
howToFix: 'Drop ALL capabilities and add back only the ones the service needs (cap_drop / cap_add).',
},
} satisfies Record<string, RuleDefinition>;
// Closed set of rule ids derived from the catalog. Typing `rule()` against this
// turns a mistyped or deleted id into a compile error instead of a runtime throw.
type RuleId = keyof typeof RULE_CATALOG;
function rule(id: RuleId, enforcement: PolicyRuleEnforcement): PolicyPackRule {
return { ...RULE_CATALOG[id], enforcement };
}
/**
* The default catalog. Frozen so callers cannot mutate the shared definitions.
* Order is intentional: gentlest posture first, strictest last.
*/
export const DEFAULT_POLICY_PACKS: readonly PolicyPack[] = Object.freeze([
{
id: 'homelab-baseline',
name: 'Homelab baseline',
tagline: 'Gentle defaults for a single-operator homelab.',
tierCopy: 'Advisory guidance that flags the few habits worth keeping without getting in your way.',
rules: [
rule('no-plaintext-secrets', 'warning'),
rule('pin-image-tag', 'warning'),
rule('run-as-non-root', 'warning'),
rule('restart-policy', 'warning'),
],
},
{
id: 'production-hardening',
name: 'Production hardening',
tagline: 'Sensible hardening for services that face real traffic.',
tierCopy: 'Warns on risky exposure, missing healthchecks, and the highest-impact misconfigurations.',
rules: [
rule('no-privileged', 'enforceable'),
rule('no-docker-socket', 'enforceable'),
rule('no-plaintext-secrets', 'enforceable'),
rule('no-host-network', 'warning'),
rule('healthcheck-defined', 'warning'),
rule('drop-capabilities', 'warning'),
rule('read-only-rootfs', 'warning'),
rule('pin-image-tag', 'warning'),
],
},
{
id: 'strict-production',
name: 'Strict production',
tagline: 'Zero-tolerance posture for critical workloads.',
tierCopy: 'The strictest baseline, intended for workloads where reproducibility and isolation are non-negotiable.',
rules: [
rule('no-privileged', 'enforceable'),
rule('no-docker-socket', 'enforceable'),
rule('no-host-network', 'enforceable'),
rule('no-broad-bind-mounts', 'enforceable'),
rule('no-plaintext-secrets', 'enforceable'),
rule('run-as-non-root', 'enforceable'),
rule('healthcheck-defined', 'enforceable'),
rule('resource-limits', 'enforceable'),
rule('pin-digest', 'enforceable'),
],
},
{
id: 'public-edge',
name: 'Public edge service',
tagline: 'Focused on services exposed to the public internet.',
tierCopy: 'Emphasizes secret leakage, exposed ports, and the misconfigurations that matter most at the edge.',
rules: [
rule('no-plaintext-secrets', 'enforceable'),
rule('no-public-db-ports', 'enforceable'),
rule('no-host-network', 'enforceable'),
rule('no-privileged', 'enforceable'),
rule('healthcheck-defined', 'warning'),
rule('pin-image-tag', 'warning'),
],
},
{
id: 'internal-service',
name: 'Internal service',
tagline: 'Least-privilege defaults for east-west internal services.',
tierCopy: 'Avoids public exposure and broad host access while keeping internal services easy to run.',
rules: [
rule('no-public-db-ports', 'warning'),
rule('no-broad-bind-mounts', 'warning'),
rule('run-as-non-root', 'warning'),
rule('drop-capabilities', 'warning'),
rule('resource-limits', 'warning'),
rule('pin-image-tag', 'warning'),
],
},
]);