L7 Firewall (Extensible Layer 7 Firewall Prototype)

An extensible Layer 7 (HTTP) firewall / request decision engine for Node.js / Express with a pluggable rule chain. Supports path allow/block rules, rate limiting, interactive challenge (captcha‑style) flow with attempt limits + temporary blocking, and optional live telemetry statistics.

Features

• Pluggable rule engine executed by priority (lowest number first).
• Built‑in rules:
  • PathRule – allow / block lists (exact or wildcard suffix *).
  • RateLimitRule – simple token bucket per key (IP by default).
  • CaptchaRule – math challenge for configured paths; includes:
    • Ephemeral pending challenges with TTL.
    • Attempt counter & configurable max attempts.
    • Automatic temporary block of key after repeated failures.
    • Issued token (header) for future bypass until expiry.
  • TurnstileRule – Cloudflare Turnstile widget integration (HTML challenge page or JSON with HTML snippet) for frictionless human verification.
  • BotDetectionRule – heuristic bot / spoof detection (UA length/patterns, OS allow/block, header presence, path entropy) returning challenge or block.
• Default allow/block fallback logic with optional short‑circuit on allow.
• Telemetry counters (allow / block / challenge / pass / fail) and per‑rule breakdown.
• Optional periodic console stats (every 1s by default) when enabled.
• Jest + Supertest tests (allow, block, challenge, rate limit).
• Zero build step (pure CommonJS JavaScript).

Quick Start

Install dependencies:

npm install

Run the server (port 80 by default; may need admin privileges on some systems – or set PORT):

npm start

In another terminal run the demo client (targets http://localhost):

npm run demo

You should see an allowed response for /pass and a blocked response for /block.

Run tests:

npm test

Adding / Customizing Rules

In src/server.js (or your own app) you can register, enable, or disable rules:

firewall.register(new PathRule({ allow: ['/public*'], block: ['/admin'] }), { id: 'paths', priority: 10 });
firewall.register(new RateLimitRule({ windowMs: 60000, max: 100 }), { id: 'ratelimit', priority: 50 });
firewall.register(new CaptchaRule({ triggerPaths: ['/login','/signup'] }), { id: 'captcha', priority: 80 });

Lower numeric priority runs earlier. Rules return decisions; first block/challenge short-circuits; allow may short-circuit if shortCircuitOnAllow enabled.

Full Customization Examples

Block extra paths dynamically:

const { firewall } = require('l7-firewall');
firewall.enable('paths', true); // ensure paths rule enabled
// Modify PathRule instance (retrieve via internal reference)
const pathRule = firewall.rules.find(r => r.id === 'paths').rule;
pathRule.block.push('/secret');

Add a new high-priority temporary block rule:

class TempBlockRule { constructor(list){ this.list=list; this.id='temp-block'; }
  evaluate(req){ return this.list.includes(req.path) ? {action:'block', status:451, reason:'Temp block'}:null; } }
firewall.register(new TempBlockRule(['/maintenance']), { id:'temp-block', priority:5 });

Rate limit a specific path harder:

const { RateLimitRule } = require('l7-firewall/rules');
firewall.register(new RateLimitRule({ windowMs: 10000, max: 5, keyFn: r => 'login:'+ (r.ip||'global') }), { id:'login-rate', priority:40 });

Captcha only on /signup and /reset:

const { CaptchaRule } = require('l7-firewall/rules');
firewall.register(new CaptchaRule({ triggerPaths: ['/signup','/reset'] }), { id: 'captcha-signup', priority: 70 });

Challenge (Captcha) Flow

When CaptchaRule decides a challenge is required it returns something like:

{
  action: 'challenge',
  status: 403,
  reason: 'Captcha required',
  challenge: {
    type: 'math',
    question: '3+4?',
    param: 'cf_answer',       // query parameter expected for answer
    header: 'x-captcha-token',// header to send back once solved
    remaining: 3              // attempts remaining before temp block
  }
}

Client handling steps:

1. Display challenge.question to user.
2. User solves locally; resend original request adding ?cf_answer=7.
3. On success response: { action:'allow', token:'<token>', reason:'Captcha passed' } (middleware example includes token if you expose it) – store this token (e.g. in a cookie or memory) and send in header x-captcha-token: <token> for subsequent protected requests.
4. If answer wrong: server returns a new challenge with reason: 'Captcha retry' and remaining decreased. After maxAttempts failures the key (IP) is temporarily blocked (Captcha failure temporary block).

CaptchaRule key options:

new CaptchaRule({
  triggerPaths: ['/login','/signup'], // where to require captcha
  challengeParam: 'cf_answer',        // query param for answer
  tokenHeader: 'x-captcha-token',     // header carrying solved token
  ttlMs: 10 * 60 * 1000,              // token validity
  maxAttempts: 3,                     // failures before block
  blockDurationMs: 5 * 60 * 1000,     // temporary block length
  challengeTtlMs: 2 * 60 * 1000       // time window to answer specific challenge
});

Returned decision meta.challengeResult values (when telemetry enabled): passed, failed, or blocked.

Turnstile (Cloudflare) Human Verification

If you prefer Cloudflare Turnstile over math captchas enable TurnstileRule with your keys:

const { TurnstileRule } = require('l7-firewall/rules');
firewall.register(new TurnstileRule({
  triggerPaths: ['/login','/signup'],
  siteKey: process.env.TURNSTILE_SITE_KEY,
  secretKey: process.env.TURNSTILE_SECRET_KEY,
  ttlMs: 10*60*1000
}), { id: 'turnstile', priority: 70 });

On first access decision:

{
  action: 'challenge',
  ruleId: 'turnstile',
  challenge: { type: 'turnstile', siteKey: '<your-site-key>', html: "<!DOCTYPE html>..." },
  reason: 'Turnstile verification required'
}

If you return this as JSON, your frontend can inject challenge.html into a modal / page. The default server demo returns JSON; you can instead detect Accept: text/html and serve raw HTML (customize via htmlTemplate). After the user completes the widget the browser submits a POST containing cf-turnstile-response which the rule verifies server-side (unless SKIP_TURNSTILE_VERIFY=1 set for tests). A successful verification caches an allow for that key for ttlMs.

Environment variables fallback: TURNSTILE_SITE_KEY, TURNSTILE_SECRET_KEY. For tests you can set SKIP_TURNSTILE_VERIFY=1 to bypass remote calls.

Bot Detection Heuristics

BotDetectionRule flags suspicious requests and either challenges (default) or blocks them. Key options:

new BotDetectionRule({
  allowUserAgents: ['^Mozilla/.*'],
  blockUserAgents: ['curl/.*','wget'],
  allowOS: [],            // only these OS allowed (empty = any)
  blockOS: ['unknown'],
  minUserAgentLength: 25,
  entropyPathThreshold: 0.6, // normalized per-char threshold
  entropyMinLength: 12,
  suspiciousAction: 'challenge', // or 'block'
  pathAllowList: ['/','/pass','/login'],
  pathBlockList: ['/bad*'],
  headerChecks: { acceptRequired: true, languageRequired: false }
});

Heuristics applied (in order): path blocklist, UA allowlist (neutral), UA blocklist, UA length, OS allow/block lists (parsed simply from UA), header presence (Accept / Accept-Language), path entropy (detects random probing). Path entropy uses Shannon entropy per character; long highly random paths are challenged. Returned challenge example:

{
  action: 'challenge',
  status: 403,
  reason: 'High-entropy path',
  ruleId: 'bot-detect',
  challenge: { type: 'bot-check', reason: 'High-entropy path', entropy: '0.731' }
}

Telemetry & Live Stats

Firewall options now include:

const fw = new Firewall({
  telemetry: true,          // collect counters
  statsEnabled: true,       // periodically log snapshot
  statsIntervalMs: 1000,    // interval for console output
  // existing options ...
});

Console sample:

[Firewall stats] { uptimeSec: 12.0, total: 42, allowed: 30, blocked: 8, challenged: 4, challengePassed: 3, challengeFailed: 1, challengeBlocked: 0 }

Programmatic access:

fw.getStats(); // returns snapshot including per-rule counts

Shut down interval (e.g., in tests): fw.close().

Building & Publishing (npm)

Pure JS – no transpile step. Ensure package.json fields (name, version, description, repository, author, license, keywords) are set. Bump version and publish:

npm version patch   # or minor / major
npm publish --access public

Dry run first if you like:

npm pack            # creates tarball you can inspect

Add a files array in package.json (already configured if you see it) to limit published content, e.g.:

"files": ["src", "README.md", "LICENSE"]

If you need 2FA: npm login --otp <code> then publish.

Using As Dependency

npm install l7-firewall

const { Firewall } = require('l7-firewall/firewall');
const { PathRule, RateLimitRule, CaptchaRule } = require('l7-firewall/rules');
const fw = new Firewall({ defaultAction: 'allow' });
fw.register(new PathRule({ block:['/forbidden'] }), { id:'paths', priority:10 });
fw.register(new RateLimitRule({ windowMs:60000, max:100 }), { id:'rl', priority:50 });
// Express usage
app.use(async (req,res,next)=>{ const d=await fw.inspect(req); if(d.action==='block') return res.status(d.status||403).json({blocked:true,reason:d.reason}); if(d.action==='challenge') return res.status(d.status||403).json(d); next(); });

Rule Return Contract

{
  action: 'allow' | 'block' | 'challenge',
  status?: number,
  reason?: string,
  ruleId?: string,
  challenge?: { type: string, question?: string, ...custom },
  token?: string,                 // (CaptchaRule) on successful challenge
  meta?: {                        // optional metadata (telemetry / rule specific)
    challengeResult?: 'passed' | 'failed' | 'blocked'
  }
}

Roadmap Ideas

• Config-driven rule engine (YAML / JSON rules) with priorities.
• Allowlist / blocklist for IPs, CIDR ranges, headers, paths.
• Rate limiting (sliding window + token bucket hybrid).
• Request body pattern signatures & size limits.
• Anomaly scoring (suspicious header combos, unusual methods).
• Caching of repeat decisions for performance.
• Pluggable modules (e.g., RegexRule, GeoIPRule, RateLimitRule).
• Metrics & structured logging (Prometheus format, JSON logs).
• Admin API to reload rules without restart.

Decision Object Shape (Simplified)

At minimum decisions include: action, optionally status, reason, ruleId plus challenge / token / meta fields as above.

Security Notes

Prototype quality: in‑memory stores (rate limits, captcha tokens, blocks) are NOT distributed or persistent. For production you would:

• Replace in‑memory Maps with Redis / durable store.
• Sign or encrypt challenge tokens instead of trusting raw random strings.
• Add input validation, structured logging, metrics export, graceful shutdown.
• Harden around spoofable headers (x-forwarded-for) behind trusted proxy only.

License

MIT