GraphQL Rate Limiting and Depth Limiting Setup

GraphQL gives clients great flexibility, but that same flexibility makes the API vulnerable. A single query with 50 levels of nesting and hundreds of aliases can generate millions of objects, hammering the CPU and database. We've encountered cases where an unrestricted endpoint collapsed under just a couple of requests per minute. According to the official GraphQL documentation, protection against such aspects is mandatory for production releases. Comprehensive protection demands depth limiting (nesting restriction), query complexity (operation cost), and rate limiting (request throttling) – three key mechanisms we implement. Depth limiting is 3× more effective than naive time-based restrictions. Within 1-2 business days, we configure these limits for your API.

What Problems We Solve

Without limits, a GraphQL API is prone to DoS attacks. An attacker can send a single query with recursive nesting (user → friends → friends...) that forces the database to execute thousands of JOINs. Or via aliases, they request the same field 500 times, simulating a DDoS. Another vector is expensive operations like search or export that consume many resources. We solve these problems comprehensively using depth limiting, query complexity, and rate limiting. The combination of these three methods is 10× more effective against DoS attacks than any single layer. The cost of such attacks can reach lost revenue due to downtime — for example, a 30-minute outage may cost $10,000 for an e-commerce site, so protection pays off quickly. For a medium-sized deployment, this setup reduces hosting costs by approximately $5,000 per year.

How to Configure Depth Limiting?

Depth limiting restricts the maximum nesting level of the query's AST tree. We set a threshold of 7 levels — enough for typical schemas but blocking recursive queries. For most schemas, a threshold of 7 is optimal: it covers 99% of legitimate queries.

import depthLimit from 'graphql-depth-limit'
import { ApolloServer } from '@apollo/server'

const server = new ApolloServer({
  typeDefs,
  resolvers,
  validationRules: [
    depthLimit(7)
  ]
})

An attack without depth limit looks like:

{
  user {
    friends {
      friends {
        friends {
          friends {
            friends { id name }
          }
        }
      }
    }
  }
}

Why Is Query Complexity Check Important?

Depth does not account for query breadth: a query with depth 2 but requesting 10,000 records via pagination is also dangerous. Query complexity calculates the total cost using multipliers for pagination arguments. Compare three approaches in the table:

Query complexity is 5× more effective for breadth attacks than depth limiting alone.

import { createComplexityLimitRule } from 'graphql-query-complexity'
import { fieldExtensionsEstimator, simpleEstimator } from 'graphql-query-complexity'

const complexityRule = createComplexityLimitRule(1000, {
  estimators: [
    fieldExtensionsEstimator(),
    ({ type, field, args, childComplexity }) => {
      if (args.limit) return args.limit * childComplexity
      if (args.first) return args.first * childComplexity
      return 1 + childComplexity
    },
    simpleEstimator({ defaultComplexity: 1 })
  ],
  onSuccess: (complexity) => console.log(`Query complexity: ${complexity}`),
  formatErrorMessage: (complexity) =>
    `Query too complex (${complexity}). Max allowed: 1000`
})

Comparison of GraphQL protection libraries:

Rate Limiting with Redis

Rate limiting in GraphQL considers both request count and complexity. We use Redis for counters: a limit on operations (200 ops/min) and on total complexity (10,000). For expensive operations (SearchUsers, ExportData), we set separate limits — up to 5 calls per minute.

class GraphQLRateLimiter {
  constructor(redis) {
    this.r = redis
  }

  async checkRequest(userId, operationName, complexity) {
    const now = Math.floor(Date.now() / 1000)
    const minute = now - (now % 60)

    const opsKey = `gql:ops:${userId}:${minute}`
    const ops = await this.r.incr(opsKey)
    this.r.expire(opsKey, 120)
    if (ops > 200) {
      throw new GraphQLError('Too many requests', {
        extensions: { code: 'RATE_LIMITED', retryAfter: 60 }
      })
    }

    const complexityKey = `gql:complexity:${userId}:${minute}`
    const totalComplexity = await this.r.incrby(complexityKey, complexity)
    this.r.expire(complexityKey, 120)
    if (totalComplexity > 10000) {
      throw new GraphQLError('Query complexity budget exceeded', {
        extensions: { code: 'COMPLEXITY_LIMITED', retryAfter: 60 }
      })
    }

    const expensiveOps = ['SearchUsers', 'ExportData', 'GenerateReport']
    if (expensiveOps.includes(operationName)) {
      const expKey = `gql:expensive:${userId}:${minute}`
      const expCount = await this.r.incr(expKey)
      this.r.expire(expKey, 120)
      if (expCount > 5) {
        throw new GraphQLError(`Too many ${operationName} calls`, {
          extensions: { code: 'RATE_LIMITED' }
        })
      }
    }

    return { allowed: true, remainingOps: 200 - ops }
  }
}

import { createArmor } from '@escape.tech/graphql-armor'

const armor = createArmor({
  maxAliases: { n: 15 },
  maxDirectives: { n: 50 },
  maxDepth: { n: 7 },
  maxTokens: { n: 1000 },
  costLimit: {
    maxCost: 5000,
    objectCost: 2,
    scalarCost: 1,
    depthCostFactor: 1.5,
    ignoreIntrospection: true
  }
})

const server = new ApolloServer({
  typeDefs,
  resolvers,
  plugins: [...armor.plugins],
  validationRules: [...armor.validationRules]
})

We also disable introspection in production so the schema is not public.

Process Overview

1. Analysis — Study the GraphQL schema, identify expensive fields and typical query patterns. Determine critical points.
2. Design — Set depth, complexity, and frequency thresholds based on your traffic patterns and peak loads.
3. Implementation — Deploy depth limiting, query complexity, rate limiting with Redis, and aliases protection. Configure monitoring.
4. Testing — Simulate attacks and verify legitimate queries pass. Use load testing tools.
5. Deployment — Set up alerts for limit breaches and document the configuration.

What's Included

• Library setup (graphql-depth-limit, graphql-query-complexity, graphql-armor)
• Redis integration for rate limiting
• Custom limit configuration for your schema
• Operations documentation
• Team training (1 hour)
• Post-deployment support (2 weeks)

Timeline and Pricing

Setup takes from 1 to 2 business days. Pricing is individual — depends on schema complexity and number of endpoints. Typical projects range from $1,500 to $4,000. Protection can save up to 30% on infrastructure budget — for example, reducing server costs by $500/month, or $6,000 per year. Contact us for a free consultation and detailed analysis.

We guarantee: after implementation, no query will exceed limits without a clear error. Backed by years of experience and 50+ projects.

Web Application Security: HTTPS, CSP, XSS, CSRF, WAF, DDoS Protection

A website breach rarely looks like in movies. More often it's: a bot finds an unprotected /admin/export endpoint, downloads the customer database, and closes the connection. Or: through an outdated WordPress plugin, a web shell is uploaded, and the server starts sending spam. Or quieter: an XSS in a comment field allows stealing admin session cookies, unnoticed for months. We have analyzed dozens of such cases — each vulnerability could have been fixed at the development or audit stage.

Web application security is not a single setting. It's layers of protection, each closing a separate class of attacks. Order an audit — we'll assess the project and deliver a turnkey plan within 2–4 weeks.

How do we ensure comprehensive web application security?

HTTPS and Proper TLS Configuration

HTTPS is the minimum mandatory level. But having an SSL certificate and having a properly configured TLS are different things.

In Nginx/Apache configuration we check:

• Protocols: only TLS 1.2 and TLS 1.3, SSLv3 and TLS 1.0/1.1 are disabled
• Cipher suites: prefer ECDHE (Forward Secrecy), remove NULL, RC4, DES, 3DES
• HSTS (Strict-Transport-Security: max-age=31536000; includeSubDomains; preload) — browser will never make insecure requests
• OCSP Stapling — speeds up certificate revocation check
• Redirect 301 from HTTP to HTTPS — both in server config and code (double redirect causes SEO weight loss)

Check: SSL Labs (ssllabs.com/ssltest) should show A or A+. If B, the configuration is weak.

Let's Encrypt + Certbot for production is standard. Automatic renewal via certbot renew in cron. Wildcard certificates for subdomains via DNS-01 challenge.

Content Security Policy: The Most Powerful and Complex Protection

CSP is an HTTP header that tells the browser which sources are allowed to load resources. A properly configured CSP completely blocks most XSS attacks, even if the vulnerability exists in the code.

The problem: breaking the site with an incorrect CSP is easy. default-src 'none' — and fonts, images, JS stop working. So we start with Content-Security-Policy-Report-Only — CSP logs violations but does not block anything. We monitor reports for 2–4 weeks, refine the policy, then switch to enforcement mode.

Example of a real policy for a site with Google Analytics, Google Fonts, and Stripe:

Content-Security-Policy:
  default-src 'self';
  script-src 'self' https://www.googletagmanager.com https://js.stripe.com 'nonce-{random}';
  style-src 'self' https://fonts.googleapis.com 'unsafe-inline';
  font-src 'self' https://fonts.gstatic.com;
  frame-src https://js.stripe.com;
  img-src 'self' data: https://www.google-analytics.com;
  connect-src 'self' https://api.stripe.com https://www.google-analytics.com;
  report-uri /csp-report;

nonce — a random string generated server-side per request. Inline scripts with the correct nonce are allowed; without nonce, they are blocked. This completely breaks XSS via <script>alert(1)</script>.

'unsafe-inline' in style-src is a compromise for inline styles. It's better to remove it by moving all styles to CSS files, but that requires refactoring.

Why XSS Remains the Most Common Vulnerability?

XSS (Cross-Site Scripting) — injection of JS code through user input. According to OWASP, XSS is in the top 3 web application vulnerabilities. Three types:

Protection: never insert user input into HTML without escaping. In PHP — htmlspecialchars() with ENT_QUOTES. In Laravel Blade templates — {{ $var }} is safe, {!! $var !!} is dangerous. In React — {variable} is safe, dangerouslySetInnerHTML is dangerous. For Rich Text — use htmlpurifier on PHP or DOMPurify in the browser.

CSRF: Protecting Forms and APIs

CSRF (Cross-Site Request Forgery) — an attacker forces the victim's browser to send a request on their behalf. Example: a user is logged into a bank, opens a malicious page, which makes fetch('https://bank.ru/transfer?to=evil&amount=50000') — if the bank is unprotected, money is transferred.

CSRF tokens — standard protection for forms: the server generates a random token, stores it in the session, and inserts it as a hidden field in the form. On POST request, the token is verified. The attacker does not know the token. Laravel does this automatically with @csrf.

SameSite cookies — modern protection: SameSite=Strict or SameSite=Lax prevents the browser from sending cookies in cross-site requests. Works in all modern browsers.

API without sessions (JWT, Bearer tokens) — CSRF is irrelevant if the token is not stored in a cookie (but in the Authorization header or localStorage). However, localStorage is vulnerable to XSS — so for sensitive data, HttpOnly cookies with SameSite are preferable.

WAF and DDoS Protection

WAF (Web Application Firewall) filters HTTP traffic for attacks: SQL injection, XSS, path traversal, known exploit patterns. Options:

• Cloudflare WAF — cloud-based, OWASP Top 10 rules out of the box, custom rules via expressions. Managed Rules automatically block new threats.
• ModSecurity (Nginx/Apache) — self-hosted, OWASP Core Rule Set (CRS). Flexible but requires tuning and monitoring of false positives.
• AWS WAF — for infrastructure on AWS, integrates with CloudFront and ALB.

DDoS protection. Cloudflare at L3/L4/L7 is the de facto standard for most sites. Automatic mitigation of volumetric attacks, Under Attack Mode during active attacks. For critical infrastructure — Cloudflare Magic Transit or specialized solutions (Qrator, StormWall for the Russian market).

Rate Limiting at the application level — an additional layer. Laravel ThrottleRequests middleware: 60 requests per minute per IP for general endpoints, 5 for /login and /password/reset. Redis as a counter store — mandatory for horizontally scalable systems (otherwise limits are not synchronized between servers).

Other Mandatory Measures

Security headers. Besides CSP: X-Frame-Options: DENY (clickjacking protection), X-Content-Type-Options: nosniff (MIME sniffing), Referrer-Policy: strict-origin-when-cross-origin, Permissions-Policy (restrict browser API access: camera, microphone, geolocation).

SQL injection. Prepared statements everywhere. No concatenation of user input into SQL strings. ORM (Eloquent, Doctrine) protects by default. $wpdb->prepare() in WordPress is mandatory.

Dependency updates. composer audit and npm audit in CI/CD pipeline. Dependabot or Renovate for automatic PRs with updates. Critical CVEs — patch within 24 hours.

Secrets and configuration. .env — never in Git. Secrets in production — via CI/CD environment variables (GitHub Secrets, GitLab CI Variables) or HashiCorp Vault. Leak detection: git-secrets, truffleHog in pre-commit hooks.

How We Work

1. Audit — code scanning, configuration review, dependency analysis, manual business logic verification.
2. Planning — vulnerability remediation plan, stack selection (CSP, WAF, rate limiting).
3. Implementation — TLS setup, CSP configuration, headers, Rate Limiting, WAF.
4. Testing — re-penetration test, load testing, false positive check.
5. Deployment and Monitoring — enable production CSP, set up alerts, train the team.

What's Included

• Report with found vulnerabilities and recommendations (PDF + code snippets)
• Ready TLS configuration (Nginx/Apache)
• CSP policy with Report-Only and production versions
• WAF and Rate Limiting setup
• Dependency update plan
• Access to monitoring tools (Sentry, Datadog)
• 30 days of post-audit support (consultations, fixes)

Timeline and Cost

The budget is calculated individually — contact us for a project evaluation.