Athenz: Fine-Grained, Role-Based Access Control at Yahoo

Ideas you will meet

Distributed Systems

Consensus, time, failure and the fundamentals of scale-out.

Web Security

OWASP risks, XSS, CSRF, CORS and secure design.

What we already know about this

Our own lessons on these ideas. Read them here before you start, so it makes sense.

• Networking and Web Security Deep Dive · 13 minThe OWASP Top 10 and secure designA map of the most common web application risks, what each means in practice and a short review checklist you can apply to any design.Read hereClose

  What the Top 10 is

  OWASP (the Open Worldwide Application Security Project) publishes an awareness list of the most critical web application risks, updated every few years from real data. Treat it as a checklist of categories to think about, not a complete standard. The names and order change between editions, so below the themes are described in stable terms with the current style of names in brackets.

  The recurring risk categories

  100%

  Figure SDA-NSD-05-1 · The risk categories as a review map · © forcsguy.com

  Category	In plain words	Typical prevention
  Broken access control	Users can act outside their permissions (view others' data, call admin features)	Check authorisation on the server for every request and object; deny by default; test with two users
  Cryptographic failures	Sensitive data exposed (no TLS, weak or home-made crypto, plaintext storage)	TLS everywhere, strong standard algorithms, hash passwords with a slow salted hash, manage keys properly
  Injection	Input interpreted as code or commands	Parameterised queries, validation, least privilege (see the attacks lesson)
  Insecure design	The design itself lacks controls	Threat modelling, abuse cases, secure patterns, limits on resources
  Security misconfiguration	Default passwords, open storage buckets, verbose errors, missing headers	Hardened baselines, infrastructure as code, regular scans
  Vulnerable and outdated components	Known flaws in libraries or platforms	Inventory dependencies, update, use automated advisories
  Identification and authentication failures	Weak passwords, no MFA, bad session handling	MFA, rate limiting, safe recovery, strong session management
  Software and data integrity failures	Untrusted updates, insecure CI/CD, unsafe deserialisation	Signed artefacts, protected pipelines, verify dependencies
  Security logging and monitoring failures	Attacks go unnoticed	Log security events, alert, practise incident response
  Server-side request forgery	The server fetches attacker-chosen URLs	Allow-lists, block internal ranges, isolation

  Authentication, authorisation, encryption, hashing

  • Authentication answers "who are you?"; authorisation answers "what may you do?". Do both on the server, every request.
  • Encryption is reversible with a key (protect data in transit and at rest). Hashing is one-way (verify without storing the original): for passwords use a slow, salted algorithm such as Argon2, bcrypt or scrypt, never plain SHA of the password. A salt is a unique random value per password that defeats precomputed tables. (Details in the passwords lesson.)
  • Never invent your own cryptography.

  Secure design in practice

  1. Threat model early. Draw the data flows, list assets, ask "who might attack this and how?" (a simple method: for each component consider spoofing, tampering, repudiation, information disclosure, denial of service and elevation of privilege).
  2. Least privilege everywhere: users, services, database accounts, cloud roles.
  3. Defence in depth: several independent controls.
  4. Secure defaults: deny first; features off until needed.
  5. Fail safe: on error, do not grant access; do not leak details.
  6. Minimise data: do not collect or keep what you do not need.
  7. Automate: dependency scanning, static analysis, secret scanning, container scanning in CI.
  8. Test: unit tests for authorisation rules, automated scans, periodic penetration tests.
  9. Prepare for incidents: logs, alerts, runbooks, key rotation, backups.

  A short review checklist for any feature

  • Who is allowed to do this, and where is that checked?
  • What input comes from users, and how is it validated and encoded?
  • What secrets and personal data are involved, and how are they protected and logged?
  • What can an attacker abuse in volume (rate limits, size limits, cost)?
  • What third-party code or services does it add, and are they maintained?
  • How would we notice if it were abused?

  Where this fits

  Security is a property of the whole system: network edge (previous lessons), application code (attacks lesson), identity (SSO and passwords lessons), data (encryption) and operations (monitoring and incident response). Make it a routine part of design reviews, not a final gate.

  Real-world reading

  Related reading

  • OWASP Cheat Sheet SeriesOWASP · OWASP FoundationDocumentation
  • AWS Security BlogAmazon Web Services engineers · Amazon Web ServicesBlogs
  • Google Security BlogGoogle engineers · GoogleBlogs
  • OWASP Top 10OWASP · OWASP FoundationDocumentation

  Show 6 moreShow fewer

  • Building Secure and Reliable SystemsGoogle engineers · O'Reilly / GoogleBooks
  • Athenz: Fine-Grained, Role-Based Access Control at YahooYahoo engineers · YahooCase Studies
  • Scalable IAM Architecture to Secure Access to 100 AWS Accounts at SegmentSegment engineers · SegmentCase Studies
  • Security Development Lifecycle at SlackSlack engineers · SlackCase Studies
  • Centralized Application Logging at eBayeBay engineers · eBayCase Studies
  • Enterprise Key Management at SlackSlack engineers · SlackCase Studies

  Open this lesson with its quiz and exercise