Optimizely compromised, served malware on Zoom, GitLab

A widespread compromise of Optimizely’s JavaScript tags briefly injected malware into major sites like Zoom, GitLab, CNN and BBC, skimming crypto and credit-card data and, on some iOS versions, attempting device exploitation. Commenters highlight how loading dynamic third‑party scripts undermines web security, debate the limits of protections like Subresource Integrity and Content Security Policy, and argue that organizations should self-host and strictly vet any code they ship to users’ browsers.

Scope and Mechanics of the Optimizely Compromise

  • Attackers gained write access to JavaScript snippets used by many Optimizely customers, including high‑traffic sites.
  • Likely vector: central snippet or project shared across tenants, not per‑customer isolation. CDN simply served what it was given; not cache poisoning.
  • Reported impacts include: form skimming (credit cards, crypto addresses) and delivery of Coruna/DarkSword iOS exploit kits to some iOS versions.
  • Some surprise that only a small amount of crypto (~$7k) was stolen given the potential reach.

Third‑Party JavaScript and Tag Managers

  • Strong criticism of loading third‑party JS (analytics, A/B testing, adtech) directly in production pages.
  • Google Tag Manager and similar tools seen as letting marketing bypass engineering review, leading to security risks and multi‑MB JS payloads.
  • Client‑side A/B testing platforms (Optimizely, VWO, etc.) viewed as particularly dangerous: compromise of one account or engine can affect millions of users.

Defensive Techniques: CSP, SRI, and Script Blocking

  • Subresource Integrity (SRI) praised for detecting unexpected script changes, but:
    • Limited when scripts change frequently by design (as with SaaS tags).
    • Some argue it’s less compelling now that browser caches are partitioned and self‑hosting is common.
  • Content Security Policy (CSP) heavily endorsed:
    • Whitelisting script domains and avoiding unsafe-eval seen as crucial.
    • Others note CSP becomes nearly unmanageable when marketing constantly adds new trackers and regional domains.
  • Browser‑side tools (NoScript, uMatrix, uBlock Origin, hosts‑file blocking) help, but:
    • Managing allowlists is tedious, and many sites break without broad enables.
    • Some sites proxy third‑party scripts through their own domains, reducing effectiveness.

Accountability, Incentives, and Regulation

  • Many blame misaligned incentives: serving malware usually yields only a PR “whoops” and little financial consequence.
  • Proposals range from stricter civil liability to making serving malware (even via third parties) potentially criminal, with debate over how that would affect open source and small maintainers.
  • Some argue large sites should vendor and audit all third‑party scripts, with controlled rollouts.

Broader Reflections: AI, the Web Model, and Article Quality

  • Some see AI as lowering the bar for attackers by automating vulnerability hunting.
  • Others criticize the entire “code in the browser” model and advocate minimal or no client‑side scripting.
  • The incident writeup itself is debated: some call it product marketing or AI‑like “slop”; others find its CSP discussion particularly useful.