What We’ve Learned From Billions of Security Reports
summary
Service processes billions of browser security reports per month and scales from ~10,000 reports/month to 16,000 reports per second peak. Ingestion runs on DigitalOcean, storage on Azure Table Storage, moved from per-node memcached to a central Redis stacking with counts to compress duplicates and cut Azure transactions to ~0.01 per report, then shifted JSON validation and per-user filtering into Cloudflare Workers at the edge. Aggregated CSP data uncovers a rogue Chrome extension and endpoint malware and informs takedowns.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:19Native Reporting
- 1:58Content Security Policy
- 4:05Reporting
- 8:18Infrastructure
- 14:36Edge state
- 15:33Massive load
- 16:01DDoS
- 17:44New Relic
- 18:58Traffic
- 21:21Report Token
- 25:54Report cost ratio
- 26:31Re Redis
- 28:49Becoming a commercial service
- 30:37Moving to the edge
- 33:56Reducing transactions
- 37:31A larger scale
- 42:57Crosssite scripting
- 43:44Malware on endpoint
- 47:25Questions
- 47:52Clarify Workers
related talks
51mWhat We’ve Learned From Billions of Security Reports
53mAdventures at HyperScale!
01h 27mSecuring at Scale Building Resilient Defences for a Growing Digital World
01h 00mHyper Speed: When Big Data Blooms
47mCloud Security at Scale and What it Means for Your Application
01h 00mBuilding Safe Systems at Scale - Lessons from Six Months at Yahoo
40mThe new reality of DDOS
20mSetting the standard in cloud security