About this tracker
Faultline is a public map of cyber attacks and major service outages. It is built in the spirit of the early COVID dashboards: one page, big numbers, a map, and a source on every row. It is not a government alert system, and it does not show advertising.
What the numbers count
The headline figures use the incident start time in UTC. “24 hours”, “7 days”, and “30 days” are rolling windows. An incident counts once, not once per country. Cyber attacks and outages are separate. Unconfirmed reports are shown beside those totals and are not added into either one. “Countries named” counts countries a cited source actually names. A worldwide cloud outage stays worldwide unless the source lists countries. Counts use severity 3 and above, so a single-region latency blip does not move the headline.
“Older ongoing” on the map keeps disruptions that started outside the selected window when the latest public status still says they are unresolved. The AWS Middle East region events are the reason that switch exists: the public health feed still describes unrestored capacity from disruptions that began in early 2026.
Confirmed, suspected, unconfirmed
Every row carries a cause label. Editorially reviewed rows are the November 2026 incidents checked by hand against the linked pages. Everything else is automatically classified from a headline and a short feed excerpt, and the page says so. A vendor status page can confirm that an outage happened and still leave the cause unknown. A news report of a hack is labeled suspected unless the seeded write-up cites an organization or agency that confirmed it. If a company or official says an outage was a bug, a bad update, or maintenance, the row says technical cause and does not call it an attack.
How ingestion works
The server fetches the sources below about every ten minutes. The browser reloads the JSON about every 45 seconds, without a full page refresh. Items are deduplicated by source URL, by a hand-written match for the seeded events, and by similar titles within three days. Country pins come from place names and cloud-region identifiers in the text, using public geographic centroids. If a place cannot be read from the source, the row stays in the feed with no pin.
Vulnerability bulletins, arrest stories, product announcements, and explainers are dropped on purpose. CISA’s Known Exploited Vulnerabilities catalog is counted on the homepage and is not drawn as incidents. Last catalog check: not yet.
Sources
No paid API key is required. Optional keys are not used.
- AWS Service Health public events. UTF-16 JSON of current public events. No key.
- Google Cloud incidents. Official incident JSON, including region names.
- Azure Status RSS. Empty when Azure reports no public incident.
- Apple System Status. Official status JSON.
- Microsoft 365 consumer status. Public consumer card. Commercial tenant health is not included.
- Statuspage feeds. The same Atlassian Statuspage JSON shape is read for Cloudflare, GitHub, Atlassian, OpenAI, DigitalOcean, Zoom, Discord, and Twilio.
- BleepingComputer. Security news RSS, filtered to incident language.
- The Record. Security news RSS.
- The Hacker News. RSS. Vulnerability write-ups are filtered out.
- Krebs on Security. RSS.
- BBC technology. RSS. General tech stories are filtered out.
- UK NCSC. Official RSS. Blogs and guidance are filtered out.
- CERT-FR. Official RSS. Product vulnerabilities are filtered out.
- CERT-EU. Security advisories RSS, filtered to incident language.
- Canadian Centre for Cyber Security. Atom feed of alerts and advisories.
- CERT-Bund. German CERT advisory RSS, filtered.
- CISA advisories. Tried each pass. Often blocked with HTTP 403 from hosting networks.
- CISA KEV. Counted only. Not plotted.
- GDELT DOC 2.0. Free news index. No key. Rate limits apply. Articles still have to pass the incident filter.
| Source | Last attempt | Fetched | Last error |
|---|---|---|---|
| Apple System Status | 19 Nov 2026, 19:21 UTC | 4 | — |
| AWS Service Health | 19 Nov 2026, 19:21 UTC | 0 | This operation was aborted |
| Azure Status | 19 Nov 2026, 19:21 UTC | 0 | — |
| Cloudflare Status | 19 Nov 2026, 19:19 UTC | 49 | — |
| Google Cloud Status | 19 Nov 2026, 19:21 UTC | 0 | This operation was aborted |
| Microsoft 365 consumer status | 19 Nov 2026, 19:19 UTC | 0 | — |
Seeded November 2026 incidents
These were written from the public record so the map is not empty on first launch. Social posts that said “about ten countries” for the refrigeration attacks are not followed. The map includes only countries named by the cited articles: seven in the November wave, plus France for the October wave.
- FAA nationwide ground stop after air-traffic network instability
- Global Microsoft 365 outage
- JPMorgan Chase online banking disruption
- Apple services outage after an internal update
- NSA internal network disruption during maintenance
- Cyberattacks on supermarket refrigeration in seven European countries
- Cyberattacks on refrigeration controllers at French supermarkets
Limits
- Automatic classification will misread some headlines. The source link is the record.
- Microsoft 365 commercial service health is behind a tenant login. Consumer status and Azure’s public RSS are fetched instead, and the November outage is cited from news reports of Microsoft’s review.
- CISA’s advisory RSS often returns 403 to datacenter addresses. The KEV JSON file is fetched separately.
- GDELT’s document API is free and rate-limits aggressively. A failed pass is recorded and skipped.
- Downdetector, paid news wires, and social listening are not used.
- Swiss NCSC and ENISA did not expose a stable keyless incident feed at build time. The European Commission joint statement with ENISA is cited for the refrigeration campaign.
- This site stores short excerpts and titles, not full articles.