One of the private keys that secured the web was literally the number 4. Eight intermediate certificate authority keys issued by Indian CA eMudhra turned out to be single-byte values on the P-384 curve, recoverable in under a second from the public certificates, and the researcher who reported it proved it by signing his disclosure message with each one. The eMudhra CA compromise has now led to 16 revoked intermediate certificates, and eMudhra plans to leave its publicly trusted TLS business, according to BleepingComputer. Browsers blocked the damage within hours, and no evidence has emerged that anyone used the keys maliciously, but there is no way to know whether someone already recovered them, in Helme’s assessment.
How does a certificate authority end up with a private key that is just the number 4?
A certificate authority is supposed to guard its private keys inside a hardware security module, a tamper-resistant device designed to keep the key inside it. eMudhra changed HSM vendors in April 2025 and used a custom Python script to move private keys between the old and new modules, according to BleepingComputer, reporting on eMudhra’s full incident report of 10 November. A bug in that script reduced each of the eight ECDSA private keys to a single byte. The recovered values were 4, 56, 66, 115, 155, 156, 230 and 237, according to Scott Helme, who verified all 16 certificates against the public records himself. With only 256 possible values, the keys could be recovered in under a second on commodity hardware.
The eight affected intermediates were six TLS issuing CAs and two code-signing CAs, all created in April 2025 and all cross-signed by Sectigo’s AAA Certificate Services root in May 2025, according to Helme. That cross-signing matters, because it meant each weak intermediate also had a chain through Sectigo’s AAA Certificate Services root. The six TLS CAs covered DV, OV and EV certificates across two generations, and the two code-signing CAs could make software look legitimately signed. eMudhra’s own account, given on Mozilla’s public incident ticket, is that a bug in the key-ceremony script created a second, weak key pair on the HSM, and a human error led to the wrong keys being used in the certificate signing requests, according to Helme’s 20 October update.
Researcher Wayne reported the problem to Apple, Google, Microsoft, Mozilla and eMudhra on Saturday 17 October, and to prove control he signed his message “We have a problem here – Wayne” with each affected intermediate key, according to Helme and Cybersecurity News. The disclosure site he used, emudhra.we-have-a-problem.com, was set up by Mozilla’s Dennis Jackson. As Helme put it, nobody had to steal anything: the keys were effectively public since April 2025, which leaves a 17-month window in which anyone could have recovered them. It is not confirmed that anyone did, and it is not confirmed how the keys left eMudhra.
How it unfolded
Mozilla received the report at 20:06 UTC on Saturday 17 October, and by 22:59 UTC the same day its OneCRL block covered all Firefox users with standard configurations, under three hours, according to Mozilla’s Security Blog, which we read directly. Mozilla also released Firefox for Android 158.0.2 on 21 October to fix an issue where some clients were not applying the OneCRL block immediately after startup. Google pushed a CRLSet block the evening of 17 October and later confirmed the root would go into Chrome Root Store v43, according to The Hacker News. eMudhra revoked the eight intermediates on 18 October. On Tuesday 20 October, Mozilla published its disclosure and released Firefox 158.0.1 plus ESR 115.43.1 and 153.5.1 for users whose custom settings may have blocked OneCRL, and NSS 3.131.1, which removes the root certificate that signed the affected intermediates and protects other applications using Mozilla’s trust store.
| Date | What happened | Source |
|---|---|---|
| Sat 17 Oct, 20:06 UTC | Wayne reports 8 compromised intermediates to Apple, Google, Microsoft, Mozilla and eMudhra | Mozilla blog; Helme |
| Sat 17 Oct, evening | Chrome pushes CRLSet block | Helme; The Hacker News |
| Sat 17 Oct, 22:59 UTC | Firefox OneCRL updated for all standard users, under 3 hours | Mozilla blog |
| Sun 18 Oct, 09:24-09:38 UTC | eMudhra revokes the 8 ECDSA intermediates, reason code cACompromise | TheTechBag, from the incident report |
| Tue 20 Oct | Mozilla discloses; Firefox 158.0.1 and ESRs; NSS 3.131.1 removes the root; Go 1.27.3 and 1.26.10 on the URGENT track; Chrome confirms Root Store v43 | Mozilla blog; The Hacker News; IFIN |
| Sat 7 Nov | eMudhra revokes 8 more RSA intermediates, reason code cessationOfOperation | BleepingComputer |
| Tue 10 Nov | eMudhra publishes its full incident report on Mozilla’s bug tracker | BleepingComputer |
| Mon 16 Nov, 23:59 UTC | G1 issuance stops; Microsoft starts removing the 10 remaining eMudhra roots | TheTechBag; The Hacker News |
The patch wave spread fast. Go shipped the fix in Go 1.27.3 and 1.26.10 on 20 October on its URGENT track, reserved for threats to ecosystem integrity, with CVE-2026-111943 assigned for the Golang trust store, according to IFIN’s threat-intel aggregation. Electron 44.9.1, 43.8.1 and 42.11.16, Python certifi 2026.10.19, Rust webpki-roots 1.0.10, Homebrew ca-certificates and NSS 3.131.1 were all patched, according to the same aggregation. Node.js and Composer ca-bundle were still pending as of 20 October, and Microsoft Edge 155.0.4283.62 had not yet updated its CRLSet, according to IFIN, a single aggregator source, so those two items should be treated as the state on 20 October rather than today.
One fact limited the damage for browser users. No TLS certificate chaining to these CAs ever appeared in Certificate Transparency logs before Chrome’s block, according to The Hacker News, citing the Chrome Root Program, so no forged certificate could have been trusted by default in Chrome, where CT is enforced, or in Safari, which also requires it. About 30,800 unexpired certificates chained to the G3 root as of 21 October, per Chrome’s count of Certificate Transparency logs cited by Ostorlab, so the population at stake was real even though no forged certificate was seen.
Why the second batch of eight revocations matters more
The first eight revocations were a cryptographic failure. The second eight were a governance failure, and they arrived with a reversal. On 19 October, eMudhra had written on Mozilla’s bug tracker that the RSA issuers “went through a different path… and are therefore not impacted”, according to the Indian Express. On 7 November it revoked those same eight RSA intermediates anyway, bringing the total to 16. The RSA keys were not weakened, they are RSA-4096 and their signatures still verify, but eMudhra could not demonstrate they were handled in line with its own rules, because the same script had re-imported key material from host-process memory instead of keeping it inside the HSM, according to Abijita and Windows Forum. They were revoked with a cessation-of-operation reason code, not a key-compromise code.
The eight RSA intermediates were six TLS issuers and two code-signing issuers, named EM DV TLS CA G2B and G2C, EM OV TLS CA G2B and G2C, EM EV TLS CA G2B and G2C, and EM OV CS CA G2B and G2C, according to Dugganusa. Four RSA G2C issuers had issued 195 end-entity certificates, 130 from EM DV TLS CA G2C, 63 from EM OV TLS CA G2C and 2 from EM OV CS CA G2C, and 29 of those 195 were for eMudhra’s own domains, according to The Hacker News. Some of the affected RSA intermediates had been actively issuing certificates since May 2026, with the last certificate issued on 3 November 2026, and eMudhra says the affected customer certificates have been replaced or migrated, according to BleepingComputer.
That is the point that makes the eMudhra CA compromise different from a single bad key. One script produced two distinct failure modes: eight keys reduced to a single byte, and eight keys handled outside the hardware that was supposed to protect them. eMudhra’s report also records that a mismatch detected in June 2025 was an early warning sign of the faulty key-generation process, and it went undetected for 16 months until October 2026, according to BleepingComputer. eMudhra has commissioned an independent review by KPMG, with findings expected by 4 December 2026, and its full inventory of all its CA keys is due 30 November 2026, according to BleepingComputer and The Hacker News. eMudhra said it would stop issuing certificates under its remaining G1 hierarchy at 23:59 UTC on 16 November, and Microsoft said it would start removing the ten remaining eMudhra roots from Windows and Edge the same day, according to TheTechBag and The Hacker News. Apple is set to end trust in all G1 certificates on 30 November, according to Abijita, a single source, so treat that date as eMudhra’s stated plan rather than Apple’s confirmed schedule.
Am I affected by the eMudhra CA compromise, and what should I check?
If your browser is up to date, you are protected from these certificates. Mozilla blocked all eight within three hours of the 17 October report, and Chrome followed the same evening. No forged certificate appeared in Certificate Transparency logs before Chrome’s block, but the risk lives in clients that do not check revocation.
- Update Chrome and Firefox to the latest version, and let your operating system install its own root-store updates.
- Check any site certificate issued by eMudhra or emSign and replace it with one from another provider.
- Update Go to 1.27.3 or 1.26.10, update Python certifi to 2026.10.19, and rebuild container images.
- Rebuild Electron apps on 44.9.1, 43.8.1 or 42.11.16, and update Rust webpki-roots to version 1.0.10.
- Check Windows machines for software signed by the EM OV CS code-signing intermediates and treat it as untrusted.
- Audit IoT and embedded devices that cannot be updated, because frozen firmware may keep trusting these certificates.
Helme demonstrated the non-browser risk first-hand on his own machines, and we did not independently reproduce his test. In his test setup against a local server, default curl accepted a forged api.stripe.com certificate, Python completed the handshake, and Node reported the connection as authorised. That establishes what the weak keys make possible, not that it happened to anyone. Windows trusts the eMudhra code-signing root directly, so malware signed with the EM OV CS intermediates would look like signed software from a real publisher to anything that does not check revocation, in Helme’s analysis, and that is a possibility to guard against, not an observed attack. Helme’s analysis also warns that firmware-frozen devices may trust the certificates until they expire in July 2039.
What is still unknown
The largest open question is whether anyone besides Wayne recovered the eight weak keys during the 17 months they were effectively public. eMudhra says it has found no evidence that the compromised keys were used maliciously, according to BleepingComputer, but absence of evidence in Certificate Transparency logs only covers browsers that enforce CT. A forged certificate used against a curl script, a CI runner or an IoT device would not be recorded there, because those clients do not check CT. KPMG’s independent review, due 4 December 2026, is the next date that could narrow that gap, and eMudhra’s full inventory of its CA keys is due 30 November. Both dates fall after the 16 November G1 shutdown, so the public record is still catching up with the incident.
One script, written to move keys between two makes of hardware security module, produced two different failures, a missed warning in June 2025, and the planned end of a certificate authority’s publicly trusted business. The browsers did their part in under three hours. The long tail is everything that is not a browser, and in Helme’s analysis it runs until the certificates expire in July 2039.


