
On April 30, 2025, eMudhra Limited held a key generation ceremony. It was supposed to be routine. The company was creating a batch of 24 intermediate Certificate Authorities (CAs) for its global TLS business, using Utimaco hardware security modules (HSMs) that meet FIPS 140-2 Level 3 standards. These intermediates sit between the root CA and the actual certificates websites use. If an intermediate key is compromised, an attacker can issue fake certificates for any website. Other
The ceremony followed the standard script, but with an approved deviation. Because the B and C series keys were intended for export to US and European data centres, the team added a test for key wrapping and unwrapping functionality. That's where things broke. A software bug in the automation script triggered the generation of a second set of keys with truncated entropy. Instead of the full 384-bit randomness required for the P-384 elliptic curve, these keys were tiny integers—all below 256.
Here's the kicker: the script assigned overlapping labels to the keys in the HSM storage. When it came time to generate the certificate signing requests, the script referenced the second, weakened set of keys, not the original strong ones. Those weak keys were then signed by eMudhra's root CA and cross-signed by Sectigo, extending trust to the resulting intermediates.
A P-384 private key should be a massive number, essentially impossible to guess. The eight affected keys were anything but. One of them, for the EM OV TLS CA - G3C intermediate, had a private key value of literally 4. Others were similarly small—56, 66, 115, 155, 156, 230, and 237. Security researcher Scott Helme put it bluntly: the entire security model rested on eight bits. Anyone could recover these private keys in milliseconds using simple math.
The flaw wasn't in the signatures themselves. It was in the key generation. Mozilla's security engineers confirmed they could recover the keys independently, without any tips from the reporter, proving the generation process was fundamentally broken.
Why didn't quality control catch this? That's the uncomfortable part. The keys passed independent audits for two years and had been in browser trust lists since mid-2025. There was even an earlier warning sign. In May 2025, Mozilla flagged that eMudhra was publishing Certificate Revocation Lists (CRLs) with invalid signatures for these exact intermediates. The fix at the time was to re-sign the CRLs with the weak keys so the signatures would validate. That resolved the error but masked the real problem: the public certificates in production were tied to broken keys. Other
On October 17, 2026, an independent researcher reported the issue to Mozilla, Google, Apple, and Microsoft, providing cryptographic proof of key control for all eight certificates. Mozilla acknowledged the report within minutes and began the process to distrust the certificates in Firefox via OneCRL, its centralized revocation list. By 22:59 UTC that same day—under three hours after the report—the block was deployed.
Google Chrome followed a similar path using CRLSets, its proprietary revocation mechanism. Apple and Microsoft acted shortly after, with Apple blocking from October 19 and Microsoft from October 20. The speed was impressive, but it masked a deeper issue: these revocation mechanisms only protect clients that consume them. Most non-browser clients do not.
Here's where things get messy. Firefox, Chrome, Safari, and Edge are protected. But what about /usr/bin/curl on macOS? Or Python scripts using the certifi bundle? Or Go applications compiled before the fix? Or container images frozen at build time?
These systems verify certificates against static bundles of trusted CAs. They don't check revocation lists in real-time. Until the bundle itself is replaced through a software update, they keep trusting the compromised intermediates. Security analysts have pointed out that macOS's built-in curl utility, in particular, may continue trusting the weak certificates because it verifies against a static /etc/ssl/cert.pem file that doesn't check revocation.
The patching landscape is uneven. Mozilla's NSS 3.131.1 update, which removes trust in the emSign ECC Root CA - G3, has been packaged by some Linux distributions quickly. Arch Linux and Alpine Linux show updated versions, but Debian's ca-certificates package was listed as pending in the latest available report. Language runtimes are on their own timelines. Go released versions 1.27.3 and 1.26.10 on October 19, hard-coding blocks for the eight compromised SubjectPublicKeyInfo hashes. Python's certifi, Rust's webpki-roots, and Electron have also issued updates. Others are still catching up.
For eMudhra Limited, the direct financial hit looks manageable. The company says the affected global TLS business represents about 1% of total revenue. For context, Q1 FY27 consolidated revenue was ₹192.46 crore, so 1% is roughly ₹1.9 crore for the quarter [stock_agent]. The India DSC and eSign businesses are unaffected because they run on separate infrastructure. Other +1
But the market reaction suggests investors see deeper risks. The stock was trading around ₹540 on October 19. By October 21, the live feed showed a price of ₹432, down about 20% [stock_agent]. That's a far larger drop than the 1% revenue exposure would justify, implying the market is pricing in reputational damage, potential liability, and the risk of further browser distrust.
Mozilla has asked eMudhra to stop issuing certificates from all intermediates chaining to the G3 root and to migrate subscribers to the G1 RSA root. eMudhra estimates about 30,000 certificates require migration, with roughly 97% coming through ACME (automated issuance). The company is working directly with affected customers to issue replacements. Other
The cross-signing relationship with Sectigo amplifies the exposure. Sectigo had cross-signed some of the affected certificates, creating a second trust path through the AAA Certificate Services root. That means there are two sets of affected certificates—eMudhra-issued and Sectigo cross-signed—widening the blast radius for revocation and migration.
Public web trust isn't governed by a regulator with fining powers. It's governed by root program policies and the CA/Browser Forum's Baseline Requirements. Mozilla's policy allows it to require additional audits or remediation plans and to refuse inclusion for CAs that cause undue risk. The company has already taken the nuclear option: emergency removal of trust in the G3 root.
The CA/B Forum's requirements mandate that CAs prepare and follow a key generation script and have a Qualified Auditor witness the process for root and certain subordinate keys. However, for a CA's own subordinates, auditor witnessing is only a "should," not a "must." That's a potential gap this incident may force the industry to close.
For organizations that relied on eMudhra-issued certificates, the immediate impact is operational. Certificates chaining to the distrusted G3 root will cause browser errors and failed connections. Whether this triggers compliance issues under standards like PCI-DSS or GDPR depends on the specific control requirements and whether any data was actually intercepted. There's no evidence so far that the weak keys were exploited, but the exposure window—from April 2025 to October 2026—is long.
This incident is a wake-up call for the entire PKI ecosystem. The revocation machinery works for browsers, but it's a patchwork for everything else. Security leaders need to inventory which trust stores their applications rely on and how those stores receive updates.
For eMudhra, the immediate priority is completing the migration to the G1 root and answering Mozilla's outstanding questions, particularly about the 2025 CRL signature bug. The company says the root cause is still under investigation. Other
For the industry, the debate will likely turn to standards. Should auditor witnessing be mandatory for all CA key generation ceremonies? Should automated post-generation checks be required to verify key entropy? Should unexplained CRL signature failures trigger immediate root-cause analysis? These are questions the CA/Browser Forum and root programs will need to address.
The padlock in your browser is only as strong as the list behind it. This time, the list was updated fast. Next time, the systems that don't auto-update might not be so lucky.