The incident report https://bugzilla.mozilla.org/show_bug.cgi?id=2083449#c82 confirms a long held suspicion: HSMs are not what they're supposed to be.
The key phrase is:
Decision (mid‑Apr 2025) to generate the B/C keys on the HSMs available before the June 2025 root store policy changes and to transfer them by key wrapping.
What does "key wrapping" mean?
Key wrapping is a cryptographic technique used to securely transfer or store encryption keys by encrypting them with another key, known as a Key Encryption Key (KEK). In the eMudhra incident, a custom Python script was used to migrate private CA keys between HSMs from different manufacturers, exposing the keys in plaintext in host memory. A programming error during this process reduced some ECDSA private keys to just 8 bits of entropy, making them trivially recoverable.
This says for me clearly: The HSM is just used as smoke and mirrors. It is possible to export the supposedly securely stored private key. From this point on, the HSM becomes just a bullet point in a marketing slide.
It has lost it's raison d'être.
And from how the words are chosen, this doesn't seem to be an uncommon procedure.
The questions raised by @fluepke are very valid ones: https://chaos.social/@fluepke/117594259242217211
I am angry about the "final" incident report, that #eMudhra has published:
https://bugzilla.mozilla.org/show_bug.cgi?id=2083449#c82
It raises questions:
- How the f*ck was the wrapped key "unwrapped" on the host? Exploit or wrong / insecure HSM usage?
- The RSA keys were also compromised, as they left the HSM and were present in plaintext within the host RAM?
- Can, we finally, get to see the damn source code used in the KGC??
- Why where the RSA certs only revoked on 7th Nov??
The incident seems much worse than just the 4.
