The incident report says an import defect truncated private scalars for eight ECDSA P-384 objects. How does that make the keys recoverable from their public keys, and what checks might have caught it?
How did the eMudhra key import bug make P-384 keys recoverable?
Answers
An ECDSA public key is derived from a private scalar. P-384 relies on that scalar being chosen from a huge range, so working backwards from the public key is normally infeasible.
The reported defect truncated the private value to its leading byte. That leaves a tiny range of possible scalar values. Someone with the public key can derive a public key for each candidate and check for a match. The curve itself wasn't broken. The private values ended up far smaller than intended.
Reply to this answer
The ceremony check that stands out is comparing the certified public key with the public key from the originally generated key. The report says that comparison wasn't done, and it would have caught the wrong object being used.
That's separate from asking whether the private key is strong. A normal-looking public key doesn't by itself prove that its private scalar was generated correctly. This particular truncation leaves such a small candidate range that a targeted check could find it, but checking for every possible key-generation failure is a different problem.
Reply to this answer
The report points to a read-back from the unwrapped object during import, not a general problem with ECDSA P-384. A wrong buffer length or array access might explain a one-byte result, but those are guesses. The description supplied doesn't establish the exact line of code.
Reply to this answer
The scope stated in the report is eight P-384 objects, identified as G3B and G3C. That doesn't establish that every key handled by the same system was affected. The useful question for an operator is whether a particular certificate contains one of the affected public keys, rather than assuming all P-384 certificates are weak.
Reply to this answer
Add your answer
Answers are reviewed before publication.
Mivluno+5d:
So the public key alone is enough to check each candidate, without needing a signature?
q.265+veltruni:
Right. The public key gives you the value to compare against after deriving a public key from each candidate scalar.