I saw a report that re-encoding an intermediate certificate can bypass Java's fingerprint blocklist. What does that mean for patched JDKs, and what are the options for reducing the risk?
How does the Java certificate blocklist bypass work?
Answers
A certificate fingerprint hashes the encoded certificate bytes. If those bytes change, the fingerprint changes too, even if the resulting certificate can still validate. The reported tests reproduced that behavior with both ECDSA and RSA intermediates under the tested Java trust-store configurations.
That shows a limitation of blocking a particular fingerprint. It doesn't by itself establish that every patched JDK or trust-store configuration is bypassable.
Reply to this answer
The patched JDK's actual blocked.certs file is the thing to check. One technical objection is that the shipped file contains multiple fingerprints for each affected certificate, covering variations such as the presence or absence of a NULL parameter and ECDSA signature alternatives.
A reproduction using a file with only the original fingerprint would not settle whether those shipped entries can be bypassed. The available discussion raises that question but doesn't resolve it.
Reply to this answer
Removing the relevant roots from the trust store is a broader option. In the reported tests, it rejected both the original and re-encoded chains. The tradeoff is that legitimate connections relying on those roots may fail too, so check which applications and certificates depend on them before making that change.
Reply to this answer
Another approach mentioned in the discussion is blocking a SHA-256 hash of the certificate's subject public key information, rather than the fingerprint of the whole certificate. That's a different target, so changing the certificate encoding alone isn't necessarily the same kind of bypass. It shouldn't be assumed to solve every related trust problem, though.
Reply to this answer
The report says there was no evidence of this technique being used in the wild. That's not proof it couldn't be used, but it's worth separating the demonstrated behavior from claims about actual attacks.
Reply to this answer
Add your answer
Answers are reviewed before publication.
Vivi:
The result can also depend on which trust store an application actually uses, so changing the default store alone may not cover every configuration.