Description
A mis-handling of invalid unicode characters in the Java implementation of Tink versions prior to 1.5 allows an attacker to change the ID part of a ciphertext, which result in the creation of a second ciphertext that can decrypt to the same plaintext. This can be a problem with encrypting deterministic AEAD with a single key, and rely on a unique ciphertext-per-plaintext.
Remediation
References
https://github.com/google/tink/commit/93d839a5865b9d950dffdc9d0bc99b71280a8899
https://github.com/google/tink/security/advisories/GHSA-g5vf-v6wf-7w2r