It is known that homomorphic encryption schemes are not secure against CCA attacks. This is because in the security game, the attacker may use homomorphic evaluation to cheat in the the challenges and ask for a decryption of a (homomorphically equivalent of the) challenge itself.
However, I found a key-recovery attack. The attacker adaptively asks for decryption of ciphertexts, and at the end she knows the secret key.
Is this of some actual value? Or, just because the scheme is known to be CCA insecure, the hypothesis of an attacker asking for decryptions is no longer valid?
Or else, maybe I am not aware of generic key-recovery attacks using CCA semantic security attacks? If this is the case, please point out a reference.