Limits on Adaptive Security for Attribute-Based Encryption

This work addresses the long quest for proving full (adaptive) security for attribute-based encryption (ABE). We show that in order to prove full security in a black-box manner, the scheme must be “irregular” in the sense that it is impossible to “validate” secret keys to ascertain consistent decryption of ciphertexts. This extends a result of Lewko and Waters (Eurocrypt 2014) that was only applicable to straight-line proofs (without rewinding). Our work, therefore, establishes that it is impossible to circumvent the irregularity property using creative proof techniques, so long as the adversary is used in a black-box manner. As a consequence, our work provides an explanation as to why some lattice-based ABE schemes cannot be proven fully secure, even though no known adaptive attacks exist.