IND-CPA PKE ⇒ IND-CCA KEM (Fujisaki–Okamoto)
Statement
Let be a PKE scheme with message space that is -correct for negligible (the decryption-failure probability, maximized over and then averaged over , is at most ), and let and be random oracles. The Fujisaki–Okamoto KEM encapsulates by sampling , encrypting it as and outputting the key ; decapsulation decrypts to , re-encrypts under , and returns if the result equals , and otherwise rejects, explicitly with or implicitly with for a uniform seed kept in the secret key. If is IND-CPA-secure, or merely one-way, the KEM is IND-CCA-secure in the random-oracle model; explicit rejection additionally needs to have -spread ciphertexts. FO99 introduced the transform, for hybrid encryption with a symmetric scheme; HHK17 decompose it into a derandomization step and a key-derivation step, covering explicit and implicit rejection, with concrete bounds for a -correct base scheme.
Sketch
Derandomizing encryption with makes ciphertexts checkable by re-encryption, so the reduction answers decapsulation queries by searching the adversary’s and queries for a message that re-encrypts to the queried ciphertext. An adversary that never queries at the challenge message has no information on ; from one that does, the reduction reads off , inverting the base encryption.
Notes
- KEM-specific versions of the transform, proved IND-CCA in the random-oracle model from a one-way PKE — Den03.