We present two efficient constructions aimed at making public key systems secure against chosen ciphertext attacks. The first one applies to any deterministic public key system and modifies it into a system that is provably as hard to break under a passive attack as the original one, but has the potential of making a chosen ciphertext attack useless to an enemy. The second construction applies to the El Gamal/Diffie-Hellman public key system. Again, the modified system is provably as hard to break under a passive attack as the original one, and under an additional cryptographic assumption, a chosen ciphertext attack is provably useless to an enemy. We also point out a connection between such public-key systems and efficient identification schemes.