Craig Gentry
The cryptographer who showed in 2009 that you can compute on encrypted data without decrypting it – a problem left open since 1978.
Fully homomorphic encryption was proposed by Rivest, Adleman and Dertouzos in 1978, and for thirty years nobody knew whether it could be built. Craig Gentry, a former lawyer, solved it in his 2009 Stanford thesis under Dan Boneh. His scheme rested on ideal lattices and a trick he called bootstrapping: ciphertexts accumulate noise with every operation, so the scheme evaluates its own decryption function homomorphically to clean the noise out before it overwhelms the result.
The first implementation was so slow it served mainly as an existence proof. Fifteen years and several generations of schemes later, the technique is in limited real use, and Gentry has gone on to contribute to areas such as program obfuscation. He spent many years at IBM Research.