Cryptography requires secret keys, and we usually assume that these can be generated uniformly at random and kept perfectly secret from an attacker. Nevertheless, there are many scenarios where we need to rely on "imperfect" secret keys for cryptography. This includes passwords and biometrics, which are not uniformly random, but come from some unspecified distribution about which we know very little. It also includes scenarios where partial information about the secret key can leak to an attacker, for example, through various "side-channel attacks". Such attacks are being used to break real-world implementations of current cryptosystems, and are the major source of difficulty when implementing cryptography in practice. My talk will explore new methods for constructing cryptosystems that remain provably secure without requiring the perfect secrecy of their secret keys.
Daniel Wichs is a postdoctoral researcher in the cryptography group at the IBM Research -- T.J. Watson Center, under the Josef Raviv memorial fellowship. In September 2011, he received his Ph.D. from New York University supervised by Yevgeniy Dodis. He received his BS and MS degrees from Stanford University in 2005. Daniel is interested in all aspects of modern cryptography and its applications to information security. He is also interested the various connections between cryptography and other areas of theoretical computer science, including complexity theory, coding theory, information theory and algorithms.