On the security of quantum cryptographic keys

The security of quantum key distribution protocols is often defined in terms of the information an adversary obtains by measuring his system. Such definitions are fundamentally flawed because of a locking property of the accessible information: Giving the adversary a single bit of information may increase the accessible information by more than one bit. We give examples of keys that are not exposure-resilient and can thus not safely be used for one-time pad encryption, even though they satisfy a measurement-based security definition. In the second part of the talk, we discuss a universally composable security definition for cryptographic keys and show how this stronger type of security can be achieved.<br/> This is joint work with Andor Bariska, Ueli Maurer and Renato Renner.