Randomness is crucial for cryptography - without randomness there can be no secrecy. However, it is not clear how to obtain the random bits necessary in a secure manner. In particular, one problem is that in general, we can *not* rely on data sampled from nature to be distributed according to a "nice" distribution (such as the uniform distribution). In this talk, I will survey some of the approaches to solving this problem in theory and practice (with an emphasis on the former).

I will mention results from a CHES 2003 paper which is joint work with Ronen Shaltiel and Eran Tromer, a FOCS 2004 paper which is joint work with Russell Impagliazzo and Avi Wigderson, and a yet unpublished joint work with Shai Halevi.