NESL Technical Report #: 2012-11-1
Abstract: Probabilistic computation is a convenient means of mechanically reasoning about a variety of information security problems. At its core, information security concerns itself with measuring or limiting the knowledge an adversary might attain from interacting with a protected system. Probabilistic inference lets one compute this knowledge explicitly as long as the interaction can be described as a program in a suitable probabilistic language that supports conditioning. Security concerns, however, require soundness guarantees on probabilistic inference which are generally not present in machine learning applications. We summarize some recent work on probabilistic computing for information security and highlight challenging aspects that still need to be addressed.
Publication Forum: NIPS Workshop on Probabilistic Programming
Page (Count): 3
Place: Lake Tahoe, CA
Public Document?: Yes
NESL Document?: Yes
Document category: Conference Paper