Many information measures compete for our attention when we want to decide whether one program leaks no more information than another: an early favourite was Shannon Entropy, of course. But more recent entries are Bayes Vulnerability, Marginal Guesswork, Guessing Entropy... and the field has become rather full. It's a bit of a zoo. All of the measures have their own advantages and disadvantages, their adherents and detractors. But do any of them have mathematical properties relevant to specification and refinement? Do they admit a program algebra? Do they allow compositional reasoning?
展开▼
