We address the problem of computing the capacity of a covert channel, modeled as a
nondeterministic transducer. We give three possible statements of the notion of
“covert channel capacity” and relate the different definitions.
We then provide several methods
allowing the computation of lower and upper bounds for the capacity of a channel.
We show that, in some cases, including the case of input-deterministic channels,
the capacity of the channel can be computed exactly
(e.g. in the form of “the largest root of some polynomial”).
