title:
Measuring information flow due to branching behaviour in programs



abstract:
Analysing the quantity of information flow due to branching behaviour in
an imperative program involves the use of non-trivial information
theoretic results. In a case of "something old, something new and
something borrowed", we show how the Fano inequality can be used to
bound flow caused by evaluating an equality test while the L1 inequality
can be used in bounding the flow into a program variable as a result of
the execution of an if statement.  Clearly the Fano and L1 inequalities
are borrowed, the use of the Fano inequality is an early result and the
use of the L1 inequality a recent one.

The bounds produced are normally used in the context of a syntax
directed program analysis of while language programs. The analysis can
establish upper and lower bounds on the leakage of confidential inputs
to non-confidential variables at the conclusion of the program. We
sketch the rest of the analysis and provide some examples using programs
containing if statements to illustrate the use of the two information
theory results.