Unwinding Conditions for Security in Imperative Languages


We study unwinding conditions for the definition of non-interference properties of a simple imperative language, admitting parallel executions on a shared memory. We present different classes of programs obtained by instantiating a general unwinding framework and show that all the programs in these classes satisfy the non-interference principle. Moreover, we introduce a subclass of secure programs which is compositional with respect to the language constructors and we discuss verification techniques.