Secure slices of insecure programs

摘要: This paper deals with the problem of protect confidentiality data manipulated by sequential programs. In this context, secure information flow refers to guarantee that program executions are free non authorized flows secret into public channels. There two established means enforce policies: static analyses, performed at compile time and all unauthorized flows; runtime monitoring, dynamically detects neutralizes invalid for current run.Both approaches have their advantages disadvantages. The main disadvantages control (IFC) is, it does not differentiate between insecure same program, therefore whole programs rejected in presence possible flows. On contrary, dynamic IFC rejects only. analysis precision comes price execution overload imposes tracking flow.This work presents slicing, a technique statically transforms probably (interfering) (non-interfering) ones. Our approach combines transformation: if detected, instead rejecting we transform eliminate way, alleviate drawbacks approaches: neither reject full nor impose run-time overhead. resulting can be seen as slice source executed without risk leaks.In also show slices computed intentionally release information, applied real programming languages such Java.