Abstract Code Injection: A Semantic Approach Based on Abstract Non-Interference

Code injection attacks have been the most critical security risks for almost a decade. These attacks are due to an interference between an untrusted input (potentially controlled by an attacker) and the execution of a string-to-code statement, interpreting as code its parameter. In this paper, we provide a semantic-based model for code injection parametric on what the programmer considers safe behaviors. In particular, we provide a general (abstract) non-interference-based framework for abstract code injection policies, i.e., policies characterizing safety against code injection w.r.t. a given specification of safe behaviors. We expect the new semantic perspective on code injection to provide a deeper knowledge on the nature itself of this security threat. Moreover, we devise a mechanism for enforcing (abstract) code injection policies, soundly detecting attacks, i.e., avoiding false negatives.