Recent developments in low-level software security

An important objective for low-level software security research is to develop techniques that make it harder to launch attacks that exploit implementation details of the system under attack. Baltopoulos and Gordon have summarized this as the principle of source-based reasoning for security: security properties of a software system should follow from review of the source code and its source-level semantics, and should not depend on details of the compiler or execution platform. Whether the principle holds - or to what degree - for a particular system depends on the attacker model. If an attacker can only provide input to the program under attack, then the principle holds for any safe programming language. However, for more powerful attackers that can load new native machine code into the system, the principle of source-based reasoning typically breaks down completely. In this paper we discuss state-of-the-art approaches for securing code written in C-like languages for both attacker models discussed above, and we highlight some very recent developments in low-level software security that hold the promise to restore source-based reasoning even against attackers that can provide arbitrary machine code to be run in the same process as the program under attack.