Research On Technologies Of Control Flow Reconstruction And Control Structure Recovery In Decompilation

Decompilation always is one of research hotspots in the field of code reverse analysis, and plays an important role in understanding and maintenance of software. Along with the broad application of embedded system in key electrical devices, and the prominent effection of code reverse analysis in guaranteeing safety and usability of system, research on decompilation of embedded system code plays an important significance.The dissertation explores and studies technologies of control flow reconstruction and analysis in decompilation.Aiming at the problems depending on specific compiler mode and control transfer implementation in classical approaches of control flow reconstruction, an abstract interpretation-based framework for control flow reconstruction is designed based on the intermediate representation BRIL. The problem of indirect transfer target identification can be solved via executing control flow reconstruction and data flow analysis simultaneously under the framework. A high-precision data flow abstract domain is constructed to instantiate the framework. The strategy of"variable value restriction"is presented to solve path explosion in iterative calculation. And that a control flow reconstruction algorithm for the framework is designed.In control flow analysis, structure semantic tree is put forward to express control structures and relationships among them. A control flow graph structuring algorithm is devised based on the strategy of"three chances"to solve the unstandardized control structure identification. And a structure semantic tree-based framework for high-level control structure recovery is designed, which solves the key problem of control structure nesting relation recovery.The dissertation implements control flow reconstruction module and high-level control structure recovery module in decompilation tool, and test results validate that the approaches above are correct and effective, which can create favourable conditions for the backward code analysis.