Leveraging abstract interpretation for efficient dynamic symbolic execution

Dynamic Symbolic Execution (DSE) is a technique to automatically generate test inputs by executing a program with concrete and symbolic values simultaneously. A key challenge in DSE is scalability; executing all feasible program paths is not possible, owing to the potentially exponential or infinite number of paths. Loops are a main source of path explosion, in particular where the number of iterations depends on a program's input. Problems arise because DSE maintains symbolic values that capture only the dependencies on symbolic inputs. This ignores control dependencies, including loop dependencies that depend indirectly on the inputs. We propose a method to increase the coverage achieved by DSE in the presence of input-data dependent loops and loop dependent branches. We combine DSE with abstract interpretation to find indirect control dependencies, including loop and branch indirect dependencies. Preliminary results show that this results in better coverage, within considerably less time compared to standard DSE