A source code oriented static detection method for integer overflow defects

Limited by incomplete call graph analysis and path feasibility analysis, current static integer overflow defect detection methods generally return results with high false positives. To reduce this inefficiency, aiming at automatic exploration of the external input triggering integer overflow defects, a new source code oriented detection method was proposed combining call graph analysis, static taint analysis and static symbolic execution, in which a fieldsensitive and flowsensitive pointer analysis method was proposed for constructing an over-approximation of the target program’s real call graph, with a static taint-sink propagation analysis carried out for calculating the potential external input reachable integer overflow defects, on which flowsensitive static symbolic execution is conducted to reduce the false positives introduced by the detection system through justifying the satisfiability of the corresponding defect constraint. Experiments prove the effectiveness of the methodin real-world integer overflow defect detection and false alarm reduction.