| Vulnerability mining techniques for binary programs are at the core of research in the field of network security,encompassing various application scenarios such as software vulnerability scanning,malware identification,and 0-day vulnerability mining.In binary program vulnerability mining techniques,the study of static vulnerability mining methods,especially methods for mining vulnerability points without executing the program,is of great significance.However,traditional static vulnerability mining methods often suffer from poor generality,high falsepositive rates,and dependence on manual modeling.Therefore,improving the accuracy,generality,and generalization ability of static vulnerability prediction models has become a key technical challenge in the field of static vulnerability mining.To address these issues,this paper proposes a complete static prediction algorithm for binary program vulnerability types based on decompilation,control flow graphs,and intermediate code representations of binary programs.By leveraging graph embedding network algorithms to sample node and path information in control flow graphs,feature vectors of functions in binary programs are extracted.Then,a neural network is built to predict vulnerability features in the feature matrix of the program.Experiments using Juliet,LAVA-M,and Real-World datasets for model training show that the proposed algorithm has high accuracy and practicality in static vulnerability mining.Based on the graph embedding network-based binary program vulnerability static prediction algorithm,this paper designs and implements a prototype system for binary program vulnerability static prediction.The system mainly includes the following components:(1)Input module for binary programs,machine instruction parsing,and intermediate representation;(2)Extended control flow graph construction and graph feature matrix extraction module for the program;(3)Neural network module for vulnerability point prediction based on graph feature matrix;(4)Visualization UI interface for algorithm output.Experimental results show that the proposed algorithm is feasible and has superior accuracy in vulnerability prediction compared to other existing static vulnerability mining models and schemes.In addition,the designed and implemented system meets the functional requirements and has advantages in terms of accuracy and performance compared to other similar vulnerability mining tools. |
