Enhanced Data Race Detection Through Dynamic Control Flow Analysis for Aspect-Oriented Program Using Ss-Lsgru and G-Csoa

Abstract

Aspect Oriented Programming (AOP) language is a high-level object-oriented language, which is widely used for generating web applications. As AOP is designed in a multi-threaded manner, the data race occurs in a program when different threads access the shared memory resource. But, none of the existing works handle the dynamic control flow of AOP, resulting in higher levels of false positives and false negatives in Data Race Detection (DRD). Therefore, in this work, an efficient framework is proposed for detecting the data race of Aspect Oriented Programming (AOP) language using Eisen Cosine Correlation distance based Entropy Variance KMeans (ECC-EVKMeans), SoftSwish – Linear Scaling Gated Recurrent Unit (SS-LSGRU), and Kullback Leibler -based Fuzzy Bayesian Inference System (KL-FBIS). Primarily, the proposed system acquires an AOP of various applications; then, its number of variables along with the methods are extracted. Further, the context-sensitive information of code is analyzed in the thread analysis phase using the control flow graph. Subsequently, the dynamic scope of pointers is evaluated in escape analysis, and the single parameterized analysis of each method is attained by the compositional pointers. Meanwhile, the test cases are generated, and the significant test cases are selected and clustered using G-CSOA and ECC-EVKMeans, respectively. Then, by using the BERT algorithm, the vector values of corresponding words in test cases are returned. Further, the vector values are separated using the KL-FBIS approach to minimize the nested loops. Eventually, the vector values are trained using the SS-LSGRU classifier and also tested with real-time AOP for detecting the race condition. The experimental results show that the proposed system detects data race with 98.26% accuracy and 98.95% precision in 37751ms. Also, the important test cases are selected with 97.85% fitness by using the proposed technique.