Robust Malicious Query Access Detection using GCNN-Attention and BiLSTM with DFA Analysis

Abstract

Language technology resource repositories are increasingly vulnerable to illicit query access in the current digital environment, which poses major hazards to intellectual property and data security. In order to efficiently identify unusual query behavior, this study suggests a hybrid deep learning architecture that combines Graph Convolutional Neural Network (GCNN) with Attention mechanism and Bidirectional Long Short-Term Memory (BiLSTM). TF-IDF feature extraction is used to alter textual queries, allowing for effective representation for categorization. Detrended Fluctuation Analysis (DFA) is used to differentiate between benign and malevolent access patterns in order to further capture long-range behavioral patterns. According to experimental data, the suggested model outperforms conventional machine learning techniques like SVM and Naïve Bayes, with an Language technology resource repositories are increasingly vulnerable to illicit query access in the current digital environment, which poses major hazards to intellectual property and data security. In order to efficiently identify unusual query behavior, this study suggests a hybrid deep learning architecture that combines Graph Convolutional Neural Network (GCNN) with Attention mechanism and Bidirectional Long Short-Term Memory (BiLSTM). TF-IDF feature extraction is used to alter textual queries, allowing for effective representation for categorization. Detrended Fluctuation Analysis (DFA) is used to differentiate between benign and malevolent access patterns in order to further capture long-range behavioral patterns. According to experimental data, the suggested model outperforms conventional machine learning techniques like SVM and Naïve Bayes, with an