Neural Network Framework based Malware Classification in Cybersecurity: A Hybrid Approach

Malware significantly endangers internet security. Criminals may get illegal activities access, expropriate confidential information, extort, and engage in further illicit activities with malware. Several criminal organizations execute targeted malware assaults. They may repurpose harmful code from their current malware copies to create more sophisticated variants that evade detection. Acquiring sufficient viral data for a custom deep learning model may be challenging. Regardless its enhancement, a distinctive deep learning model will exhibit inadequate results on testing data due to its inability to comprehend the intricate information inside the augmented data. The pre-trained model exceeds the unique model using both limited and supplemented data. Pre-trained algorithms are instructed to use augmented data to get more precise and varied information. Malware classification requires enhanced recall and precision. Our hybrid malware classification approach employs Deep Neural Networks (DNN), Recurrent Neural Networks (RNN), and Convolutional Neural Networks (CNN) to address data imbalance and scarcity. To optimize data use, we refine our pre-trained model using novel attributes. In a subsequent step, we evaluated the hybrid models using optimum weights to see which one performed best on the test data; this yielded an accuracy of 96.76% as well as a Recall, Precision, F1-Score and Specificity of 97%.