A Novel ANN-Based Support Vector Machine for improving Classification Accuracy in Intrusion Detection Systems

Abstract

This work recommends a innovative hybrid context for intrusion discovery, specifically tailored to address the challenges posed by the composite and energetic landscape of cyber threats, utilizing the CICIDS2017 dataset, that affords a complete and realistic illustration of modern web transportation. This approach integrates the strengths of Artificial Neural Networks (ANN) and Support Vector Machines (SVM) to develop classification accurateness and toughness in identifying malicious activities within network traffic. The use of ANN in this framework serves a crucial role in feature learning and representation. By employing a multi-layered architecture, ANN can autonomously extract intricate features from raw network data, capturing both spatial and temporal dependencies that are indicative of various types of network intrusions. This adaptive feature extraction capability enables the system to adapt to evolving attack strategies and effectively differentiate between normal and anomalous network behavior. Complementing the feature learning aspect, Support Vector Machines (SVM) are employed for classification tasks, leveraging their ability to construct optimal hyper planes in high-dimensional feature spaces. SVMs excel in binary classification tasks, making them well-suited for distinguishing between normal and intrusive network traffic. By integrating SVM into the framework, the researcher aims to exploit its robustness and generalization capability to improve the general performing and reliability of this intrusion detecting method. Extensive experimentation and evaluation are handled using the CICIDS2017 dataset to assess the effectiveness of the recommended approach. Reasonable analyses are performed against traditional SVM and ANN classifiers, as well as other contemporary intrusion detecting methods, to benchmark the execution in terms of classification accuracy. The results demonstrate consistent improvements achieved by hybrid approach, highlighting its efficacy in detecting various types of network intrusions while minimizing false positives. This research work presents a comprehensive and effective framework for intrusion detection, leveraging the synergies between ANN and SVM to enhance classification accuracy to 97.20%. By advancing the high-tech in intrusion detecting techniques, this study contributes to strengthening network defense measures and mollifying the threats posed by progressing cyber coercions in modern computing environments.