@article{4045,
  author = {Mohsin Ali, Abdul Razaque, Damelya M. Yeskendirova, Talgat A. Nurlybayev, Nessibeli Y. Askarbekova and Zarina A. Kashaganova},
  title = {Impact of Neural Networks on Improving Cloud Computing Security with AI-powered Smart Intrusion Detection },
  journal = {Journal of Information Security Research},
  year = {2024},
  volume = {15},
  number = {2},
  doi = {https://doi.org/10.6025/jisr/2024/15/2/59-71},
  url = {https://www.dline.info/jisr/fulltext/v15n2/jisrv15n2_2.pdf},
  abstract = {This work introduces a highly effective method for identifying harmful network activity by leveraging the power of artificial neural networks. These networks, particularly adept at analyzing deep packet inspection-based systems for detecting intrusions, have been the cornerstone of our research. The method was rigorously tested using a diverse range of harmless network data, including images, dynamic link library files, logs, music files, word processing documents, and malicious shell code files from the exploit and vulnerability database, exploitdb. The proposed artificial neural network design successfully distinguished between harmless and harmful network traffic. The developed neural network consistently achieved a 99% accuracy rate, a 0.99 average area under the receiver operator characteristic curve (AUC-ROC), and a false positive rate of less than 2% across various 10-fold cross-validation tests. These results underscore our classification method’s reliability, accuracy, and effectiveness. This innovative strategy for identifying harmful network traffic holds significant potential for enhancing intrusion detection systems in both traditional network traffic analysis and cyber-physical systems analysis, such as smart grids. Furthermore, we present a new intrusion detection system (IDS) that combines a multilayer perceptron (MLP) network with an artificial bee colony (ABC) and fuzzy clustering algorithms. The MLP network distinguishes between standard and irregular network data packets, while the ABC method fine-tunes the connections and prejudices of the MLP during its training phase. We used the CloudSim model and the NSL-KDD data set to validate our method, evaluating it with criteria including mean absolute error (MAE), root mean square error (RMSE), and the kappa coefficient. Our findings demonstrate that our proposed technique surpasses current leading methods in the field.
},
}