TY  - JOUR
AU  - Singh, Ankit Kumar
AU  - Singh, Aastha
AU  - Agnihotri, Avanish Kant
AU  - Trivedi, Amit
AU  - Nadeem, Mohd
PY  - 2026
DA  - 2026/07/01
TI  - Artificial Neural Network-Based Intelligent Framework for Multiclass Network Intrusion Detection in Modern Cybersecurity Systems
JO  - Recent Progress in Science and Engineering
SP  - 014
VL  - 02
IS  - 03
AB  - The advent of cloud computing, the Internet of Things (IoT), and digital services has led to an increase in the number and complexity of cyberattacks. The intrusion detection methods currently used, based on machine learning and artificial intelligence, include SVM, random forest, deep learning, convolutional neural networks, and LSTM. These methods are more advanced than the signature method since they are more efficient. However, despite these advancements, several issues with intrusion detection systems remain. They include high false-positive rates, computational complexity, limited scalability, inability to detect zero-day attacks, and poor real-time performance. To overcome such challenges, this paper suggests the development of an intelligent system for network intrusion detection using artificial neural networks (ANNs). This method is intended to increase the accuracy of cyber threat detection and minimize the number of false positives through adaptive and nonlinear learning. Data preprocessing, data encoding, feature normalization, feature selection, and feedforward neural networks are some of the methods employed in the classification of data traffic into malicious and normal traffic. The types of intrusions considered in this study include DoS, Probe, R2L, and U2R attacks. According to experiments performed using intrusion detection benchmark data, the proposed ANN algorithm delivers accuracy, precision, recall, and F1 score values of 97.6%, 96.8%, 97.2%, and 97.0%, respectively, while maintaining a low false-positive ratio of only 2.1%. The comparative analysis further demonstrates that the ANN classifier outperforms other classification techniques, such as random forests, decision trees, and support vector machines. The findings thus demonstrate that the ANN can be successfully used for detecting attacks and improving the security of networks, cloud computing, and the IoT.
SN  - 3067-4573
UR  - https://doi.org/10.21926/rpse.2603014
DO  - 10.21926/rpse.2603014
ID  - Singh2026
ER  -