Intelligent Traffic Classification Using Convolutional Neural Networks
Year: 2026
Type: conference-paper
Abstract: A vital part of present day community control and cybersecurity systems is correct network visitors categorization. The effectiveness of traditional port-based and payload-based visitors identity strategies has been greatly dwindled by means of the good sized use of encrypted verbal exchange protocols, which has brought about the need for smart statistics-driven strategies. In order to conquer the problems presented by way of encrypted and numerous community traffic, this observe proposes an clever visitors categorization framework based totally on a multi-channel convolutional neural network (CNN). This research's fundamental contribution is a completely unique spatiotemporal waft illustration that concurrently captures packet-stage byte shape, temporal dynamics, and aggregated statistical properties by converting every community flow into a 3-channel tensor. The CNN can learn hierarchical and discriminative features due to the fact to this illustration, which gets rid of the need for manually created characteristic engineering. To offer reliable and constant learning, the advised architecture combines many convolutional blocks with function fusion and regularization techniques. Numerous research overlaying a variety of software categories had been accomplished making use of a combination of publicly reachable datasets and encrypted site visitors lines. The results demonstrate good type accuracy, brief convergence, and balanced performance during training. To confirm the effectiveness and generalizability of the proposed method, training curves, confusion matrices, and consistent with-magnificence F1-rankings are examined. All things considered, the recommended multi-channel CNN architecture provides a scalable, encryption-resistant solution for intelligent site visitor classification in state-of-the-art network scenarios.
Author: Yassen Abdelkhaleq
Institution: Iraqi University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203708967