Algorithm for Detection and Defense of Neural Network Technology Based on Neural Network and Multimedia

Authors

  • Hui Ke Department of Network and Information Security, Chongqing Vocational Institute of Safety &Technical, Wanzhou 404120, Chongqing, China

Abstract

 

Neural networks have been widely usedin thefields ofimage recognition, voice recognition, and natural language processing among others. However, the neural network model is less robust against adversarial samples. That is, when small perturbations are artificially added in the input data, the output of the model will change. This phenomenon, known as adversarial example attack, can lead to misclassification and performance degradation of the model. In response to this problem, the academic community has proposed a variety of adversarial sample attack detection and defense methods. Adversarial attack detection is intended to detect adversarial samples and filter them out so that they are not input into the model. On the other hand, the purpose of adversarial defense is to increase the robustness of the model during training, making it more stable against adversarial samples. At present, there are still some problems in regard to the detection and defense of adversarial sample attacks. Therefore, further research and exploration are still of great significance. In this study, we examine “Adversarial Sample Attack and Defense in Neural Networks” to determine the protection capabilities of neural network technology. Through the experiment, the detection experiment of adversarial samples is carried out, and the three attack methods of C&W, FGSM and FIA are detected by using the adversarial detection algorithm based on neural network technology, and the detection success rate is recorded. Experimental results show that the average detection success rate of C&W, FGSM and FIA using the adversarial
detection algorithm based on neural network technology is 97.257%, 95.354% and 94.602%, respectively. This indicates that the algorithm has a high detection success rate for these three attack methods, can effectively identify adversarial samples, and improve the robustness and accuracy of the model.


Keywords: system attack, neural network, against the samples, image recognition

Cite As

H. Ke, "Algorithm for Detection and Defense of Neural Network Technology Based on Neural Network
and Multimedia", Engineering Intelligent Systems, vol. 33 no. 4, pp. 377-384, 2025.


Downloads

Published

2025-07-01

Issue

Vol. 33 No. 4 (2025): International Journal of Engineering Intelligent Systems

Section

Article

License

The submission of a paper implies that, if accepted for publication, it will not be published elsewhere in the same form, in any language, without the prior consent of the publisher. Before publication, authors are requested to assign copyright to CRL Publishing Ltd. This allows CRL to sanction photocopying, and to authorize the reprinting of issues or volumes according to demand. Authors' traditional rights will not be jeopardized by assigning Copyright in this way, as they retain the right to reuse the material following publication, and to veto third-party publication.