Research on Deep Learning and Feature Aggregation Techniques for Web Security
Enregistré dans:
| Publié dans: | Journal of Web Engineering vol. 24, no. 2 (2025), p. 291 |
|---|---|
| Auteur principal: | |
| Publié: |
River Publishers
|
| Sujets: | |
| Accès en ligne: | Citation/Abstract Full Text Full Text - PDF |
| Tags: |
Pas de tags, Soyez le premier à ajouter un tag!
|
| Résumé: | With the rapid development of internet technologies, Web services have been widely applied in various fields, including finance, healthcare, education, e-commerce, and the Internet of Things, bringing great convenience to humanity. However, Web security threats have become increasingly severe, with side-channel attacks (SCA) emerging as a covert and highly dangerous attack method. SCAs exploit non-explicit information, such as network traffic patterns and response times, to steal sensitive user data, posing serious threats to user privacy and system security. Traditional detection methods primarily rely on rule-based feature engineering and statistical analysis, but these methods show significant limitations in terms of detection performance when dealing with complex attack patterns and high-dimensional, large-scale network traffic data. To address these issues, this paper proposes a side-channel leakage detection method based on SSA-ResNet-SAN. The SSA (sparrow search algorithm) is an optimization mechanism, intelligently searching for globally optimal feature subsets to enhance the model’s feature selection capabilities and global optimization performance. Combined with deep residual networks (ResNet) and the signature aggregation network (SAN), the method performs a comprehensive analysis of both single-attribute and aggregated-attribute features in network traffic, thereby improving the model’s accuracy and robustness. Experimental results demonstrate that SSA-ResNet-SAN significantly outperforms existing methods on multiple practical datasets. On the Google dataset, the use of aggregated attribute features enables SSA-ResNet-SAN to achieve an accuracy of 93%, which is substantially higher than that of other models. Furthermore, in multi-class tasks on the Baidu and Bing datasets, SSA-ResNet-SAN exhibits strong robustness and applicability. These experimental results fully validate the outstanding performance of SSA-ResNet-SAN in side-channel leakage detection, providing an efficient and reliable solution for the field of Web security. |
|---|---|
| ISSN: | 1540-9589 1544-5976 |
| DOI: | 10.13052/jwe1540-9589.2426 |
| Source: | Computer Science Database |