WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM
The frequency of cyber-attacks has been rising in recent years due to the fact that startup developers have failed to overlook security issues in the core web services. This stated serious concerns about the security of the web. Therefore, this paper proposes a hybrid model built on the base of Con...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
University of Zakho
2025-01-01
|
| Series: | Science Journal of University of Zakho |
| Subjects: | |
| Online Access: | http://sjuoz.uoz.edu.krd/index.php/sjuoz/article/view/1404 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841533377359904768 |
|---|---|
| author | Sarbast H. Ali Arman I. Mohammed Sarwar MA. Mustafa Sardar Omar Salih |
| author_facet | Sarbast H. Ali Arman I. Mohammed Sarwar MA. Mustafa Sardar Omar Salih |
| author_sort | Sarbast H. Ali |
| collection | DOAJ |
| description |
The frequency of cyber-attacks has been rising in recent years due to the fact that startup developers have failed to overlook security issues in the core web services. This stated serious concerns about the security of the web. Therefore, this paper proposes a hybrid model built on the base of Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU) and an attention mechanism to detect vulnerabilities in application code. Particularly, the model can help detect attacks based on Structured Query Language Injection (SQLi), Cross-Site Scripting (XSS), and command injection. When using the dataset SXCM1, our model achieved 99.77%, 99.66% and 99.63% for training, validation and testing, respectively. The results obtained on data from the DPU-WVD dataset are even better because it was 99.97%, 99.98% and 99.99% for training, validation and testing, respectively. These results significantly outperform the state-of-the-art models and can strongly identify vulnerabilities in web applications. Through training, on both the SXCM1 and DPU-WVD datasets, the model achieved an accuracy rate of 99.99%. The results show that this combination model is highly effective at recognizing three vulnerability categories and surpasses cutting-edge models that usually specialize in just one type of vulnerability detection.
|
| format | Article |
| id | doaj-art-b4ecb0f01c644bcc818970d5f0a2da1e |
| institution | Kabale University |
| issn | 2663-628X 2663-6298 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | University of Zakho |
| record_format | Article |
| series | Science Journal of University of Zakho |
| spelling | doaj-art-b4ecb0f01c644bcc818970d5f0a2da1e2025-01-16T02:32:49ZengUniversity of ZakhoScience Journal of University of Zakho2663-628X2663-62982025-01-0113110.25271/sjuoz.2025.13.1.1404WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISMSarbast H. Ali0Arman I. Mohammed1Sarwar MA. Mustafa2Sardar Omar Salih3Duhok Technical College, Duhok Polytechnic University.Duhok Technical College, Duhok Polytechnic UniversityCollege of Science, University of DuhokDuhok Technical Institute, Duhok Polytechnic University The frequency of cyber-attacks has been rising in recent years due to the fact that startup developers have failed to overlook security issues in the core web services. This stated serious concerns about the security of the web. Therefore, this paper proposes a hybrid model built on the base of Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU) and an attention mechanism to detect vulnerabilities in application code. Particularly, the model can help detect attacks based on Structured Query Language Injection (SQLi), Cross-Site Scripting (XSS), and command injection. When using the dataset SXCM1, our model achieved 99.77%, 99.66% and 99.63% for training, validation and testing, respectively. The results obtained on data from the DPU-WVD dataset are even better because it was 99.97%, 99.98% and 99.99% for training, validation and testing, respectively. These results significantly outperform the state-of-the-art models and can strongly identify vulnerabilities in web applications. Through training, on both the SXCM1 and DPU-WVD datasets, the model achieved an accuracy rate of 99.99%. The results show that this combination model is highly effective at recognizing three vulnerability categories and surpasses cutting-edge models that usually specialize in just one type of vulnerability detection. http://sjuoz.uoz.edu.krd/index.php/sjuoz/article/view/1404CNNWeb vulnerabilitiesDeep LearningXSSSQL injection |
| spellingShingle | Sarbast H. Ali Arman I. Mohammed Sarwar MA. Mustafa Sardar Omar Salih WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM Science Journal of University of Zakho CNN Web vulnerabilities Deep Learning XSS SQL injection |
| title | WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM |
| title_full | WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM |
| title_fullStr | WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM |
| title_full_unstemmed | WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM |
| title_short | WEB VULNERABILITIES DETECTION USING A HYBRID MODEL OF CNN, GRU AND ATTENTION MECHANISM |
| title_sort | web vulnerabilities detection using a hybrid model of cnn gru and attention mechanism |
| topic | CNN Web vulnerabilities Deep Learning XSS SQL injection |
| url | http://sjuoz.uoz.edu.krd/index.php/sjuoz/article/view/1404 |
| work_keys_str_mv | AT sarbasthali webvulnerabilitiesdetectionusingahybridmodelofcnngruandattentionmechanism AT armanimohammed webvulnerabilitiesdetectionusingahybridmodelofcnngruandattentionmechanism AT sarwarmamustafa webvulnerabilitiesdetectionusingahybridmodelofcnngruandattentionmechanism AT sardaromarsalih webvulnerabilitiesdetectionusingahybridmodelofcnngruandattentionmechanism |