To search, Click below search items. |
| All Published Papers Search Service |
|
|
|
Title |
A Deception-Driven AI Framework for Detecting Cyberattacks in Mobile Financial Applications Using GANs, Reinforcement Learning, and SVM |
|
Author |
Chidiebere, Abigail Nnenna, Abiodun Esther Omolara, Mohammed Hammawa |
| Citation |
Vol. 26 No. 3 pp. 23-36 |
|
Abstract |
Mobile financial applications have become essential infrastructures for banking, digital payments, and financial services. However, their rapid adoption has significantly expanded the cyberattack surface of financial institutions, exposing users and service providers to increasingly sophisticated threats such as credential theft, account takeover, malware injection, and automated fraud attacks. Conventional rule-based and signature-driven security mechanisms often struggle to detect these evolving threats because they rely on previously known attack patterns and static detection rules. This study proposes a deception-driven artificial intelligence framework for detecting cyberattacks targeting mobile financial applications. The framework integrates Generative Adversarial Networks (GANs) to generate realistic synthetic attack scenarios, Reinforcement Learning (RL) to dynamically adapt deception strategies, and Support Vector Machine (SVM) classification to identify malicious interactions. The proposed system operates within a layered deception architecture that deploys honey accounts, honey transactions, and authentication traps to lure attackers into controlled environments while simultaneously collecting behavioral intelligence for threat analysis and detection. Experimental evaluation using benchmark cybersecurity datasets (NSL-KDD and CICIDS2017) augmented with GAN-generated attack samples demonstrates the effectiveness of the proposed approach. The integrated GAN?SVM detection model achieved 96.5% classification accuracy with a false positive rate of 1.8%, outperforming reinforcement learning based detection (93.2% accuracy) and standalone SVM classification (88.9% accuracy). Additional ROC?AUC analysis and ablation experiments further confirm the advantages of combining generative attack simulation with adaptive deception strategies. The results indicate that integrating deception technologies with artificial intelligence can significantly improve cyber threat detection performance while maintaining low false alarm rates. The proposed framework provides a scalable and proactive cybersecurity approach capable of enhancing the resilience of financial technology systems against evolving cyber threats. |
|
Keywords |
Deception technology; honeypots; honey transactions; mobile banking security; generative adversarial networks; reinforcement learning; support vector machines; anomaly detection |
|
URL |
