Self-Learning Attack Generation toward Edge-Based Consumer Web Security Testing

doi:10.1109/mce.2026.3731752

Self-Learning Attack Generation toward Edge-Based Consumer Web Security Testing

2026-09-10

0

PRE

AI

DOI：10.1109/mce.2026.3731752

Original

Original request for help

Save

Abstract

En 中文

Security evaluation of web-based services for consumer electronics (CE) applications is increasingly important, as many device management platforms, support systems, and companion services rely on PHP-based web technologies. Existing black-box fuzzing approaches often suffer from limited vulnerability coverage due to the large input space, while existing XSS datasets may be incomplete or biased toward known payload patterns. This paper presents an automated web security testing system that leverages Generative Adversarial Networks (GANs) to learn and generate diverse attack patterns from limited samples. WordPress is adopted as a practical PHP-based experimental carrier because of its widespread use in consumer-facing services, while the proposed testing process focuses on general PHP-based input handling, payload execution, and response analysis rather than WordPress-specific functions. The framework is intended to support local attack input generation as a future edge or gateway deployment option. The present experimental validation focuses on payload generation and vulnerability testing in a controlled web testbed rather than on embedded hardware. The results show that the generated payloads can support security testing of consumer web services under limited and imperfect attack data.

Keywords:

Cross-site scripting

Payloads

Testing

Training

Modeling

Security

Learning (artificial intelligence)

Sequences

Sequential analysis

Generative adversarial networks

Self-Learning Attack Generation toward Edge-Based Consumer Web Security Testing

Abstract

Journal

Organization

Cited Papers