2026-10-11T19:55:57Z https://www.scidb.cn/oai

10.57760/sciencedb.00zzl 2026-08-26T13:41:42Z

2026-08-26 Large Language Model-Generated Honeypot Response Dataset for Web Vulnerability Attacks doi:10.57760/sciencedb.00zzl en This dataset is designed for the research and construction of web vulnerability attack trapping and honeypot dynamic response. It mainly includes four typical types of web vulnerability attack requests and their corresponding honeypot response payload data: SQL injection, path traversal, remote code execution, and cross site scripting. The dataset addresses the problems of traditional honeypots relying on fixed response templates, single vulnerability interaction features, and difficulty in adapting to diverse attack requests. A large language model is introduced to dynamically generate HTTP response data with corresponding vulnerability interaction features based on different vulnerability attack requests, vulnerability types, and contextual information. The data files include four types of data: SQL injection, path traversal, remote code execution, and cross site scripting. Each piece of data contains vulnerability numbers, HTTP response status generated by honeypots, response headers, response bodies, and related generated results. This dataset can be used for research on dynamic honeypots driven by big language models, web vulnerability response simulation, attack trapping, vulnerability interaction feature generation, and network security big model capability evaluation. It provides a data foundation for analyzing the understanding and response capabilities of different big language models to web vulnerability attack requests, as well as researching adaptive honeypot response mechanisms for unknown or variant attacks.  vulnerability response payload simulation; large language model; agent; honeypot 袁天伟 吕世超 RESTRICTED dataset Science Data Bank