太原理工大学学报2026，Vol.57Issue(1)：177-185,9.DOI:10.16355/j.tyut.1007-9432.20250418
任务重要度感知的结果篡改自适应检测方法研究
Adaptive Detection Method for Result Tampering Based on Task Importance Awareness
霍盈宇 1刘琪 2赵智慧 2陈永乐2
作者信息
- 1. 太原理工大学工业互联网安全山西省重点实验室,山西 太原
- 2. 太原理工大学工业互联网安全山西省重点实验室,山西 太原||太原理工大学计算机科学与技术学院(大数据学院),山西 太原
- 折叠
摘要
Abstract
[Purposes]In distributed collaborative computing environments,to address the chal-lenge that computation results are tampered and detection costs are high,an adaptive detection strate-gy is designed by incorporating task importance.This strategy will improve detection efficiency and re-duce unnecessary resource consumption.[Methods]In this paper,a task-importance-aware adap-tive detection method was proposed for result tampering.A large language model was used to automat-ically decompose the complex tasks submitted by users'and quantify the importance of each sub-task.Then,detection resources were dynamically allocated according to task importance.Critical sub-tasks were prioritized,and detection frequency was adaptively adjusted to balance coverage and resource overhead.The simulation experiments were conducted by using the GSM8K dataset and Worfbench dataset.[Results]The results show that the proposed method significantly reduces detection over-head compared with traditional fixed-frequency strategies.The overhead is reduced by approximately 68%,with no noticeable degradation in detection capability.The results verify the method's effective-ness and feasibility.[Conclusions]This work achieves targeted detection for high-importance tasks,improving detection performance while maintaining good applicability and scalability.It is help-ful for scenarios that require high result integrity,such as autonomous driving,multi-agent collabora-tion,and edge computing.关键词
多方协作计算/结果篡改检测/自适应检测/任务重要度/大语言模型赋能Key words
distributed collaborative computing/result tampering detection/adaptive detection/task importance/large language model-enabled分类
信息技术与安全科学引用本文复制引用
霍盈宇,刘琪,赵智慧,陈永乐..任务重要度感知的结果篡改自适应检测方法研究[J].太原理工大学学报,2026,57(1):177-185,9.
