Reinforcement learning-driven temporal knowledge graph reasoning for secure data provenance in distributed networks

被引：0

作者：

Yunxiang Qiu ^{[1
]}

Yuting Tang ^{[2
]}

Liangguo Chen ^{[2
]}

Shuyu Jiang ^{[1
]}

Ying Huang ^{[1
]}

Xingshu Chen ^{[2
]}

机构：

[1] Sichuan University,School of Cyber Science and Engineering

[2] Huaxin,Huaxin Consulting Design and Research Institute Co., Ltd.

来源：

Peer-to-Peer Networking and Applications | 2025年 / 18卷 / 4期

关键词：

Data provenance; Data security; Distributed networks; Temporal knowledge graph; Reinforcement learning;

D O I：

10.1007/s12083-025-01977-y

中图分类号：

学科分类号：

摘要：

Distributed networks have become a cornerstone of modern computing systems, enabling scalable, efficient, and interconnected data processing across diverse computing paradigms, including Cloud Computing, IoT, and Edge Computing. However, ensuring data security and privacy within these distributed networks remain a significant challenge, particularly in tracing the provenance and flow of sensitive data due to their dynamic and decentralized nature. To tackle this challenge, we propose TR-DPro, a novel data provenance scheme that combines Temporal Knowledge Graph (TKG) and Reinforcement Learning (RL) to achieve adaptive and accurate data provenance in distributed networks. TKG captures the temporal and relational dynamics of data flow, while RL optimizes the reasoning process by dynamically learning efficient provenance paths, reducing computational complexity and improving accuracy. Simulation experiments conducted on a private dataset of network traffic and security events demonstrate that the proposed framework, TR-DPro, can effectively and accurately trace data flow paths while outperforming existing methods in key performance metrics.

引用