低功耗嵌入式软件信息时序突发异常检测方法  

Method of Time Series Burst Anomaly Detection in Low-Power Embedded Software

作　　者：石巍[1] SHI Wei(Guangxi College of Water Resources and Electric Power,Nanning 530023,China)

机构地区：[1]广西水利电力职业技术学院,广西南宁530023

出　　处：《系统仿真技术》2024年第4期352-357,共6页System Simulation Technology

摘　　要：针对低功耗嵌入式软件信息时序突发异常问题,提出一种时序突发异常检测方法。建立扩展语义接口自动机模型,对低功耗嵌入式软件信息展开分析,学习构件之间的交互与通信行为;通过最小二乘支持向量机方法,检测软件信息时序中存在的离群点,提高异常检测精度;建立长短期记忆(LSTM)网络模型,根据设定的阈值实现低功耗嵌入式软件信息时序突发异常检测。实验结果表明,所提方法可有效检测软件信息时序的状态和迁移变化情况,同时在检测精度和效率方面表现出良好的性能。A time series burst anomaly detection method is proposed to address the issue of time series burst anomalies in low-power embedded software.An extended semantic interface automata model is established to analyze low-power embedded software information,and learn the interaction and communication behavior between components.The least squares support vector machine method is used to detect outliers in software information time series and improve anomaly detection accuracy.A LSTM network model is employed to achieve time series burst anomaly detection in low-power embedded software information based on a set threshold.The experimental results show that the proposed method can effectively detect the state and transition changes of software information time series,while exhibiting good performance in detection accuracy and efficiency.

关 键 词：嵌入式软件信息 扩展语义接口自动机模型 最小二乘支持向量机 时序异常检测 长短期记忆网络 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]