基于用户操作的APP软件异常分析  

Abnormal Analysis of APP Software Based on User Operation

作　　者：彭可磊 姜瑛[1,2] 刘海毅 PENG Ke-lei;JIANG Ying;LIU Hai-yi(Yunnan Key Laboratory of Artificial Intelligence,Kunming University of Science and Technology,Kunming 650504,China;School of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650504,China)

机构地区：[1]昆明理工大学云南省人工智能重点实验室,云南昆明650504 [2]昆明理工大学信息工程与自动化学院,云南昆明650504

出　　处：《计算机技术与发展》2025年第3期62-68,共7页Computer Technology and Development

基　　金：国家自然科学基金(62162038);国家重点研发计划项目(2018YFB1003904);云南省计算机技术应用重点实验室开放基金(2020101)。

摘　　要：随着智能手机的广泛普及,APP软件已成为日常生活中不可或缺的一部分。然而,APP软件的频繁使用也随之带来了越来越多的软件异常,这些软件问题不仅会影响用户体验,还可能造成数据丢失或安全风险。为了深入理解用户操作导致APP软件异常的运行过程,该文从用户使用APP软件过程记录的日志信息中抽取了相关的日志元素,提出了一种基于用户操作的APP软件异常分析方法。首先,利用双向变压器编码器(Bidirectional Encoder Representations from Transformers,BERT)解析日志文本,以捕捉其中上下文的语义;其次,通过日志元素的时序特性分析元素间的因果关系;此外,基于日志元素的分布特征计算互信息,分析日志元素间的依赖程度,并融合上述多种关系特征构建日志元素关系图;最后,借助图遍历算法揭示用户操作引起软件异常的运行过程。实验验证,该方法的准确率和F1-score分别达到86.6%和74.1%。对比实验显示,该方法优于同类方法,验证了该方法的有效性。With the widespread popularity of smart phones,mobile applications(APP)have become an indispensable part of daily lives.However,the frequent use of APP software has also brought about more and more software anomalies,which not only affect the user experience,but also may cause data loss or security risks.In order to deeply understand the running process of APP software anomalies caused by user operations,we extract relevant log elements from the log information recorded by users using APP software processes,and propose an APP software anomaly analysis method based on user operations.Firstly,the Bidirectional Encoder Representations from Transformers(BERT)is used to parse the log text to capture the semantics of the context.Secondly,the causal relationship between log elements is analyzed through the temporal characteristics of log elements.In addition,mutual information is calculated based on the distribution characteristics of log elements,the degree of dependence between log elements is analyzed,and the log element relationship diagram is constructed by integrating the above various relationship characteristics.Finally,the operation process of software anomalies caused by user operations is revealed by graph traversal algorithm.Experimental results show that the precision rate and F1-score of the proposed method are 86.6%and 74.1%,respectively.Comparative experiments show that the proposed method is superior to similar methods,which verifies its effectiveness.

关 键 词：用户操作 关联分析 APP软件异常 互信息 特征融合 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]