Research And Implementation Of Android Aging Detection Method Based On GAN-ASD

Software aging refers to the phenomena such as performance degradation,system downtime,and failures that occur in a long-running software system.In the face of high load and high power consumption scenarios such as live video,edge computing,and mobile games,mobile devices have the risk of software aging,which seriously affects the user experience.Accurate detection of aging in Android system software and implementation of rejuvenation operations to improve user fluency are of great significance to commercial software operations.However,with the continuous development of Android operating system and the increasing complexity of application types,it becomes more difficult to identify software aging in Android system.Combined with the theory of system state division,this thesis proposes an Android aging detection method based on generative adversarial network technology,named as GAN-ASD,which can accurately identify the running state of Android system.Such method facilitates the timely execution of anti-aging operation,ensure the stable operation of the system,and optimize the user experience.First,in our GAN-ASD method,it uses the combined aging indicators to comprehensively analyze the software aging phenomenon from aspects of users and systems.Data set processing method that meets the characteristics of software aging is proposed to fill in the real-time data record samples that are missing due to software aging.This makes the system operation data set for analysis as complete as possible to present the situation of the system.Second,we use a time-series-data generation method based on the boundary equilibrium generative adversarial network(BEGAN).It takes the joint time series data set as input,and aims to dynamically generate fitting data with the characteristics of the Android operating environment.The K-Means clustering algorithm trained by this fitting data set has the function of real-time detection of the software aging of Android devices.Third,in order to verify the effectiveness of our GAN-ASD,this thesis compares the precise anti-aging method guided by GAN-ASD with similar related methods.We propose two indicators to evaluate our method,that is,user experience coefficient and anti-aging cost coefficient.The experiments results show that in terms of evaluation indicators,our method performs best,which verifies that the software aging detection method based on GAN-ASD is effective.Finally,based on the GAN-ASD method,we develop the web application with microservice architecture,which can detect software aging in real time in the Android software aging detection service.