Research On Android Privacy Data Protection Mechanism Based On Machine Learning

With the extensive use of the Android operating system,the number of Android-based applications is increasing.However,the lack of supervision and management of many Android third-party application markets has led to a growing number of malware and variants on android platforms.This phenomenon causes a serious threat to users' information security.How to effectively prevent the spread of Android malware and improve the security of Android system is critical to protect the security of users' privacy information.In this paper,we comprehensively analyze the advantages and disadvantages of the current Android privacy protection model,and then propose a new Android privacy data protection scheme based on machine learning algorithm.The scheme provides a perfect privacy protection strategy,which is divided into the application risk assessment module and the privacy protection module.For application risk assessment module,it mainly uses the machine learning classification algorithm to detect the malware on the Android platform to assess its risk level,which is the core part of the privacy protection scheme.Aiming at the detection of Android malware,we propose a new hybrid feature analysis method which combines the advantages of static analysis and dynamic analysis.The permission request,API calls and dynamic runtime behavior information are extracted as the hybrid feature vectors.In addition,the feature selection algorithm is used to further optimize the extracted information to remove some of the useless features.For privacy protection module,it mainly uses mock data to return the access request of the untrusted application based on the result of the risk assessment table,and protects the user's privacy information.The user can set up mock data protection for the sensitive data requested by the application to return real or mock data based on the actual situation.The use of the hybrid analysis method in this paper improves the accuracy of the classification and achieves better detection results.In the experiments,the scheme we designed can effectively prevent malicious application access to the user's sensitive information and can well avoid the risk of crashes when applications do not get the required permissions,and it effectively protects the user's privacy data security.