Research On Security And Trustworthiness Of Intelligent Terminal Software

With the development of information technology,intelligent terminal has become an indispensable part of people's life.Currently,Android intelligent terminal occupies the majority of the market.Due to the free and open-source characteristics of the Android platform,various kinds of application software are born to be widely used in people's requirements.Meanwhile,driven by profits,many illegal software developers are attracted.In recent years,the security threats against Android Software are gruadually increasing,including malicious fees,privacy theft,bundled download,etc.With the rapid increase of software requirement and application,how to ensure the information security of intelligent in users is a challenge task today.Based on the above situation,this thesis focuses on the research on the security and trustworthiness of Android application software.In terms of security,due to the malware software has become the main threat to its ecosystem,the research on detecting malware software to promote effectively identification of users is conducted.In terms of trustworthiness,since the problem of software trustworthiness has become increasingly prominent for some unknown applications,the research on trustworthiness evaluation is performed to ensure its safe operation.The main contents of the thesis are summarized as follows:1.For the security of application software,an Android malware detection method combining frequent item set and machine learning is proposed.In the method,many static features from various applications are firstly extracted,and the differences between benign software and malicious software is analyzed as feature indicators.Then,Apriori algorithm-based method is designed to mine the frequent item sets of the permissions,and combining with the application programming interface(API),the feature selection algorithm is adopted to extract important features.Finally,three machine learning classifiers are used to achieve the learning and classification of features so as to achieve Android malware detection.The results of experiments show that frequent item sets of permissions are helpful to distinguish malware.At the same time,the method of fusing frequent item sets and machine learning not only has a very high accuracy rate,but also has a low consumption of system resources.The accuracy rate of random forest classifier achieves 99.28%,which effectively improves the security of software.2.In the Android open environment,software trustworthiness evaluation is an important guarantee of software security.a trustworthiness evaluation method based on SVM and improved evidence theory is proposed for Android intelligent terminal application software.In this method,at the training stage,attribute evidence is formed by measuring the operation behavior,CPU change,memory change,network traffic change and other attributes from application software samples.Then,attribute evidence as input to train SVM which associated with each attribute.At the evaluating stage,according to the measurement results of application software,the classification probability is formed by SVM's classification model which corresponding to each attribute.Then,on the basis of setting adaptive weight to each attribute evidence by entropy weight method,the multi-attribute is synthesized based on improved evidence theory to evaluate the software trustworthiness.The results of experiments show that this method has a higher accuracy than that using the single attribute.At the same time,this method not only avoids the conflict problem in the classical evidence combination,but also has a more comprehensive evaluation scope compared with the existing evaluation methods.