Security Detection System Based On The Android Paltform

In recent years, with the arrival of the era of mobile Internet, mobile terminals has become one of the main equipment for people to obtain information, The popularity of smart phones has brought rapid development of the Android system. Due to the open source Android system, non-strict signature mechanism and non-strict application release management, Android may be the most vulnerable operating system for hackers, more and more lawless profiteering by malware on Android.Now Android malware has a number of large, spread quickly, multi-variant, and caused great harm, therefore need effective security testing system, enabling the detection of unknown malware. This subject aims at based on the Android operating system, combining machine learning, focus on the functional behavior of the malware analysis and recognition, innovate and improve the traditional detection methods, through studying samples of known software to build safety inspection of models to determine unknown malicious software.This paper first of all from the Android platform structure design thought of the characteristics and perspective analysis of the safety mechanism on the platform itself, while samples of malicious software in the library and analysis of malicious acts, more targeted detection methods proposed. And then design and implements a security detection system from the static detection and dynamic detection of aspect. Static detection decompile the application installation file to extract static characteristic properties, characteristics of these properties can be achieved from a certain point of that software feature, then use the detected model calculated from the training phase, to determine the static detection of malicious. Dynamic behavior detection is established by intercepting system calls for state sequence; Mining contains the application of behavioral patterns in the behavior of the state transition diagram behind. Recognition based on hidden Markov models built to detect the presence of application security testing models of abnormal behavior. At the same time, in order to verify the correctness of the system, we design a test. Test the static detection system, designed to compare the system internal longitudinal comparison test lateral and similar products. Test the dynamic detection system, designed a system load test, the best parameter selection test and system precision tests. The final test results show that the static detection and dynamic detection has good complementary system for the detection of unknown malware has better detection effect. The advantage of this paper to achieve security detection systems detect more own characteristics and behavior recognition of malware for the Android platform, making the system more precise detection of the unknown software.