Android True Random Number And Device Feature Generation Methods Based On Process Scheduling

The key is one of the essential components to secure Android devices.Generating a key needs random numbers.In order to acquire high-quality random numbers,a noise source with high-entropy is required.On the other hand,we need to identify smart phone when provide personalized services to different mobile phone.Therefore,features of the device need to be collected.Generating random numbers requires a noise source with high-entropy,while identifying devices need a source with low-entropy.In this paper,we analyze the randomness of the process scheduling information of Android system by calculating the information entropy.A noise source with high-entropy and a relatively unique process access density characteristic are found.Based on this,we propose methods for generating true random numbers and extracting device features.On the one hand,the true random number generator uses physical random events.Besides,rare physical true random number generators are integrated in Android devices.Therefore,it is particularly important to find a noise source with high-entropy from the system.In this paper,we analyze the randomness of the side channel information in the Android system and find that the rules of the PID calling order reflect the change of interruption times.A sequence of time t,which is required to increase the same number of interruptions,follows the negative exponential distribution.The sequence is a Poisson stream that owns a memoryless property,that is,each value of the sequence is only related to adjacent values.So,the increased number of interruption times during the identical time can be regarded as a noise source.Random numbers can be generated by preprocessing the noise source information,quantization and coding.The random numbers are evaluated by calculating the min-entropy.The effects of different interval partition methods are compared.The relationship among sampling interval,the variance of the noise source and the min-entropy of random numbers is analyzed.The correlation between several processes and the min-entropy of random numbers is discussed.Our experimental results show that random numbers with higher entropy can be obtained by this method for dividing into five parts with mean non-uniform quantization,sampling per 1ms and suppressing the running of the ksoftirqdr process.On the other hand,we extract features of Android devices by getting information that can be modified with root privilege such as IMEI with sensitive permissions.In this paper,we propose two methods for extracting the feature of devices with zero-permission.Multiple processes are running at the same time in the system.The state changes with different systems,and the PID calling density that is process scheduling information is different.A device operates stably for a period of time,and the PID calling density remains basically unchanged.The stable operation of the equipment owns low entropy that reflects less randomness.In this paper,we propose two methods for extracting features of devices by analyzing the PID calling density and regarding the running process information whose pid is less than 300 as the source.The threshold of the similarity matching algorithm is determined by calculating the false acceptance rate,false rejection rate and equal error rate.The methods for extracting features of devices are evaluated by calculating precision,recall and F_β score.