The Research Of Inherent Randomness In GNU/Linux And Its Application In Random Number Generators

In the trend of rapid development of the computer science and technology,the computer performance has become the focus of the academic and industry currently.At present,multi-core processors have widely used in different computer platforms,at the same time there introduces some new techniques,which increases the complexity of computer operating systems.On another hand,the traditional random number generators usually harvest the event time fluctuation as the entropy source in the GNU/Linux systems,which is limited and difficult to achieve.What's more,in some cases that this type of random numbers generated by traditional means cannot meet the demand of some security areas(such as encryption applications).Hence,it is necessary to find a new method as the entropy source of random number generators.Nicholas Mc Guire suggested a new method called ESRNG to generate true random numbers by utilizing the inherent randomness of computer system at 14 th Real Time Linux Workshop.He gave the method to extract the entropy and implemented Galton board porotype,and this method has important significance for finding new entropy sources.However,there are multiple constraints in the implementation process,for example,for the Galton board model,the N value of race condition and the HISIZE used need many experiments to be determined,and the number of threads it used is also different,so it has strong platform dependency.In this paper,based on the Linux system,we have improved this approach: According the relationship between the race condition occurrence and the N value,we design a new method to determine the N value by using the Python fit method,what's more,in our porotype,there is only one hardware parameter,and the threads used is fixed,therefore,it eliminates the platform dependency.And according to the ISO/IEC 18031:2011 standard,we designed a non-physical true random number generator called DDSRNG,which is suitable for encryption applications.Finally,we use two common statistical test tools – ENT and NIST Special Publication 800-22 to verify the randomness of the DDSRNG.The results show that the random number generator designed in this paper passes the test and it is better than ESRNG for generating random numbers which is suitable for the encryption applications.Meanwhile,the results indicate that different platforms have different randomness performance.