Seogchung Seo, National Security Research Institute, Korea; Sangwoon Jang, National Security Research Institute, Korea; Yewon Kim, Kookmin University, Korea; and Yongjin Yeom, Kookmin University, Korea In this talk, the speaker share our project to devise entropy estimation methods for SW noise sources. Since SW noise sources have quite different features from HW noise sources, it is required to develop entropy estimation methods considering their own properties. Our project contains two different approaches but they will be combined as a single entropy estimation framework eventually.
The first one is analyzing all bytes consisting a noise source differently from the concept of SP 800-90B which chooses the most promising byte and analyzes it. To estimate accurate entropy of each byte, we use some statistical methods from SP 800-90B. After computing min-entropy, the first method updates the estimated entropy by reflecting the Pearson correlation of inter-bytes. We have developed a software tool for estimating entropy for SW noise sources based on the above concepts.
The second one is on the estimation of the entropy of SW noise source using current standard documents CMVP IG and SP 800-90B. Suppose that we collect entropy from system calls provided by OS such as GetTickCount and GetSystemTime. The amount of estimated entropy depends highly on the time interval between adjacent entropy harvests. Also, we need to generalize the statistical tests in AIS.31 since it is difficult to collect required amount of data from the noise sources. We suggest entropy estimation methodologies suitable for SW environments and provide several experimental results showing the accuracy of our methods.