Research Article

Convolution Neural Network-Based Sensitive Security Parameter Identification and Analysis

Table 2

Estimated entropy values for noise sources by byte index through SP800-90B. Target noise sources are mainly used as entropy sources for software cryptographic modules in Windows 10. The smaller the entropy per byte (sum total/number of bytes), the more patterns appear in the image, and vice versa; it becomes increasingly difficult to recognize the pattern. The noise sources CryptGenRandom and CoCreateGuid, commonly used in Windows’s software cryptographic modules, have quite high entropy values and do not differ much. Therefore, even in their images, it seems indistinguishable by the eyes.

Target noiseGetTickCountGetProcessHeapCoCreateGuidCryptGenRandomGetProcessTimes

Byte index00.3030.0007.8637.8837.882
10.0270.0007.8857.8817.008
20.0027.8767.8847.8770.587
30.0007.8527.8827.8820.001
47.8827.8787.888
50.9967.8807.871
60.0007.8867.865
70.0003.9757.880
85.9397.879
97.8817.861
107.8697.879
117.8837.874
127.8627.864
137.8807.889
147.8837.879
157.8817.865
Sum total of 0.33224.606120.211126.01715.478
Average of min-entropy 0.0833.0767.5137.8763.870