joshuaehill
commented
1 year ago You can't form a restart data set using a single large sample. Instead, you need to follow the instructions in the paragraph that you have highlighted, and take the first 1000 (for your noise source, single bit) raw samples from each of 1000 distinct entropy source restarts. For hardware, a "restart" generally involves power cycling the hardware between taking each 1000 symbol sample. The first bullet under you highlighted text tells you how to represent this in a file.
For the NIST tool, you only need to provide the row dataset, as the column dataset is generated by the tool internally using the row dataset.
The particulars of the formatting each sample from the noise source are the same as for the large raw data sample: you need one raw sample per byte; for your source (as it produces a bitstring), this will be a sequence of single byte integers (uint8_t), where each value is either 0x00 or 0x01, each representing one bit of output from your noise source.
The data used to form the restart data is described in the first bullet: the data file should contain the first 1000 symbols output from the noise source in the first restart in the order they were produced, followed by the first 1000 symbols output from the noise source in the second restart in the order they were produced, followed by the first 1000 symbols output from the noise source in the third restart in the order they were produced, and so on, until you finish the file with the first 1000 symbols output from the noise source in the 1000th restart in the order they were produced. The result will be a single file with 1 million bytes which you use with ea_restart.
Implementing this as a matrix is unlikely to be helpful, unless you are trying to implement the tool yourself (whence, it may be useful, as the column dataset is just the transpose of the row dataset).
xiaoyaowana
I have a little difficulty in understanding this sentence “the row dataset is M[1][1] ||...|| M[1][c] || M[2][1] ||...|| M[2][c] || ... || M[r][1] ||...|| M[r][c]”.What consists of the M[1][1]..or M[r][c](element of matrix M).As I noted in the figure，I'm confused about the elements of the matrix.
I may have mistaken the format description of the input file and output file, but unfortunately, I have not yet obtained the output file.
I read the document of "row dataset" format described in SP800-90B Section 3.1.4.1.
I have a 0-1 binary bit stream of 10M in total. Looking at the document, is it means that this bit stream should be constructed into a 1000 1000 matrix? In Python, that is reshape M.(1000，1000), where M is the bitstream data. But I take this 1000 1000 matrix as input for earestart and still fails. I'm sorry, I don't know the right anwser for the parameter of "bits per_ symbol" ,but I tried every number from 1 to 8 and failed. I hope to get a reply about what format of bitstream can run through ea_restart program, thank you!(my bitstream already pass the ea_iid program)