Design Of Entropy Extractor Based On SRAM PUFs

With the development of the Internet,the information security is quite important,whether it is the encryption of personal information,or the protection of trade secrets in the business field.Physical Unclonable Functions(PUFs)can satisfy the requirement of the security to a certain extent.SRAM PUFs have the advantages of simple structure,easy to implement,good anti-prediction and so on.However,the oscillating characteristics of cells in the SRAM is not clear.The existing architecture of entropy extractors based on SRAM PUFs is not efficient in compressing large numbers of cell data.If the cell information is leaked,.the entropy source in existing architecture cannot be replaced quickly.An efficient design of entropy extractor based on SRAM PUFs is proposed in this paper.Only the noisy cells of high min-entropy are selected to generate full entropy seeds in this design.Therefore,it can reduce the amount of data to be processed significantly and improve the efficiency of seed generation.The oscillating characteristics of the noisy cells inside SRAM are measured,and screening strategies for filtering out the selected noisy cells inside the SRAM are put forward.Finally,based on the strategies,two entropy extractors(based on different conditioning algorithms)are designed and implemented on FPGA,which can generate full entropy seeds.The length of seeds generated by this design is 256bit.The number of the selected cells which are used to generate seeds is from 0.5%to 4%of all SRAM cells.Compared with the architecture for generating seeds by using all SRAM cells’ start-up values,it is shown that the proposed design in this paper is efficient and widely applicable.This design can provide stable full entropy seeds for other modules.This work will greatly accelerate the application of PUFs in real life scene,and have great application prospect.