Font Size: a A A

Design And FPGA Implementation Of Indefinite Model Based APUFs With High Reliability

Posted on:2020-04-26Degree:MasterType:Thesis
Country:ChinaCandidate:F K DanFull Text:PDF
GTID:2428330626950761Subject:Integrated circuit engineering
Abstract/Summary:PDF Full Text Request
As a new cryptography primitive,Physically Unclonable Functions(PUFs)are used to extract the manufacturing process variations as keys or authentication information,which effectively resist physical attacks.PUFs are lightweight and low cost circuit.Therefore,it has become an important part of the security field.Strong PUFs can generate a large number of challenge-response,and are mainly used in authentication systems.However,with the development of attack technology,side channel attack,fault injection attack and machine learning attack have become the threats of PUFs.And machine learning attack has become the most important threat of strong PUFs.As a typical strong PUFs,APUFs has a prediction accuracy rate of more than 99%.In order to resist machine learning attacks,an IM-APUFs(Indefinite-Model Based Arbiter PUFs)is proposed in this thesis,which is composed of 4-1 MUX and inverter.Challenges are used to configurate four siganal pathway in IM-APUFs.Compared with the standard APUFs circuit,2-1mux is replaced with 4-1mux in IM-APUFs.The model of IM-APUFs is indefinite based on the substantial change,which can effectively resist machine learning attacks.Simultaneously,inverters are inserted in IM-APUFs,which effectively improved the uniqueness.The majority voting and fault-tolerant protocol are used to guarante the accuracy of authentication.The experimental results show that the machine learning prediction accuracy of IM-APUFs decreases from about 95%of the standard APUFs to about 52%,the uniqueness increases from about 22%to about 43.6%,the reliability is about 94.34%,the authentication error rate is less than 1.26×10-8,and the uniformity and bit-aliasing are nearly 50%.Compared with standard APUFs,the uniformity,bit-aliasing,uniqueness and security of IM-APUFs have been greatly improved.Simultaneously,IM-APUFs has certain advantages in security and low resource consumption compared with the structure or scheme against machine learning attack.The fault-tolerant protocol based on IM-APUFs shows that it meets the requirement of authentication.Therefore,the proposed IM-APUFs can be widely used in lightweight security authentication system,which has great practical significance.
Keywords/Search Tags:Physically Unclonable Functions, Modeling attack, Machine Learning, Field Programmable Gate Array, Indefinite-Model Based Arbiter PUFs
PDF Full Text Request
Related items