Multi-APUF response confusion structure, control method thereof and identity verification system

By designing a multi-APUF response obfuscation structure in identity authentication, and using the responses of multiple APUFs to generate LFSR output, the problem of traditional strong PUFs being vulnerable to attack is solved, and efficient and secure identity authentication is achieved, suitable for IoT devices with limited resources .

CN119094132AActive Publication Date: 2024-12-06ANHUI POLYTECHNIC UNIV

Patent Information

Application Number
CN202411166317.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-06
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

Traditional strong PUFs are susceptible to modeling attacks and machine learning algorithms in identity authentication, and IoT devices with limited resources cannot afford traditional identity authentication solutions.

Method used

A multi-APUF response obfuscation structure is designed, using three responses of five APUFs as the initial seeds of three-level LFSR and the responses of two APUFs as feedback coefficients to generate final responses through iteration and XOR operations to enhance resistance to machine learning Ability to attack.

Benefits of technology

Improve the security and resource utilization efficiency of identity authentication, can effectively resist machine learning attacks such as LR, CMA-ES and ANN, maintain high uniqueness and uniformity, and maintain high reliability under temperature and voltage changes. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119094132A_ABST
    Figure CN119094132A_ABST
Patent Text Reader

Abstract

The invention discloses a multi-APUF response confusion structure and a control method thereof, and an identity verification system, the multi-PUF response confusion structure integrates a plurality of APUF responses in a three-level LFSR, the responses of three APUFs serve as initial seeds to be input into the LFSR, the other two APUFs serve as feedback coefficients of the LFSR cycle process, and in order to achieve the purpose of thorough confusion, the multiple APUFs are input into the three-level LFSR. And finally, the response read when the maximum number of cycles is 7 is responded, so that the resistance to machine learning attacks is greatly improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of identity authentication technology, and more specifically, the present invention relates to a multi-APUF response obfuscation structure and a control method thereof, and an identity authentication system. Background Art

[0002] With the development of the Internet of Things, people's living standards are constantly improving, but at the same time, it also brings a lot of opportunities for hackers. Traditional security mechanisms store keys in electrically erasable programmable read-only memory and non-volatile static random access memory (SRAM), and combine encryption algorithms to achieve information encryption and identity authentication. In order to ensure the security of cryptographic key storage, people have developed tamper-proof devices with multiple countermeasures to resist physical attacks. However, in many IoT devices, resources are limited, so they cannot afford traditional identity authentication solutions.

[0003] Physically Unclonable Function (PUF) is a promising and cost-effective hardware security primitive that has been widely used in devices. A PUF is a circuit whose input is called stimulus and whose output is called response. The combination of stimulus and response is called stimulus-response pair (CRPs). Due to the uncontrollable random variations introduced during the manufacturing process, PUF circuits are considered to be unclonable and unpredictable. Based on the number of CRPs, PUFs can be divided into weak PUFs and strong PUFs. Weak PUFs have a small number of CRPs. ROPUF, SRAM PUF, and Butterfly PUF are typical weak PUFs that can produce reliable CRPs. Due to the limited CRPs of weak PUFs, they cannot be the basis for device authentication. In contrast to weak PUFs, strong PUFs are characterized by a large number of CRPs, and the number of CRPs grows exponentially with the input bit width, such as Arbiter PUF (APUF) and BR PUF. Strong PUFs can be prime candidates for some security applications, such as lightweight authentication, hardware obfuscation, and advanced encryption. Although strong PUF-based protocols provide an effective and lightweight solution for authentication of IoT devices, they are not completely reliable and are vulnerable to some serious attacks such as modeling attacks.

[0004] Since weak PUF has a limited number of stimulus-response pairs (input-output pairs), it is difficult to apply it to identity authentication. Since strong PUF has an exponential number of stimulus-response pairs, strong PUF is used for identity authentication. Since the stimulus-response pairs have a strong correlation, they are easily cracked by robot learning algorithms. Summary of the invention

[0005] The present invention provides a multi-APUF response obfuscation structure, aiming to improve the above-mentioned problem.

[0006] The present invention is implemented as follows: a multi-APUF response obfuscation structure, the structure comprising:

[0007] A first AND gate and a second AND gate, a first APUF to a fifth APUF, and a first flip-flop to a third flip-flop;

[0008] The two input terminals of the first AND gate are respectively connected to the output terminal of the first APUF and the output terminal Q of the third trigger.

[0009] Two input terminals of the second AND gate are respectively connected to the output terminal of the second APUF and the output terminal Q of the third flip-flop;

[0010] A first selector, wherein two input terminals are respectively connected to the output terminal of the third APUF and the output terminal Q of the third trigger, and the output terminal is connected to the input terminal D of the first trigger;

[0011] A first XOR gate, two input terminals of which are respectively connected to the output terminal Q of the first trigger and the output terminal of the first AND gate;

[0012] A second selector, wherein two input ends are respectively connected to the output end of the first XOR gate and the output end of the fourth APUF, and the output end is connected to the input end D of the second trigger;

[0013] A second XOR gate, two input terminals of which are respectively connected to the output terminal Q of the second flip-flop and the output terminal of the second AND gate;

[0014] A third selector, two input ends of which are respectively connected to the output end of the second XOR gate and the output end of the fifth APUF, and an output end of which is connected to the input end D of the third trigger;

[0015] The output terminals of the first trigger and the second trigger are connected to the input terminal of the third XOR gate, and the output terminal of the third XOR gate outputs a response R;

[0016] When the port configuration signal Sel=0, the first selector to the third selector output the third APUF to the fifth APUF output signals respectively, and when the port configuration signal Sel=1, the first selector to the third selector output the other signal respectively.

[0017] The present invention is implemented as follows: a control method based on the above-mentioned multi-APUF response confusion structure, the control method is specifically as follows:

[0018] (1) enabling the first APUF to the fifth APUF and obtaining respective responses from the first APUF to the fifth APUF;

[0019] (2) The Sel selection port is configured as 0, and the first selector selects to output the responses of the third APUF to the fifth APUF to the input terminals D of the first trigger to the third trigger respectively. When CLK passes a pulse, the first trigger to the third trigger respectively saves the responses of the third APUF to the fifth APUF;

[0020] (3) The Sel selection port is configured as 1. Every time CLK passes a pulse, the multi-PUF response confusion structure completes an iteration and updates the state quantities x0, x1, and x2. The state quantity after the jth iteration is

[0021] (4) When the iteration number j=7, the output response R7 of the third XOR gate is selected for output.

[0022] Furthermore, after the j+1th iteration, each state quantity can be shown as follows:

[0023]

[0024] in, are the state quantities output from the output terminals Q of the first to third triggers after the jth iteration, are the state quantities output from the output terminals Q of the first to third triggers after the j+1th iteration, and g0 and g1 are the feedback coefficients output by the first and second AND gates, respectively.

[0025] Furthermore, after the seventh iteration, the final response R7 can be expressed as:

[0026]

[0027] in, They respectively represent the state quantities of the output terminals Q of the first and second triggers after the 7th iteration.

[0028] The present invention is implemented in this way: an identity authentication system, the system comprising:

[0029] The device end is a server, the device end is integrated with the multi-APUF response obfuscation structure as claimed in claim 1, and the server is integrated with the multi-APUF response obfuscation model as claimed in claims 2 to 4;

[0030] In the registration phase, the multi-APUF response obfuscation model is trained to obtain the initial state of the multi-PUF response obfuscation model, including the initial state quantity and feedback coefficients g0, g1;

[0031] In the authentication phase: randomly select the stimulus signal C u , the excitation signal C uand the number of iterations j = 7 acts on the multi-APUF response obfuscation structure in the device end, and obtains the response R′7, the excitation signal C u The multi-APUF response confusion model of the server is applied with the iteration number j=7 to obtain the response R7. If |R7-R′7| is less than the threshold, the authentication is successful.

[0032] Furthermore, the registration process of the registration phase is as follows:

[0033] The server sends an excitation signal C = {c1, c2, ...c n};

[0034] The device side sets the first APUF to the fifth APUF for each stimulus c i Response Send to the server;

[0035] The server responds to the response signal and the corresponding configuration signal (c i , γ i ) Train the PUF model and obtain the initial state of the PUF model, including the initial state quantity And feedback coefficients g0 and g1.

[0036] Furthermore, after registration, the external interfaces of the first APUF to the fifth APUF are burned using the OTP technology.

[0037] Furthermore, the authentication process of the authentication phase is as follows:

[0038] The device sends its own ID to the server for ID retrieval. If the ID exists in the server's registration list, the server will randomly select the configuration stimulus signal C u Send to the device;

[0039] When the device port receives the stimulus signal C u After that, the excitation signal C u Acting on the first APUF to the fifth APUF, obtaining a response R7′ with the iteration number j=7;

[0040] The device sends a cycle number j=7 and a response R7′ to the server, and the server responds to the stimulus signal C u and the initial state of the multi-APUF response obfuscation model stored by the server, computing the response R7 at iteration number j=7;

[0041] If |R7-R′7| is less than the threshold, the authentication is successful and a response signal is sent to the device.

[0042] The present invention adopts a response obfuscation configuration to enhance its ability to resist machine learning attacks. It includes five APUFs, where the responses of three APUFs are used as the initial seeds of the three-stage LFSR, and the responses of the remaining two APUFs are used as feedback coefficients to control the final output of the LFSR and its participation in the feedback loop; in order to maximize the ability to resist machine learning attacks, this structure utilizes the maximum number of cycles (2 n -1)7, generate the final response by iterating and XORing the states of the previous two stages. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the structure of a multi-APUF response obfuscation structure provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of the structure of an identity authentication system based on a multi-APUF response obfuscation structure provided by an embodiment of the present invention;

[0045] Figure 3 A schematic diagram of data interaction during the registration phase provided by an embodiment of the present invention;

[0046] Figure 4 A schematic diagram of data interaction in the authentication phase provided by an embodiment of the present invention;

[0047] Figure 5 A schematic diagram of the uniqueness results of the multi-APUF response obfuscation structure in V7 provided by an embodiment of the present invention;

[0048] Figure 6 A graph showing the deviation test results under one million data provided by an embodiment of the present invention;

[0049] Figure 7 A schematic diagram of the uniqueness of a multi-APUF response obfuscation structure provided by an embodiment of the present invention;

[0050] Figure 8 The reliability of the multi-APUF response obfuscation structure under different factors provided by the embodiment of the present invention, where (a) is voltage and (b) is temperature;

[0051] Fig. 9 The attack results of different attack methods provided in the embodiments of the present invention, where (a) is LR, (b) is CMA-ES, and (c) is ANN. DETAILED DESCRIPTION

[0052] The specific implementation modes of the present invention are further explained in detail below by describing the embodiments with reference to the accompanying drawings, so as to help those skilled in the art to have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0053] Figure 1 A schematic diagram of a structure of a multi-APUF (strong PUF) response obfuscation structure provided in an embodiment of the present invention. For ease of explanation, only the part related to the embodiment of the present invention is shown. The multi-APUF response obfuscation structure includes:

[0054] A first AND gate and a second AND gate, the first APUF to the fifth APUF, and the first trigger to the third trigger, the two input ends of the first AND gate are respectively connected to the output end of the first APUF and the output end Q of the third trigger, and the two input ends of the second AND gate are respectively connected to the output end of the second APUF and the output end Q of the third trigger; a first selector, the two input ends are respectively connected to the output end of the third APUF and the output end Q of the third trigger, and the output end is connected to the input end D of the first trigger; a first XOR gate, the two input ends are respectively connected to the output end Q of the first trigger and the output end of the first AND gate; a second selector, the two input ends are respectively connected to the output end of the first XOR gate and the output end of the fourth APUF, and the output end is connected to the input end D of the second trigger; a second XOR gate, the two input ends are respectively connected to the output end Q of the second trigger and the output end of the second AND gate; a third selector, the two input ends are respectively connected to the output end of the second XOR gate and the output end of the fifth APUF, and the output end is connected to the input end D of the third trigger; the output ends of the first trigger and the second trigger are connected to the input end of the third XOR gate, and the output end of the third XOR gate outputs a response R;

[0055] When the port configuration signal Sel=0, the first selector to the third selector output the third APUF to the fifth APUF output signals respectively, and when the port configuration signal Sel=1, the first selector to the third selector output the other signal respectively.

[0056] x0, x1, and x2 are three initial seeds. The feedback coefficients g0 and g1 are determined by the responses of the first PUF and the second PUF. When the coefficient is 1, the state value of the last stage is XORed with the state value of this stage. This function is implemented by inputting the response of the PUF and the state value of the last stage into the AND gate, so that the response of the PUF determines whether the last stage participates in the feedback. The selector here has two functions: one is to control the initial seed input of the D flip-flop and save the seed on the rising edge provided by CLK; the other is to enable the path of the LFSR to ensure its normal operation. In this configuration, the D flip-flop is used to save the state value of each stage, and the XOR gate operates on these state values. The control flow of the multi-APUF response confusion structure is as follows:

[0057] (1) enabling the first APUF to the fifth APUF and obtaining respective responses from the first APUF to the fifth APUF;

[0058] (2) The Sel selection port is configured as 0, and the first selector selects to output the responses of the third APUF to the fifth APUF to the input terminals D of the first trigger to the third trigger respectively. When CLK passes a pulse, the first trigger to the third trigger respectively saves the responses of the third APUF to the fifth APUF;

[0059] (3) The Sel selection port is configured as 1. Every time CLK passes a pulse, the multi-PUF response confusion structure completes an iteration and updates the state quantities x0, x1, and x2. The state quantity after the jth iteration is

[0060] (4) When the number of iterations j=7, the output response R of the third XOR gate is selected for output.

[0061] The new state quantities x0, x1, and x2 are the state quantities outputted from the output terminals Q of the first to third triggers respectively, and g0 and g1 are the feedback coefficients outputted from the first and second AND gates respectively.

[0062] The expressions of the first APUF to the fifth APUF5 can be expressed by the linear additive delay model of the APUF, and the responses generated by the five APUFs are:

[0063]

[0064] Wherein, P is the characteristic vector of APUF, which is only related to the excitation, and ω1 to ω5 are the delay parameters of the first APUF to the fifth APUF respectively; are the initial values ​​of the state quantities x0, x1, and x2 respectively;

[0065] The initial state quantities are obtained from the five PUFs. And the feedback coefficients (g0, g1), the state of the overall LFSR (linear feedback shift register) is recorded as The state after j LFSR iterations is S j , its iterative function is represented by T, and the j-th expression is as follows:

[0066] S j =T j (S 0 )(2)

[0067] An n-level LFSR can only traverse at most 2 n -1 state, the present invention is a three-stage LFSR, therefore, the maximum number of iterations j = 2 3 -1-7.

[0068] The following discusses the expression of the XOR gate of the AND gate. According to the truth table of the AND gate, the expression of the value v after m and n are ANDed is the multiplication of m and n, and the expression is: v = m & n = mn; similarly, the expression of m and n through the XOR gate is: v = m^n = m + n - 2mn;

[0069] After the j+1th iteration, S j+1 The various status values ​​can be as follows:

[0070]

[0071] After the seventh iteration, the final response R can be expressed as:

[0072]

[0073] in, Respectively represent the state quantities x0 and x1 after the 7th iteration;

[0074] Figure 2 A structural diagram of an identity authentication system flow based on a multi-APUF response obfuscation structure is provided for an embodiment of the present invention. For ease of description, only the part related to the embodiment of the present invention is shown. The system includes:

[0075] The device side and the server, the device side is integrated with the above-mentioned multi-APUF response obfuscation structure, and the server is integrated with a multi-APUF response obfuscation model;

[0076] During the registration phase, the server trains the multi-APUF response obfuscation model to obtain the initial state of the multi-PUF response obfuscation model, including the initial state quantity and feedback coefficients g0, g1;

[0077] In the authentication phase: randomly select the stimulus signal C u , the excitation signal C u and the number of iterations j = 7 acts on the multi-APUF response obfuscation structure in the device end, and obtains the response R′7, the excitation signal C u The multi-PUF response confusion model with iteration number j=7 is applied to the server to obtain response R7. If |R7-R′7| is less than the threshold, the authentication is successful.

[0078] The following is a detailed description of the registration process in the registration phase and the authentication process in the authentication phase, as follows:

[0079] (1) The registration process during the registration phase;

[0080] Traditional strong PUF needs to store a large number of CRPs (stimulus response pairs) in the server database, which wastes server resources to a certain extent. The present invention selects APUF to model through the linear additive delay model, and the storage model is much more resource-saving than storing CRPs.

[0081] During the registration phase, Figure 3 As shown, the server first sends a sufficient number of incentives C = {c1, c2, ...c n}, the device will five APUF for each stimulus c i Response The server then sends the response signal and the corresponding configuration signal (c i , Υ i ) Train the PUF model and obtain the initial state of the PUF model, including the initial state quantity and feedback coefficients g0, g1;

[0082] It should be noted that after registration, the external interface of each APUF needs to be burned. OTP technology can be used to burn it, so as to prevent criminals from collecting CRPs for each APUF. After registration is completed, the server port will send an ACK signal to indicate that the registration is completed. When the device receives the ACK signal, it will start OTP to burn the external interface of each APUF.

[0083] (2) The certification process during the certification phase;

[0084] During the authentication phase, Figure 4 As shown, the device first needs to send its own ID to the server for ID retrieval. If it is in the data, the configuration excitation signal C will be randomly selected. u The device port receives the excitation signal C u After that, the excitation signal C u Acting on the first to fifth APUFs in the multi-PUF response obfuscation structure, the response R7′ with the iteration number j=7 is obtained. The device sends the cycle number j=7 and the response R7′ to the server. The server first responds to the excitation signal C u And the initial state of the PUF model stored in the server, through the iteration number j = 7 and formulas (3) ~ (4), the server's response R7 is calculated. R'7 is compared with R7. If |R7-R'7| is greater than a threshold T, the authentication is canceled. On the contrary, if |R7-R'7| is less than the threshold, the authentication is successful and a response signal is sent to the device to inform the device that it can communicate.

[0085] Uniformity: The consistency of a PUF response is determined by the balance of 1s and 0s. A significant imbalance makes it easier for an attacker to predict the response. In fact, if the ratio of "1 / 0" is 50%, the consistency of the PUF is ideal. The consistency of a PUF can be measured by the average Hamming weight (HW) of the response sequence, which is defined by the following formula:

[0086]

[0087] Among them, r i Represents the output of APUF for the i-th excitation signal.

[0088] The proposed multi-PUF response obfuscation structure was implemented using two Xilinx Virtex-7 boards at 1.0V and 25℃. In each PUF, 10,000 CRPs were measured. Figure 5 The histogram of uniformity shows that the average uniformity of the proposed structure is 50.19% and the standard deviation is 1.9%.

[0089] Deviation test: The security characteristics of PUF rely on the randomness of the generated response. The complexity of an attacker trying to exploit or destroy a PUF depends mainly on the degree of randomness of its stimulus-response mapping. At the same time, the proposed structure is tested to prevent the generated response from containing lengthy sequences of "1" or "0". The experimental results are shown in Figure 2. Figure 6 As shown in the figure, black and white pixels represent "1" and "0", respectively. This shows that the response distribution is uniform, reflecting that the response generated by this structure has good statistical properties.

[0090] Uniqueness: Uniqueness is measured by evaluating how a particular PUF circuit differs between different devices. Ideally, when the same PUF circuit is implemented on multiple devices, it should produce an average Hamming distance (HD) of approximately 50% when comparing the responses produced by the same stimulus. The HD between different PUFs is defined as follows:

[0091]

[0092] Among them, R i [n] and R j [n] represents the response of the i-th PUF and the j-th PUF under the same stimulus, and N represents the total number of bits of the response. For m PUFs with N-bit responses, the average Hamming distance is usually used to measure their uniqueness, as shown in formula (7):

[0093]

[0094] Ten different structures were used, so m is 10 and N is 10,000. The responses generated at a standard voltage of 1.0V and room temperature of 25°C were statistically analyzed. The final statistical graph is as follows: Figure 7 As shown in Figure 2, the uniqueness of the multi-PUF response obfuscation structure is 49.90, with a standard deviation of 0.5%.

[0095] Reliability: Ideally, a PUF within a device should perfectly reproduce its output. However, environmental factors such as temperature and voltage fluctuations may affect the output of the PUF, so a reliability metric is needed to evaluate the consistency of the PUF design in reproducing the response. For a device, reliability is obtained by finding N response samples under baseline conditions and comparing them with N responses under different operating conditions and baseline conditions to obtain the on-chip average HD, which is defined as follows:

[0096]

[0097] Where K represents a PUF that obtains K N-bit responses from the same stimulus, R i,p Represents R i Sample responses under different circumstances.

[0098] Voltage and temperature experiments were carried out on the structure. Figure 8 (a) and (b) are dot-line graphs of two experiments. In the voltage experiment, 1.0V was used as the reference voltage and the voltage range was ±10%. In the temperature experiment, 25℃ was used as the reference temperature and the experiment was conducted between 20-80℃. In the voltage experiment, the reliability was 93.60%. In the temperature experiment, the reliability was 96.58%.

[0099] Modeling attack test: Three machine learning methods were used: LR, CMA-ES, and ANN. In our experiments, LR uses the Rprop iterative function, CMA-ES uses the fitting function of the average Hamming distance, and ANN uses the logical approximation method of machine learning attack. During the model training process, 70% of the CRP dataset was randomly assigned to the training set and the remaining 30% was assigned to the test set. APUF, 2-XOR-APUF, 3-XOR-APUF, and MPUF were replicated for comparison. The experimental results are shown in Figure 2. Fig. 9 As shown in (a), (b) and (c).

[0100] From the above results, we can see that in the LR attack, APUF has been completely broken. In the CMA-ES attack based on high reliability, 2-XOR-APUF and 3-XOR-APUF have become unsafe as the data set grows. In the ANN attack, MPUF has also been broken, and the prediction rate is close to 90% under a CRP of 20,000. It can be seen that in the three attacks, the multi-PUF response obfuscation structure can effectively resist the three attacks.

[0101] The present invention studies and collects PUFs implemented on advanced FPGAs in recent years, and makes a comprehensive comparison of the proposed PUFs. As shown in Table 1, the reliability of the PUF proposed by the present invention is lower than that of other works, but higher than that of CAPUF. AP . This is the result of integrating multiple APUFs, where unstable factors will be magnified in the overall structure. Since the structure adopts a method of combining five different APUFs with each other, one of the same stimuli responds differently on different PUFs, which can also affect the difference in the final output response of the overall structure. Compared with other structures, the uniqueness of this structure is the highest. Uniformity Due to the strict control of the layout constraints of the FPGA board, we use a NAND gate-based SR latch circuit as an arbitrator to alleviate the routing tilt introduced between the switch chain and the arbitrator, and use PDL as an adjustment unit to make the ratio of '0' and '1' for each APUF the same. The uniformity of this structure is similar to other variants that also use APUFs such as CA PUF AP Compared with FLAM PUF, its uniformity is higher than both. And in the whole comparison, its uniformity is also the highest.

[0102] Table 1 Performance comparison with the most advanced PUF

[0103]

[0104]

[23] M.Thirumoorthi, M.Jovanovic, M.Mirhassani and M.Khalid, "Design andEvaluation of a Hybrid Chaotic-Bistable Ring PUF," in IEEE Transactions onVery Large Scale Integration(VLSI)Systems, vol.29, no.11, pp.1912-1921, Nov.2021, doi:10.1109 / TVLSI.2021.3111588.

[0105]

[12] H.Nassar,L.Bauer and J.Henkel,"CaPUF:Cascaded PUF Structure forMachine Learning Resiliency,"in IEEE Transactions on Computer-Aided Design ofIntegrated Circuits and Systems,vol.41,no.11,pp.4349-4360,Nov.2022,doi:10.1109 / TCAD.2022.3197539.

[0106]

[14] L.Wu,Y.Hu,K.Zhang,W.Li,X.Xu and W.Chang,"FLAM-PUF:A Response–Feedback-Based Lightweight Anti-Machine-Learning-Attack PUF,"in IEEETransactions on Computer-Aided Design of Integrated Circuits and Systems,vol.41,no.11,pp.4433-4444,Nov.2022,doi:10.1109 / TCAD.2022.3197696.

[0107]

[29] Gu,C.,Liu,W.,Cui,Y.,Hanley,N.,Oneill,M.,&Lombardi,F.(2019).AFlip-Flop Based Arbiter Physical Unclonable Function(APUF)Design with HighEntropy and Uniqueness for FPGA Implementation.IEEE Transactions on Emerging.

[0108]

[30] Ali,Rashid,et al."A machine learning attack-resilient strong PUFleveraging the process variation of MRAM."IEEE Transactions on Circuits andSystems II:Express Briefs 69.6(2022):2712-2716.

[0109]

[31] ALogic Encryption-enhanced PUF Architecture to Deceive MachineLearning-based Modeling Attacks.

[0110] At the same time, the proposed structure was tested for machine learning under 150,000 CRPs, and the results are presented in Table 2. It can be seen that under 5,000 CRPs, APUF has been broken by LR, CMA-ES, and ANN. When using more CRPs, the prediction rate of this structure is closer to the ideal value of 50% than previous work, and has good resistance to the three algorithms.

[0111] Table 2 Comparison of anti-machine learning capabilities

[0112]

[0113]

[15] YJCui, CYGu, QQMa, Y.Fang, CHWang, M.O'Neill and WQLiu, "Lightweight Modeling Attack-Resistant Multiplexer-Based Multi-PUF (MMPUF) Design on FPGA," Electronics, vol.9, no.5, pp.815:1-815:21, May 2020.

[0114]

[13] Y.Gao et al., "Obfuscated challenge-response: A secure lightweight authentication mechanism for PUF-based pervasive devices," in Proc.IEEEInt.Worksh.on Secur.,Priv.and Trust for IoT, Sydney, Australia, Mar. 2016, pp.1–6.

[0115] Aiming at the problem that traditional PUF is difficult to resist machine learning in identity authentication, this paper proposes a multi-PUF response obfuscation structure based on LFSR, which can resist machine learning. This technology uses five APUFs, of which the responses of three APUFs are used as initial seeds, and the remaining two are used as feedback coefficients to control the on-off state of the feedback path. Based on the above experimental results, it is shown that this structure has high uniqueness and uniformity, which are 49.90% and 50.19% respectively. Its reliability under temperature and voltage also reaches 91.10% and 94.29% respectively.

[0116] The present invention has been described exemplarily. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, or the concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A multi-APUF response obfuscation structure, It is characterized in that The structure comprises: A first AND gate and a second AND gate, a first APUF to a fifth APUF, and a first flip-flop to a third flip-flop; The two input terminals of the first AND gate are respectively connected to the output terminal of the first APUF and the output terminal Q of the third trigger. Two input terminals of the second AND gate are respectively connected to the output terminal of the second APUF and the output terminal Q of the third flip-flop; A first selector, wherein two input terminals are respectively connected to the output terminal of the third APUF and the output terminal Q of the third trigger, and the output terminal is connected to the input terminal D of the first trigger; A first XOR gate, two input terminals of which are respectively connected to the output terminal Q of the first trigger and the output terminal of the first AND gate; A second selector, wherein two input ends are respectively connected to the output end of the first XOR gate and the output end of the fourth APUF, and the output end is connected to the input end D of the second trigger; A second XOR gate, two input terminals of which are respectively connected to the output terminal Q of the second flip-flop and the output terminal of the second AND gate; A third selector, two input ends of which are respectively connected to the output end of the second XOR gate and the output end of the fifth APUF, and an output end of which is connected to the input end D of the third trigger; The output terminals of the first trigger and the second trigger are connected to the input terminals of the third XOR gate, and the output terminal of the third XOR gate outputs a response R.

2. A control method based on the multi-APUF response obfuscation structure of claim 1, It is characterized in that When the port configuration signal Sel=0, the first selector to the third selector output the third APUF to the fifth APUF output signals respectively, and when the port configuration signal Sel=1, the first selector to the third selector output the other signal respectively.

3. A control method based on the multi-APUF response obfuscation structure of claim 1, It is characterized in that The control method is specifically as follows: (1) enabling the first APUF to the fifth APUF and obtaining respective responses from the first APUF to the fifth APUF; (2) The Sel selection port is configured as 0, and the first selector selects to output the responses of the third APUF to the fifth APUF to the input terminals D of the first trigger to the third trigger respectively. When CLK passes a pulse, the first trigger to the third trigger respectively saves the responses of the third APUF to the fifth APUF; (3) The Sel selection port is configured as 1. Every time CLK passes a pulse, the multi-PUF response confusion structure completes an iteration and updates the state quantity x 0 、x 1 、x 2 , the state after the jth iteration is (4) When the number of iterations j = 7, the output response of the third XOR gate is selected to be R 7 Output.

4. The control method of the multi-APUF response obfuscation structure as claimed in claim 3, Its characteristic is that After the j+1th iteration, the various state quantities can be shown as follows: in, are the state quantities output from the output terminals Q of the first to third triggers after the jth iteration, are the state quantities of the output terminals Q of the first to third triggers after the j+1th iteration, g 0 , g 1 are the feedback coefficients of the first AND gate and the second AND gate outputs respectively.

5. The control method of the multi-APUF response obfuscation structure as claimed in claim 3, It is characterized in that After the seventh iteration, the resulting response R 7 Expressed as: in, They respectively represent the state quantities of the output terminals Q of the first and second triggers after the 7th iteration.

6. An identity verification system, It is characterized in that The system comprises: The device end is a server, the device end is integrated with the multi-APUF response obfuscation structure as claimed in claim 1, and the server is integrated with the multi-APUF response obfuscation model as claimed in claims 2 to 4; During the registration phase, the server trains the multi-APUF response obfuscation model to obtain the initial state of the multi-APUF response obfuscation model, including the initial state quantity And the feedback factor g 0 , g 1 ; In the authentication phase: randomly select the stimulus signal C u , the excitation signal C u and the number of iterations j = 7 acts on the multi-APUF response obfuscation structure in the device end to obtain the response R 7 ′, excitation signal C u The multi-PUF response confusion model of the server is applied with the iteration number j=7 to obtain the response R 7 , if |R 7 -R′ 7 | is less than the threshold, the authentication is successful.

7. The identity verification system according to claim 6, It is characterized in that The registration process of the registration phase is as follows: The server sends an excitation signal C to the device. 1 、c 2 ,...c n }; The device side sets the first APUF to the fifth APUF for each stimulus c i Response Send to the server; The server responds to the response signal and the corresponding configuration signal (c i , Υ i ) Train the PUF model and obtain the initial state of the PUF model, including the initial state quantity And the feedback factor g 0 , g 1 .

8. The identity verification system according to claim 7, It is characterized in that After registration, the external interfaces of the first APUF to the fifth APUF are burned using the OTP technology.

9. The identity verification system according to claim 6, It is characterized in that The authentication process in the authentication phase is as follows: The device sends its own ID to the server for ID retrieval. If the ID exists in the server's registration list, the server will randomly select the configuration stimulus signal C u Send to the device; When the device port receives the stimulus signal C u After that, the excitation signal C u Acting on the first APUF to the fifth APUF, the response R with the iteration number j=7 is obtained 7 ′; The device sends the number of cycles j = 7 and responds to R 7 ′, sent to the server, the server responds to the excitation signal C u And the initial state of the multi-APUF response obfuscation model stored in the server, calculate the response R at iteration number j = 7 7 ; If | R 7 -R′ 7 |If it is less than the threshold, the authentication is successful and a response signal is sent to the device.

Citation Information

Patent Citations

  • Dynamic multi-key obfuscation PUF structure and authentication method thereof

    CN109005040A

  • Statistics and modeling attack resistant circuit structure based on response feedback PUF and response generation method

    CN118432848A

  • Reliable multi-information entropy physical unclonable function (PUF) for internet of things security

    US20230092828A1

Cited By

  • Physical unclonable function (PUF) circuit based on XOR iteration and operation method thereof

    CN119848948A