CN105978694B - The strong physics unclonable function device and its implementation of anti-modeling attack - Google Patents

Abstract

The present invention is applicable to the fields of information security and integrated circuit technology, and provides a strong physical unclonable function device that is resistant to modeling attacks, including: a Boolean confusion module, which is used to pass input excitation through multiple weak physical unclonable functions and Boolean logic After the components are reprocessed, the output response realizes the unpredictable Boolean logic relationship; the stimulus division module is used to divide the input stimulus into valid stimulus and invalid stimulus; the attack detection module is used to detect the invalid stimulus and identify the modeling attack, and process all The invalid incentive and the modeling attack; a response calculation module, used to perform response calculation on the effective incentive through a strong physical unclonable function device. Also provided is an implementation method of a strong physical unclonable function device based on anti-modeling attack. Thereby, while ensuring the randomness and stability of the strong physical unclonable function device, the present invention can actively detect and passively defend against modeling attacks that seriously threaten the safety of the strong physical unclonable function device.

Description

Strong Physical Unclonable Function Device Against Modeling Attacks and Its Implementation Method

technical field

The invention relates to the field of information security and the field of integrated circuits, in particular to a hardware security design, in particular to a strong physical unclonable function device resistant to modeling attacks and an implementation method thereof.

Background technique

With the widespread use of electronic devices, security and privacy have become important issues. The key that is considered to be permanently stored and unknown to the attacker is the core of traditional cryptography. However, many attack methods have been able to crack the key, which makes the key insufficient to ensure security. In order to effectively solve the security problem, the Physical Unclonable Function (Physical Unclonable Function, PUF) came into being, which is a hardware component that can deal with the security problem more effectively.

The PUF uses the inevitable process deviation during chip manufacturing to generate a specific input-output pair, also known as a Challenge-Response Pair (CRP). Even with the same circuit design, the process deviation in the manufacturing process makes the PUF of different chips face the same input stimulus, which may produce different output responses, that is, different CRP. Since the process deviation itself is difficult to control and predict, these CRPs can neither be predicted before PUF fabrication nor replicated after PUF fabrication. This has a greater advantage over traditional keys. This characteristic of PUF makes it widely used in the field of security, such as intellectual property protection, authentication, authentication, identification, etc.

Broadly speaking, PUFs can be divided into two categories: weak PUFs and strong PUFs. The strength and weakness here do not refer to their security level, but the number of CRP, and their characteristics are as follows.

A weak PUF has only a small amount of CRP, and in most cases a weak PUF has only one CRP. For example, the dielectric particle layer PUF is a weak PUF. During manufacture, a layer of dielectric particles is randomly sprinkled. Because their distribution is difficult to predict, the dielectric particle layer PUF is determined according to the randomly covered dielectric particle layer. The size of the capacitance produces a response. As another example, Static Random Access Memory (SRAM) PUF is another kind of weak PUF. Due to the influence of process deviation, each SRAM unit has different electrical characteristics. When the chip is powered on, different SRAM cells will randomly and independently store 0 or 1, and naturally form a CRP. Some other storage units such as flash memory, dynamic random access memory, memristor, etc. also have similar characteristics, so they can be used to construct weak PUFs. Since the weak PUF based on the storage unit will only generate a response when it is powered on, it is more secure than a key stored in a non-volatile memory. Since weak PUFs only have a small amount of CRP, their CRP generally has a dedicated one-time security channel for manufacturers to obtain after the chip is manufactured, and then it is difficult for attackers to steal CRP through this channel again.

Compared with weak PUFs, strong PUFs possess a large amount of CRP. Arbitration PUF is a typical strong PUF. It determines the response by comparing the delay of two path propagation jumps. Each path is composed of multiple sub-paths, and the selection of sub-paths is determined by incentives. Different incentives The two constructed paths are different, and their delays are also different, resulting in random responses. In order to improve the security of arbitration PUF, arbitration PUF has many other extensions, such as feedforward arbitration PUF, XOR arbitration PUF, lightweight arbitration PUF, current mirror PUF, etc. Feedforward arbitration PUF guides the selection of sub-paths in the remaining paths by comparing the delay of some paths; XOR arbitration PUF XORs the responses of multiple arbitration PUFs as the final response; lightweight arbitration PUF still uses XOR gates The XOR calculation is performed on the responses of multiple arbitration PUFs, but multiple response bits are obtained at one time; the current mirror PUF introduces current instead of time delay for comparison. The ring oscillator PUF is another strong PUF. Its response is determined by comparing the frequency of different ring oscillators, and the compared ring oscillator is determined by the excitation. However, the hardware overhead of the ring oscillator PUF is large, and the number of CRP Not as good as the arbitration PUF. Since a strong PUF has a large number of CRPs, such as 10 ³⁸ CRPs, and it is impossible for an attacker to read all the CRPs within a reasonable time, such as 100 years, and does not know which CRPs will be used in actual applications, the The CRP access of strong PUF is generally not restricted. After the chip is manufactured, the manufacturer will select a specific CRP for subsequent applications through the CRP access port.

With the increasing research and application of PUF, the security of PUF is also seriously threatened. The modeling attack method based on machine learning seriously threatens the security of strong PUF such as arbitration PUF. Since a strong PUF has a large number of CRPs, if an independent circuit is designed for each CRP, obviously the hardware overhead is huge. Therefore, there is a certain correlation between different CRPs of a strong PUF. The modeling attack uses machine learning to infer this Association, thereby cracking the CRP of the strong PUF. The modeling attack first establishes a CRP model with relevant physical properties as unknowns based on the circuit structure of the strong PUF, and then uses machine learning to infer these relevant physical properties from part of the CRP data obtained. Predict the relevant physical properties, predict its response, and realize the cracking of the strong PUF. Commonly used machine learning methods include support vector machines, logistic regression, and evolutionary strategies. Using modeling attack to crack the arbitration type PUF and ring oscillation PUF, the CRP prediction accuracy can reach more than 99%. Even though feed-forward arbitration PUF, XOR arbitration PUF, lightweight arbitration PUF and current mirror PUF make the structure of strong PUF more complex, the average CRP prediction accuracy obtained by using modeling attacks can still reach more than 90%, which shows that strong PUF is safe. Sexuality is seriously threatened.

In summary, there are obviously inconveniences and defects in the actual use of the prior art, so it is necessary to improve it.

Contents of the invention

In view of the above-mentioned defects, the purpose of the present invention is to provide a strong physical unclonable function device and its implementation method against modeling attacks, the purpose is to ensure the randomness and stability of the strong physical Passive defense against modeling attacks that seriously threaten the security of strong physical unclonable function devices, thus effectively resisting modeling attacks.

In order to achieve the above object, the present invention provides a strong physical non-clonable function device resistant to modeling attacks, including:

The Boolean confusion module is used to output the response after the input stimulus is reprocessed by multiple weak physical unclonable function devices and Boolean logic elements, so as to realize the unpredictable Boolean logic relationship;

An incentive division module is used to divide input incentives into valid incentives and invalid incentives;

An attack detection module, configured to detect the invalid stimulus and identify a modeling attack, and process the invalid stimulus and the modeling attack;

The response calculation module is used to calculate the response of the effective stimulus through the strong physical unclonable function device.

According to the strong physical unclonable function device of the present invention, the Boolean confusion module includes:

The weak PUF sub-module is used for the weak physical unclonable function device to process the input excitation to obtain Boolean logic configuration bits;

The Boolean determination sub-module is used to input the response of the weak physical unclonable function device, and reprocess the response through the determined input-output Boolean logic relationship to obtain an output response; and/or

The manufacturer obtains the response of the weak physical unclonable function device through a one-time secure channel to obtain the actual Boolean logic of the Boolean obfuscation module.

According to the strong physical unclonable function device of the present invention, the incentive division module includes:

The division rule submodule is used to define the division rules of the effective incentives and invalid incentives;

A division execution submodule, configured to compare the input stimulus with the division rule to obtain the category of the input stimulus;

The division rules include: the input stimulus is divided into a valid stimulus set and an invalid stimulus set; the valid stimulus set is the input set legally used by the strong physical unclonable function device in normal applications, and the invalid stimulus set is the strong physical A set of inputs that are illegally used by a non-clonable function device in normal applications.

According to the strong physical unclonable function device of the present invention, the incentive division module is realized by the hardware circuit resources of the Boolean obfuscation module, and the division rule is determined according to the type of the Boolean obfuscation module;

When the input excitation division module is a switch type Boolean confusion module or a switch normally open type Boolean confusion module, the Boolean confusion modules are connected in series, and the division rules include:

Divide the input stimuli C ₁ to C _4m , the response of the weak PUF sub-module and the input value together determine the effective value or HiZ of the output; if the input stimuli enables the transition from T ₀ to After the switch-type Boolean confusion modules S _B1 , S _B2 ,..., S _Bm finally reach T ₁ , the input excitation is an effective excitation; if the input excitation makes the transition from T ₀ unable to pass through the switch Type Boolean confusion modules S _B1 , S _B2 ,...,S _Bm finally reach T ₁ , it is an invalid incentive;

When the Boolean obfuscation module is a switch type Boolean logic obfuscation module, the switch of the Boolean obfuscation module is controlled by the output control byte of the weak PUF sub-module.

According to the strong physical unclonable function device of the present invention, the attack detection module includes:

The incentive counting submodule is used to count the invalid incentives according to the judgment result of the input incentives;

The attack processing sub-module is configured to trigger processing to deal with the attack when the number of invalid incentive counts reaches the attack threshold.

According to the strong physical unclonable function device of the present invention, the response calculation of the response calculation module to the input stimulus is generated by comparing the time delays of two path propagation jumps, and the sub-paths of each path are determined by the input Incentive decisions;

The processing of the propagation process of the transition through the path by the response calculation module may also combine and reuse hardware circuit resources of the Boolean obfuscation module.

The present invention provides a method for implementing a strong physical unclonable function device based on anti-modeling attacks, including:

The Boolean obfuscation step is to output the response after the input stimulus is reprocessed by multiple weak physical unclonable function devices and Boolean logic elements, so that the Boolean logic relationship is unpredictable;

Divide the incentive steps, and divide the input incentives into valid incentives and invalid incentives;

Detecting an attack step, detecting the invalid stimulus to identify a modeling attack, processing the invalid stimulus and the modeling attack;

In the response calculation step, the response calculation is performed on the effective stimulus through a strong physical unclonable function device.

According to the implementation method of the present invention, the Boolean confusion step also includes:

The weak physical unclonable function device processes the input stimulus to obtain Boolean logic configuration bits;

Input the response of the weak physical unclonable function device, and reprocess the response through the determined input-output Boolean logic relationship to obtain an output response; and/or

The manufacturer obtains the response of the weak physical unclonable function device through a one-time secure channel to obtain the actual Boolean logic.

According to the implementation method of the present invention, the response calculation step also includes:

The calculation of the response of the input stimulus is generated by comparing the time delays of the propagation jumps of the two paths, the sub-paths of each of the paths are determined by the input stimulus, and processing the propagation process of the jumps through the paths can also be Reusing the hardware circuit resources of the Boolean obfuscation step.

According to the implementation method of the present invention, the input excitation division step includes:

defining rules for dividing said effective incentives and ineffective incentives;

comparing the input stimulus with the division rule to derive the category of the input stimulus;

The division rules include: the input stimulus is divided into a valid stimulus set and an invalid stimulus set; the valid stimulus set is the input set legally used by the strong physical unclonable function device in normal applications, and the invalid stimulus set is the strong physical The set of inputs illegally used by the non-clonable function device in normal applications;

The attack detection steps include:

counting the invalid stimuli according to the judgment result of the input stimuli;

When the number of counts of invalid incentives reaches an attack threshold, processing to deal with attacks is triggered.

Through the structure and design of the strong physical unclonable function device resistant to modeling attacks, the present invention can actively detect and passively defend against serious threats to the safety of the strong physical unclonable function device while ensuring the randomness and stability of the strong physical unclonable function device Modeling attacks to effectively resist modeling attacks.

Description of drawings

Fig. 1 is a system structure diagram of a strong physical non-clonable function device resistant to modeling attacks of the present invention;

Fig. 2 is a preferred embodiment of the strong physical unclonable function device of the present invention against modeling attacks;

Fig. 3 is a schematic diagram of the implementation method of the strong physical unclonable function device based on the anti-modeling attack of the present invention;

Fig. 4 is a schematic flow diagram of the implementation method of the strong physical unclonable function device based on the anti-modeling attack of the present invention;

Fig. 5A is one example of the Boolean confusion module in the strong physical unclonable function device resistant to modeling attacks of the present invention;

Fig. 5B is the second example of the Boolean confusion module in the strong physical unclonable function device resistant to modeling attacks of the present invention;

Figure 5C is the third example of the Boolean confusion module in the strong physical unclonable function device resistant to modeling attacks of the present invention;

Fig. 6 is a schematic diagram of an embodiment in which the incentive division module is implemented by a switch-type Boolean confusion module in the strong physical unclonable function device resistant to modeling attacks of the present invention;

7 is a schematic diagram of an embodiment of the response calculation module of the multiplexing Boolean confusion module of the strong physical unclonable function device resistant to modeling attacks in the present invention;

Fig. 8 is a schematic diagram of the effective number of incentives of an example of a strong physical non-clonable function device resistant to modeling attacks in the present invention;

Fig. 9 is a schematic diagram of randomness evaluation of an example of a strong physical unclonable function device resistant to modeling attacks in the present invention;

Fig. 10 is a schematic diagram of the stability evaluation of an example of a strong physical unclonable function device resistant to modeling attacks in the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

Fig. 1 shows that the present invention provides a strong physical unclonable function device 100 resistant to modeling attacks, including:

The Boolean confusion module 10 is used to output the response after the input stimulus is reprocessed by multiple weak physical unclonable function devices and Boolean logic elements, so as to realize the unpredictable Boolean logic relationship;

An incentive division module 20, configured to divide input incentives into valid incentives and invalid incentives;

An attack detection module 30, configured to detect the invalid stimulus and identify a modeling attack, and process the invalid stimulus and the modeling attack;

The response calculation module 40 is configured to perform response calculation on the effective stimulus through a strong physical unclonable function device.

In a preferred embodiment of the strong physical unclonable function device 100 resistant to modeling attacks of the present invention, as shown in Figure 2, the Boolean confusion module 10 includes:

The weak PUF sub-module 11 is used for the weak physical unclonable function device to process the input excitation to obtain Boolean logic configuration bits;

The Boolean determination sub-module 12 is used to input the response of the weak physical unclonable function device, and reprocess the response through the determined input-output Boolean logic relationship to obtain an output response; and/or

The manufacturer obtains the response of the weak physical unclonable function device through a one-time secure channel to obtain the actual Boolean logic of the Boolean obfuscation module 10 . The response of the weak PUF sub-module 11 is the Boolean logic configuration bit of the Boolean obfuscation module 10. For different chips, the Boolean logic configuration bits are also different, so that the Boolean logic relationship of the Boolean obfuscation module 10 is not the same. , the manufacturer uses a dedicated one-time security channel to obtain the response in the weak PUF; because the response of the weak PUF sub-module 11 is random, the response of the Boolean sub-module 12 is also random, and it is unpredictable before manufacturing, making the Boolean The obfuscation module 10 improves security by introducing a weak PUF that is naturally resistant to modeling attacks. The weak PUF, the original strong physical unclonable function device and corresponding devices constitute a strong physical unclonable function device 100 that is resistant to modeling attacks.

In this embodiment, the incentive division module 20 includes:

The division rule sub-module 21 is used to define the division rules of the valid incentives and invalid incentives;

The division execution sub-module 22 is used to compare the input stimulus with the division rule to obtain the category of the input stimulus;

The division rules include: the input stimulus is divided into a valid stimulus set and an invalid stimulus set; the valid stimulus set is the input set legally used by the strong physical unclonable function device in normal applications, and the invalid stimulus set is the strong physical A set of inputs that are illegally used by a non-clonable function device in normal applications.

Preferably, as shown in Figure 6, the excitation division module 20 is realized by the hardware circuit resources of the Boolean confusion module 10, and the division rule is determined according to the type of the Boolean confusion module 10;

When the excitation division module 20 is a switch type Boolean confusion module or a switch normally open type Boolean confusion module, the Boolean confusion module 10 is connected in series, and the division rules include:

Divide the input stimuli C ₁ to C _4m , the response of the weak PUF sub-module 11 and the input value together determine the effective value or HiZ of the output; if the input stimuli makes the jump starting from T ₀ can pass through the switch-type Boolean confusion modules S _B1 , S _B2 ,..., S _Bm and finally reach T ₁ , then the input stimulus is an effective stimulus; if the input stimulus makes the transition from T ₀ unable to pass through The switch-type Boolean confusion module S _B1 , S _B2 ,...,S _Bm finally reaches T ₁ , which is an invalid excitation;

When the Boolean obfuscation module 10 is a switch type Boolean logic obfuscation module, the switch of the Boolean obfuscation module 10 is controlled by the output control byte of the weak PUF sub-module 11 .

Further, for easy understanding, the excitation division module 20 divides the input excitations C ₁ ~C _4m by a switch-type Boolean confusion module, and the response of the weak PUF together with the values of A, B, C, and D determine the value of Y to be P , or HiZ, whether the output value is P, see Fig. 5A ~ Fig. 5C. At this time, the division rule is that if the input excitation makes the jump from T0 pass through the switch-type Boolean confusion modules S _B1 , S _B2 ,...,S _Bm and finally reach T ₁ , then the excitation is an effective excitation, otherwise for ineffective incentives.

Further, the attack detection module 30 includes:

The incentive counting sub-module 31 is used to count the invalid incentives according to the judgment result of the input incentives;

The attack processing sub-module 32 is configured to trigger processing to deal with an attack when the number of invalid incentive counts reaches an attack threshold.

Preferably, the following takes the example of the response calculation module 40 in FIG. 7 as an example for illustration. The response calculation module 40 shown in FIG. 7 reuses the Boolean obfuscation module 10 shown in FIGS. 5A-6 to save hardware overhead. As shown in Figure 7, the response calculation of the response calculation module 40 to the input excitation is produced by comparing the time delays of two path propagation jumps, and the sub-paths of each path are determined by the input excitation; the response calculation module 40 Processing the propagation process of the transition through the path can also combine and reuse the hardware circuit resources of the Boolean obfuscation module 10 . The response calculation module 40 calculates the response for the effective stimulus, and the response is generated by comparing the time delays of the propagation transitions of the two paths, and the sub-paths of each path are determined by the stimulus. In order to save hardware overhead, the response calculation module 40 and the Boolean obfuscation module 10 can be combined and multiplexed, as shown in FIG. 7 . When the effective stimulus is ready, a jump starts from T and arrives at two paths through multiple switch-type Boolean confusion modules S _U1 , S _U2 ,...S _Um and S _B1 , S _B2 ,...S _Bm For the arbiter, for different stimuli, the circuit paths that the jumps pass through in the switch-type Boolean confusion module are also different, so the time delays are different. Finally, the arbiter compares the order of the jumps to get the final response R.

Fig. 3 is a flow chart of the first embodiment of the implementation method of the present invention based on the strong physical unclonable function device resistant to modeling attacks, which can pass the strong physical unclonable function resistant to modeling attacks as shown in Figs. 1-2 The device 100 is implemented, including the following steps:

Step S301, the Boolean obfuscation step, outputs the response after reprocessing the input stimulus through a plurality of weak physical unclonable function devices and Boolean logic elements, so as to realize the unpredictable Boolean logic relationship;

Step S302, divide the incentive step, and divide the input incentive into valid incentive and invalid incentive;

Step S303, detecting an attack step, detecting the invalid incentive to identify a modeling attack, and processing the invalid incentive and the modeling attack;

Step S304, the step of calculating the response, calculating the response of the effective stimulus by means of a strong physical unclonable function device.

More preferably, in the step S301, the Boolean confusion step also includes:

The weak physical unclonable function device processes the input stimulus to obtain Boolean logic configuration bits;

Input the response of the weak physical unclonable function device, and reprocess the response through the determined input-output Boolean logic relationship to obtain an output response; and/or

The manufacturer obtains the response of the weak physical unclonable function device through a one-time secure channel to obtain the actual Boolean logic.

Fig. 4 is a flow chart of a specific embodiment of the implementation method of the strong physical unclonable function device based on modeling attack resistance in the present invention, which can be achieved through the physical unclonable function device resistant to modeling attack as shown in Figs. 1-2 100 implementation, in order to have better defense and active detection against modeling attacks, in the specific implementation process, it also includes:

Step S401, defining the rules for dividing the valid incentives and invalid incentives;

Step S402, comparing the input stimulus with the division rule to obtain the category of the input stimulus;

The division rules include: the input stimulus is divided into a valid stimulus set and an invalid stimulus set; the valid stimulus set is the input set legally used by the strong physical unclonable function device in normal applications, and the invalid stimulus set is the strong physical The set of inputs illegally used by the non-clonable function device in normal applications;

Step S403, counting the invalid incentives according to the judgment result of the input incentives;

Step S404, when the counted number of invalid incentives reaches the attack threshold, triggering processing to deal with the attack;

Through the attack detection module 30 built in the chip, it is used to actively detect whether the strong physical unclonable function device 100 resistant to modeling attacks is subjected to modeling attacks. Since modeling attacks are based on machine learning methods, and machine learning needs to obtain training sets, the attacker is bound to ask the anti-building Strong physical unclonable function device 100 for modulo attacks with invalid stimuli as inputs. The attack detection module 30 counts invalid stimuli, and when the number of invalid stimuli reaches the attack threshold, it can be considered that the strong physical unclonable function device 100 resistant to modeling attacks is subjected to a modeling attack, thereby executing a series of processing methods for the attack. The attack threshold and the corresponding processing method here are set by the manufacturer according to the security level of the actual application. For example, when the number of invalid incentives is small, the processing method of automatic power-off can be triggered; when the number of invalid incentives is large, for For application scenarios with a very high level of security, the processing method of chip self-destruction can be implemented.

Furthermore, the theoretical calculation of the existence probability of effective incentives is carried out. Let N _CB be the number of bits of the input excitation, then the total number of excitations of the strong physical unclonable function device 100 resistant to modeling attacks is Let N _SI be the input number of a single switch-type Boolean obfuscation module, and N _S be the number of switch-type Boolean obfuscation modules of the excitation division module 20, then N _S =N _CB /(N _SI -1). Since the switch-type Boolean obfuscation modules are connected in series, the jump from T ₀ to reach T ₁ must pass through each switch-type Boolean obfuscation module. For a switch-type Boolean obfuscation module, the total number of switches is Assuming that for a switch-type Boolean obfuscation module, the probability that the switch controlled by a bit of the weak PUF response is turned off is P _OFF , then the probability that at least one switch is turned on in a switch-type Boolean obfuscation module is Therefore, the probability that the jump of T ₀ can reach T ₁ through N _S switch-type Boolean confusion modules is Since the response randomness of a weak PUF can generally reach 50%, that is, the probability of turning off the switch for 0 and turning on the switch for 1 is similar, so P _OFF ≈50%. When N _SI =5 and N _CB =128, the probability of effective excitation is 99.95%, which shows that effective excitation can be provided for practical applications. If this probability needs to be further increased, a switch normally open Boolean obfuscation module can also be used to divide the input stimulus.

Further, a theoretical calculation is performed on the number of effective incentives. For a switch-type Boolean obfuscation module, according to the response of the weak PUF, how many switches are turned on, then the switch-type Boolean obfuscation module can propagate the P value to Y under how many input combinations. There are combinations of i switches that are turned on and N _SS -i switches that are turned off. , so the mathematical expectation of the number of input combinations that a switch-type Boolean obfuscation module can propagate P to Y is:

Among them, i is calculated from 1 to N _SS , which means that considering a single switch-type Boolean obfuscation module, there is 1 input combination that can transfer the P value to Y, there are 2 input combinations, ... until N _SS input combinations;

Denotes the probability that i switches are on and N _SS -i switches are off.

There are N _S switch-type Boolean confusion modules in the whole stimulus division module. If the j-th switch-type Boolean confusion module has i _j switches turned on, then the total number of effective incentives is M, and the mathematical expectation of the total number of effective incentives is:

Among them, a plurality of summation symbols means to be considered separately: all switch-type Boolean obfuscation modules have only one switch on, N _S -1 switch-type Boolean obfuscation modules have only one switch on and one switch-type Boolean obfuscation module There are 2 switch-on situations, N _S -1 switch-type Boolean obfuscation modules have only one switch on and 1 switch-type Boolean obfuscation module has 3 switch-on situations, ... until all switch-type Boolean obfuscation modules are all The switch is turned on. When N _SI =5 and N _CB =128, the mathematical expectation of the total number of effective incentives is 7.9×10 ²⁸ , which can last for 10 ¹⁴ years. Further, theoretically calculate the probability of the attacker obtaining effective incentives when collecting the training set. If the attacker chooses incentives at random, the mathematical expectation of choosing an effective incentive is:

When N _SI =5 and N _CB =128, this value is only 2.33×10 ^-8 %.

Further, a theoretical calculation is performed on the total number of incentives that an attacker needs to collect to collect a specific number of valid incentives. Assuming that the training set needs N _TS effective incentives, if the i-th incentive is randomly selected to make the training set contain N _TS effective incentives, it means that there are N _TS -1 effective incentives in the previous i-1 incentive selection, and The incentive for the i-th choice is valid. Therefore, the mathematical expectation of the total number of stimuli required to obtain N _TS effective stimuli is:

When N _SI =5 and N _CB =128, if N _TS =2×10 ⁶ , the value is about 8.6×10 ¹⁵ , which means that it takes more than 10 years to collect an effective training set.

Specifically, the integrated circuit simulation software (Simulation program with integrated circuit emphasis, SPICE) was used to simulate the strong PUF instance based on the SMIC (Semiconductor Manufacturing International Corporation, SMIC) 180nm process, and a total of 10 instances were generated. Firstly, the effective CRP of the strong PUF is evaluated, as shown in Figures 8 to 10, on average, the strong PUF has 1.74×10 ¹⁴ effective stimuli, which is enough for practical applications. According to these effective incentives, when the attacker randomly selects the training set, there is only a 0.0009% probability of being able to select effective incentives. Then the randomness of the strong PUF is evaluated, as shown in FIG. 9 , on average, the randomness of the strong PUF is 49.99%, which is very close to the ideal value of 50%. Finally, the stability of the strong PUF was evaluated, as shown in FIG. 10 , on average, the stability of the strong PUF was 95.40%.

Compared with the existing strong PUF in terms of performance and security, the randomness and stability of the present invention are very close to the existing strong PUF, but the security against modeling attacks is obviously superior to the existing ones in the following points: Existing technologies: (1) the present invention includes weak PUFs in strong PUFs to passively resist CRP modeling; (2) the attacker has only a 0.0009% possibility of randomly selecting a valid CRP; (3) the present invention can actively detect and build mod attack.

In summary, the present invention uses a weak physical unclonable function device to construct a Boolean confusion module, and the Boolean logic relationship between the input and output of the Boolean confusion module is unpredictable before manufacture; the Boolean confusion module is used to construct an incentive division module, so The excitation division module divides the input excitation set of the strong physical unclonable function device into an effective excitation set and an invalid excitation set; the attack detection module constructed by using the Boolean confusion module, the attack detection module can detect whether the input excitation is valid; use the Boolean confusion The module builds a response computation module that computes an output response for an input stimulus. Thereby, while ensuring the randomness and stability of the strong physical unclonable function device, the present invention can actively detect and passively defend against modeling attacks that seriously threaten the safety of the strong physical unclonable function device, thereby effectively resisting the modeling attack.

Certainly, the present invention also can have other multiple embodiments, without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and deformations according to the present invention, but these corresponding Changes and deformations should belong to the scope of protection of the appended claims of the present invention.

Claims (10)

1. a kind of strong physics unclonable function device of anti-modeling attack characterized by comprising

Boolean obscures module, for locating input stimulus again through multiple weak physics unclonable function devices and boolean logic element Output response after reason realizes that Boolean logic relationship is unpredictable；

Division module is motivated, for the input stimulus to be divided into effectively excitation and invalid excitation；Wherein, if the input stimulus Module is obscured through boolean and exports the response, then the input stimulus is effectively excitation, otherwise is invalid excitation；

Attack detection module handles the nothing for identifying modeling attack by detecting the number that the invalid excitation occurs Effect excitation and modeling attack；

Response computation module, for carrying out response computation to effective excitation by strong physics unclonable function device.

2. strong physics unclonable function device according to claim 1, which is characterized in that the boolean obscures module packet It includes:

Weak PUF submodule handles to obtain Boolean logic for the weak physics unclonable function device to the input stimulus Configuration bit；

Boolean determines submodule, leads to for the Boolean logic relationship by determining input and output and/or by disposable safe Road is reprocessed to obtain the response by the Boolean logic configuration bit.

3. strong physics unclonable function device according to claim 2, which is characterized in that the excitation division module packet It includes:

Division rule submodule, the division rule for defining effective excitation with motivating in vain；

It divides implementation sub-module and obtains the input stimulus for comparing the input stimulus and the division rule Classification；

The division rule includes: that the input stimulus is divided into effectively excitation set and gathers with invalid excitation；Effectively excitation collection It is combined into the strong physics unclonable function device legal input set used in normal use, invalid excitation collection is combined into institute State the input set that strong physics unclonable function device illegally uses in normal use.

4. strong physics unclonable function device according to claim 3, which is characterized in that the excitation division module is by institute The hardware circuit resource realization that boolean obscures module is stated, the division rule is determined according to the type that the boolean obscures module；

When the excitation division module be switching mode boolean obscure module or switch open type boolean obscure module when, the boolean Obscure module to connect in the form of concatenated；

When it is that switching mode Boolean logic obscures module that the boolean, which obscures module, pass through the output control of the weak PUF submodule Byte processed controls the switch that the boolean obscures module.

5. strong physics unclonable function device according to claim 4, which is characterized in that the attack detection module packet It includes:

Counting submodule is motivated, for when the input stimulus is invalid excitation, the number motivated in vain to generation to be counted Number；

Attack processing submodule, for being identified as the modeling and attacking and trigger when the count number reaches attack threshold value Cope with the processing of the modeling attack.

6. strong physics unclonable function device according to claim 1, which is characterized in that the response computation module is to institute The response computation of input stimulus is stated to generate by comparing the time delay that two paths propagate jump, the subpath in every path by The input stimulus determines；

Communication process of the jump through the path described in the response computation resume module can also merge and be multiplexed the boolean Obscure the hardware circuit resource of module.

7. a kind of implementation method of the strong physics unclonable function device based on anti-modeling attack characterized by comprising

Boolean obscures step, by input stimulus after multiple weak physics unclonable function devices and boolean logic element reprocessing Output response realizes that Boolean logic relationship is unpredictable；

Partiting step is motivated, input stimulus is divided into effectively excitation and invalid excitation；Wherein, if the input stimulus is mixed through boolean Module of confusing exports the response, then the input stimulus is effectively excitation, otherwise is invalid excitation；

Attack detecting step, the number by detecting the invalid excitation generation identify modeling attack, handle described invalid sharp It encourages and is attacked with the modeling；

Response computation step carries out response computation to effective excitation by strong physics unclonable function device.

8. implementation method according to claim 7, which is characterized in that the boolean obscures step further include:

The weak physics unclonable function device handles the input stimulus to obtain Boolean logic configuration bit；

By the Boolean logic relationship and/or disposable safe channel of determining input and output to the Boolean logic configuration bit Reprocessing obtains the response.

9. implementation method according to claim 7, which is characterized in that the excitation partiting step includes:

The division rule for defining effective excitation and motivating in vain；

The input stimulus and the division rule are compared, obtain the classification of the input stimulus；

The division rule includes: that the input stimulus is divided into effectively excitation set and gathers with invalid excitation；Effectively excitation collection It is combined into the strong physics unclonable function device legal input set used in normal use, invalid excitation collection is combined into institute State the input set that strong physics unclonable function device illegally uses in normal use；

The attack detecting step includes:

When the input stimulus is invalid excitation, the number motivated in vain to generation is counted；

When the count number reaches attack threshold value, it is identified as the modeling and attacks and trigger the place for coping with the modeling attack Reason.

10. implementation method according to claim 8, which is characterized in that the response computation step further include:

The response computation of the input stimulus is generated by comparing the time delay that two paths propagate jump, the son in every path Path determines by the input stimulus, and handles the communication process of the jump through the path and can also be multiplexed the boolean Obscure the hardware circuit resource of step.