A Computationally Efficient Tensor Regression Network-Based Modeling Attack on XOR Arbiter PUF and Its Variants

Santikellur, Pranesh; Chakraborty, Rajat Subhra

doi:10.1109/tcad.2020.3032624

Cited by 23 publications

(13 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Then simple machine learning is utilized to resolve the model parameters. This process is similar with other PUF attack studies [31], [32].…”

Section: Mathematical Model Attacksupporting

confidence: 90%

“…For 6-XOR-nCk-PUF, the number neurons of hidden layer of their setting is (2 4 , 2 5 , 2 4 ), the neuron sum is slightly larger than that of our method, and the attack results of those two method are similar as shown in Table . 6. For 7-XOR-nCk-PUF, the hidden layers' neurons number of their setting is (2 6 , 2 7 , 2 6 ), which is larger that of our setting (64, 32,16). The prediction accuracy of our method is slightly lower than that of the method displayed in [35], however, our setting has a slight advantage in training time, as shown in Table . 6, the prediction accuracy and train time of our method for 7-XOR-100(10)-PUF are 98.64% and 2045.97 seconds compared with that of the other method's 99.13% and 2460.64 seconds.…”

Section: Attack and Mlp Attack Comparisonmentioning

confidence: 71%

“…Yu et al [19] optimized combined power and modeling attacks to crack lightweight strong PUFs: voltage regulator (VR) PUFs. Santikellur et al [20] employded a novel machine learning-based modeling technique called efficient CANDECOMP/PARAFAC-tensor regression network (CP-TRN), a variant of CP-decomposition-based tensor regression network, to reduce the computational resource requirement of model building attacks on XOR APUF.…”

Section: Introductionmentioning

confidence: 99%

“…In addition, note that one nCk-PUF construction [29] leverages an emerging nano element (i.e., memristor), which is not readily accessible to common labs. Therefore, it is preferable to simulate the nCk-PUF structure to ease its security examination, especially considering the fact that the usage of simulated CRP is a commonly used strong PUF modeling resilience evaluation strategy [11], [13], [14], [16], [20], [30]. For the convenience of the readers and the research community, we have released the source code at https://github.com/users/sunshine-plh/projects/1.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Security Evaluation of n-Choose-k-Sum PUFs Against Modeling Attacks

et al. 2021

View full text Add to dashboard Cite

A physical unclonable function (PUF) generates hardware intrinsic volatile secrets by exploiting uncontrollable manufacturing randomness. Several PUF candidates use challenge-response pairs (CRPs) to enhance their security. A practically plausible idea for PUF is to use the technique called n-Choose-k-Sum (nCk) structure/topology, exemplified by the memristive device based on strong PUF (mrSPUF). The previous analysis claimed that nCk-PUF is resilient to modeling attacks. This paper provides numerical security evaluations on simulated CRPs -we first construct a mathematical model of nCk-PUF, and then analyze the possibility of modeling attacks using machine learning (ML). Our analysis demonstrates that our modeling attacks could break the unpredictability of nCk-PUF even though the nonlinearity of nCk-PUF could be increased due to XOR-operations on its circuit. Our analysis demonstrates that our modeling attacks could break the unpredictability of nCk-PUF even though the non-linearity of nCk-PUF could be increased due to XOR-operations on its circuit. Consequently, our analysis suggest that n-Choose-k-Sum PUFs (including mrSPUF) sharing similar topology could be insecure. It should be careful when utilizing nCk-PUF for strong PUF applications, and it is suggested to add some security blocks to protect the challenge-response interface, such as the Lockdown-PUF with additional random number generator (RNG), TREVERSE with additional Hash.

show abstract

“…Then simple machine learning is utilized to resolve the model parameters. This process is similar with other PUF attack studies [31], [32].…”

Section: Mathematical Model Attacksupporting

confidence: 90%

Section: Attack and Mlp Attack Comparisonmentioning

confidence: 71%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Security Evaluation of n-Choose-k-Sum PUFs Against Modeling Attacks

et al. 2021

View full text Add to dashboard Cite

show abstract

“…R is the modeling method used in [12], T is the modeling method used in [14], M* is the modeling method first proposed in [19] and then revisited in [17]. S also is the modeling method used in [15]. We used the simulator to generate 10 instances of 128-stage {2, 3, 4, 5, 6, 7, 8, 9 and 10}-XOR Arbiter PUF variants.…”

Section: A Experimental Setupmentioning

confidence: 99%

Elaborating on Sub-Space Modeling as an Enrollment Solution for Strong PUF

Pour

Hély

Beroulle

et al. 2022

2022 18th International Conference on Distributed Computing in Sensor Systems (DCOSS)

View full text Add to dashboard Cite

A novel hardware authentication primitive against modeling attacks

Liu

Cheng

Sun

et al. 2023

Circuit Theory & Apps

View full text Add to dashboard Cite

Traditional hardware security primitives such as physical unclonable functions (PUFs) are quite vulnerable to machine learning (ML) attacks. The primary reason is that PUFs rely on process mismatches between two identically designed circuit blocks to generate deterministic math functions as the secret information sources. Unfortunately, ML algorithms are pretty efficient in modeling deterministic math functions. In order to resist against ML attacks, in this letter, a novel hardware security primitive named neural network (NN) chain is proposed by utilizing noise data to generate chaotic NNs for achieving authentication. In a NN chain, two independent batches of noise data are utilized as the input and output training data of NNs, respectively, to maximize the uncertainty within the NN chain. In contrast to a regular PUF, the proposed NN chain is capable of achieving over 20 times ML attackresistance and 100% reliability with less than 39% power and area overhead.deterministic math functions, machine learning (ML) attacks, neural network (NN) chain, physical unclonable functions (PUFs)

show abstract

A Computationally Efficient Tensor Regression Network-Based Modeling Attack on XOR Arbiter PUF and Its Variants

Cited by 23 publications

References 23 publications

Security Evaluation of n-Choose-k-Sum PUFs Against Modeling Attacks

Security Evaluation of n-Choose-k-Sum PUFs Against Modeling Attacks

Elaborating on Sub-Space Modeling as an Enrollment Solution for Strong PUF

A novel hardware authentication primitive against modeling attacks

Contact Info

Product

Resources

About