DLA-PUF: deep learning attacks on hardware security primitives

Physically Unclonable Function (PUF) circuits are finding wide-spread use due to increasing adoption of IoT devices. However, the existing strong PUFs such as Arbiter PUFs (APUF) and its compositions are susceptible to machine learning (ML) attacks because the challenge-response pairs have a linear relationship. In this paper, we present a Recurrent-Neural-Network PUF (RNN-PUF) which uses a combination of feedback and XOR function to significantly improve resistance to ML attack, without significant reduction in the reliability. ML attack is also partly reduced by using a shared comparator with offset-cancellation to remove bias and save power. From simulation results, we obtain ML attack accuracy of 62% for different ML algorithms, while reliability stays above 93%. This represents a 33.5% improvement in our Figure-of-Merit. Power consumption is estimated to be 12.3μW with energy/bit of ≈ 0.16pJ.

We propose a scheme to prevent the machine learning (ML) attacks against physically unclonable functions (PUFs). A silicon PUF is a security primitive in a semiconductor chip that generates a unique identifier by exploiting device variations. However, some PUF implementations are vulnerable to ML attacks, in which an attacker tries to obtain the mathematical clone of the target PUF to predict its responses. Our scheme adds intentional noise to the responses to disturb ML by an attacker so that the clone fails to be authenticated, while the original PUF can still be correctly authenticated using an error correction code (ECC). The effectiveness of this scheme is not very obvious because the attacker can also use the ECC. We apply the countermeasure to n-XOR arbiter PUFs to investigate the feasibility of the proposed scheme. We explain the relationship between the prediction accuracy of the clone and the number of intentional noise bits. Our scheme can successfully distinguish a clone from the legitimate PUF in the case of 5-XOR PUF.

Communication security is essential for the proper functioning of the Internet of Things. Traditional approaches that rely on cryptographic keys are vulnerable to side-channel attacks. Physical Unclonable Functions (PUFs), leveraging unavoidable and irreproducible variations of integrated circuits to produce responses unique for individual PUF devices, are emerging as promising candidates as security primitives to provide keyless solutions. Before a PUF can be adopted for real applications, the PUF must be thoroughly examined to understand its various properties for its application feasibility. In this paper, we study XOR PUFs for broad ranges of values for circuit architecture parameters. XOR PUFs have been extensively studied, and have been shown to be unable to withstand machine learning attacks for 64-bit XOR PUFs with less than ten component PUFs. Attack methods employed in existing studies need a large number of challenge-response pairs (CRPs), which are obtainable only if the PUF has an open access interface. When PUF-embedded devices equipped with mutual authentication or response obfuscating techniques, it is difficult for attackers to accumulate large numbers of CRPs. With only a small number of accumulated CRPs available to attackers, small size PUFs, like XOR PUFs with a small number of component PUFs and stages, may become resistant to machine learning attacks. Since smaller sizes mean less resource-demanding, it is worthwhile to examine such PUFs which have usually been considered unsafe against attacks. Such are thoughts that have been motivating us in this paper to explore the PUF performances for a wide range of values of the PUF architecture parameters.

Silicon physical unclonable function (PUF) has emerged as a promising spoof-proof solution for low-cost device authentication. Due to practical constraints in preventing phishing through a public network or insecure communication channels, simple PUF-based authentication protocol with unrestricted queries and transparent responses is vulnerable to modeling and replay attacks. In this article, we present a modeling attack resistant PUF-based mutual authentication scheme to mitigate the practical limitations in applications where a resource-rich server authenticates a device with no strong restriction imposed on the type of PUF design or any additional protection on the binary channel used for the authentication. Our scheme uses an active deception protocol to prevent machine learning (ML) attacks on a device with a monolithic integration of a genuine strong PUF (SPUF), a fake PUF, a pseudorandom number generator (PRNG), a register, a binary counter, a comparator, and a simple controller. The hardware encapsulation makes the collection of challenge-response pairs (CRPs) easy for model building during enrollment but prohibitively time consuming upon device deployment through the same interface. A genuine server can perform a mutual authentication with the device using a combined fresh challenge contributed by both the server and the device. The message exchanged in clear cannot be manipulated by the adversary to derive unused authentic CRPs. The adversary will have to either wait for an impractically long time to collect enough real CRPs by directly querying the device or the ML model derived from the collected CRPs will be poisoned. The false PUF multiplexing is fortified against the prediction of waiting time by doubling the time penalty for every unsuccessful guess. Our implementation results on field-programmable gate array (FPGA) device and security analysis have corroborated the low hardware overheads and attack resistance of the proposed deception protocol.

Electronic payment systems have leveraged the advantages offered by the RFID technology, whose security is promised to be improved by applying the notion of Physically Unclonable Functions (PUFs). Along with the evolution of PUFs, numerous successful attacks against PUFs have been proposed in the literature. Among these are machine learning (ML) attacks, ranging from heuristic approaches to provable algorithms, that have attracted great attention. Our paper pursues this line of research by introducing a Fourier analysis based attack against PUFs. More specifically, this paper focuses on two main aspects of ML attacks, namely being provable and noise tolerant. In this regard, we prove that our attack is naturally integrated into a provable Probably Approximately Correct (PAC) model. Moreover, we show that our attacks against known PUF families are effective and applicable even in the presence of noise. Our proof relies heavily on the intrinsic properties of these PUF families, namely arbiter, Ring Oscillator (RO), and Bistable Ring (BR) PUF families. We believe that our new style of ML algorithms, which take advantage of the Fourier analysis principle, can offer better measures of PUF security.

Physically unclonable functions (PUFs) have been widely proposed as hardware primitives for device identification and cryptographic key generation. While strong PUFs provide an exponential number of challenge–response pairs (CRPs) for device authentication, most are vulnerable to machine-learning (ML)-based modeling attacks. Regarding the inherent hard learning problem, adversarial-based PUFs pseudo-randomly switch between several predefined internal states to poison the ML algorithms. However, existing adversarial-based PUFs sacrifice reliability or security against ML attacks. Furthermore, most works only evaluated the ML resistance by empirical experiments, which provides limited evidence for ML resistance. This work explored a novel adversarial-based PUF, Configurable Poisoning PUF (CP PUF), that simultaneously maintains reliability and protects the internal state by a reliability trapdoor. We then connect the proposed CP PUF with the learning parity with noise (LPN) problem to provide a theoretical security guarantee. Experiments show that the reliability of our proposed CP PUF keeps at least 88.9% and outperforms the existing adversarial-based PUFs under different levels of noisy environments. Moreover, the proposed CP PUF achieved a 50.16% accuracy under empirical ML attacks, a near-optimal performance.

Physically unclonable functions (PUFs) have been explored as lightweight hardware primitives for the purpose of realizing robust security via strong authentication or secure key/ID generation. PUF harness manufacturing process variations for the purpose of generating binary keys or binary functions. An ideal strong PUF is a binary function that maps an m-bit input challenge to an uniquen-bit output response, making it attractive for authentication applications. Unfortunately, real strong PUF implementations suffer from reliability issues where the same challenge may produce different responses in the presence of noise. To overcome this problem, strong PUF leverages the availability of exponential number of challenge-response pairs (CRPs). A successful authentication event requires acquiring multiple CRPs and applying a threshold. In contrast, weak PUFs produce limited keys and are required to be highly reliable. Multiple techniques have been developed to achieve the necessary reliability. An additional prerequisite for strong PUFs is resilience against model-building attacks (cloning) by an adversary, who has observed a few CRPs, to prevent successful prediction of future CRPs. In this work, we first illustrate a strong PUF design that re-purposes a weightless neural network (WNN). Second, we showcase the robustness of WNN-based strong PUFs with respect to machine learning attacks, while providing desirable uniqueness and reliability metrics. Finally, we employ an initial entropy source of highly reliable weak PUF bits mapped to weightless neural networks (WNNs) for the purpose of creating a near-ideal strong PUF in terms of reliability. Our results show that it is possible to create highly reliable WNN–based strong PUFs with < 65 % ML accuracy by using as few as 32 initial reliable weak PUF bits.

The Internet of things (IoT) ecosystem has grown exponentially with the convergence of various technologies such as deep learning, sensor systems, and advances in computing platforms. With such a highly pervasive nature of “smart” devices, the nature of data being collected and processed can be increasingly private and require safeguards to ensure the data’s integrity and security. Physically unclonable functions (PUFs) have emerged as a lightweight, viable security protocol in the Internet of Things (IoT) framework. Malicious modeling of PUF architectures has proven to be difficult due to the inherently stochastic nature of PUF architectures. In this work, we show that knowledge of the underlying PUF structure is unnecessary to clone a PUF. We tackle the problem of cloning PUF-based edge nodes in different settings such as unencrypted, encrypted, and obfuscated challenges in an IoT framework. We present a novel non-invasive, architecture-independent, machine learning attack for robust PUF designs and can handle encryption and obfuscation-based security measures on the transmitted challenge response pairs (CRPs). We show that the proposed framework can successfully clone different PUF architectures, including those encrypted using two (2) different encryption protocols in DES and AES and with varying degrees of obfuscation. We also show that the proposed approach outperforms a two-stage brute force attack model. Finally, we offer a machine learning-based countermeasure, a discriminator, which can distinguish cloned PUF devices and authentic PUFs with an average accuracy of 96%. The proposed discriminator can be used for rapidly authenticating millions of IoT nodes remotely from the cloud server.

Previous works have shown excellent prospects for implementing strong physical unclonable functions (PUFs) with memristive crossbar circuits. Here we first propose two techniques for boosting the robustness of such PUFs to machine learning (ML) attacks. The general idea behind both proposals is to maximize the contribution of each crosspoint device to the PUF output to make the response less predictable. Specifically, we present results for choosing an optimal ratio of selected rows and columns, and investigate in detail the improvements in robustness due to the balancing of device conductances in the crossbar array. The effectiveness of the proposed algorithm for conductance balancing is confirmed by modeling the response of a two-sided PUF based on a 20 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\times20$ </tex-math></inline-formula> crossbar memristive circuit with a multilayer perceptron network. Second, we explore some open questions which require in-depth analysis. Specifically, we quantify the effect of device nonlinearity and device analog-tunability. We show that nonlinear, analog memristive PUFs outperform the PUFs that have either linear or digital devices. Finally, we explore the effect of stuck-at-fault devices (nonideal yield) on PUFs uniformity. Indeed, by modeling this hardware imperfection, we show that the proposed algorithm results in a more-robust PUF.

Physical Unclonable Functions (PUFs) are promising security primitives for resource-constrained network nodes. The XOR Arbiter PUF (XOR PUF or XPUF) is an intensively studied PUF invented to improve the security of the Arbiter PUF, probably the most lightweight delay-based PUF. Recently, highly powerful machine learning attack methods were discovered and were able to easily break large-sized XPUFs, which were highly secure against earlier machine learning attack methods. Component-differentially-challenged XPUFs (CDC-XPUFs) are XPUFs with different component PUFs receiving different challenges. Studies showed they were much more secure against machine learning attacks than the conventional XPUFs, whose component PUFs receive the same challenge. But these studies were all based on earlier machine learning attack methods, and hence it is not clear if CDC-XPUFs can remain secure under the recently discovered powerful attack methods. In this paper, the two current most powerful two machine learning methods for attacking XPUFs are adapted by fine-tuning the parameters of the two methods for CDCXPUFs. Attack experiments using both simulated PUF data and silicon data generated from PUFs implemented on field-programmable gate array (FPGA) were carried out, and the experimental results showed that some previously secure CDC-XPUFs of certain circuit parameter values are no longer secure under the adapted new attack methods, while many more CDC-XPUFs of other circuit parameter values remain secure. Thus, our experimental attack study has re-defined the boundary between the secure region and the insecure region of the PUF circuit parameter space, providing PUF manufacturers and IoT security application developers with valuable information in choosing PUFs with secure parameter values.

Physically unclonable functions (PUFs) are emerging as hardware primitives for key-generation and light-weight authentication. Strong PUFs represent a variant of PUFs which respond to a user challenge with a response determined by its unique manufacturing process variations. Unfortunately many of the Strong PUFs have been shown to be vulnerable to model building attacks when an attacker has access to challenge and response pairs. In mounting a model building attack, typically machine learning is used to build a software model to forge the PUF. Researchers have long been interested in designing Strong PUFs that are resistant to model building attacks. However, with innovations in application of machine learning, nearly all Strong PUFs presented in the literature have been broken. In this paper, first we present results from a set of experiments designed to show that if certain randomness properties can be met, cascaded structure based Strong PUFs can indeed be made machine learning (ML) attack resistant against known ML attacks. Next we conduct machine learning experiments on an abstract PUF model using Support Vector Machines, Logistic Regression, Bagging, Boosting and Evolutionary techniques to establish criteria for machine learning resistant Strong PUF design. This paper does not suggest how to harvest the process variation, which remains within the purview of a circuit designer; rather it suggests what properties of the building blocks to aim for towards building a machine learning resistant Strong PUF - thus paving the path for a systematic design approach.

In this paper, a physical unclonable function (PUF), a type of hardware security device, is proposed to overcome the limitations of existing security schemes. A $32\times 32$ crossbar array using TiOx/Al2O3- based memristors was fabricated, and electrical characteristics including its set voltage distribution were analyzed. The memristor switching characteristics model is described in a simplified space-charge-limited current (SCLC) regime. Based on this I-V model, selected bit-line current PUFs (SBC-PUFs) were designed with $32\times 32$ , $64\times 64$ , and $128\times 128$ crossbar arrays. The entropy source of these PUFs is the set voltage deviation in the fabricated memristors. Due to these characteristics, the SBC-PUF can exploit the broad resistance distribution near the switching region, including the internal resistance distributions of the high resistance state (HRS) and low resistance state (LRS). The SBC-PUF performance was evaluated for randomness/uniformity, correctness/reliability, and uniqueness by calculating the Hamming weight and intra/inter Hamming distance of challenge-response pairs (CRPs). The designed structure demonstrates high-security performance due to the high value of these indicators and the large number of CRPs. Furthermore, the devised PUF has a higher prediction error rate than arbiter PUF in machine learning attacks. This study verified that the SBC-PUF using the memristor of the crossbar array structure is safe enough to be used for hardware security.

Physical unclonable functions (PUFs) are used to extract unique signatures from silicon-based chips which can be used for chip authentication and producing unclonable cryptographic keys. However, researchers have found that PUFs are vulnerable to various machine learning modeling attacks. In this work, we introduce a unique hybrid PUF structure that uses Challenge-Response Pairs (CRPs) from an Arbiter PUF and feeds them to an XOR-Inverter based Ring Oscillator to generate responses which makes the PUF less vulnerable to machine learning modeling attacks. From the results, it is found that the prediction accuracy when different machine learning classifier algorithms are employed to attack the PUF, is drastically reduced and lies in the range of 3.5% to 6.8%, whereas the ANN-based model accuracy obtained is in the range of 5.4% to 7.5%. Our study indicates that the new design’s vulnerability in terms of prediction accuracy against different machine learning modeling attacks is less by 51.6% for ML and 54.1% for ANN compared to other delay-based PUF designs.

Strong physically unclonable function (PUF) offers a promising solution to low-cost hardware identification and authentication for Internet of Things (IoT) applications. The continuous advancement of machine learning (ML) technology makes the PUF resilience to ML attacks a major design priority. This paper presents a robust and area-efficient strong PUF design with high ML attack resilience. The proposed PUF architecture based on inverter amplifiers operating in the subthreshold region achieves both low energy consumption and high supply and temperature scalability. The proposed nonlinearity topology effectively enhances PUF resilience to various ML attacks with low implementation area and cost. The proposed strong PUF design was designed and implemented using a 28-nm CMOS process. The measurement results show that the proposed PUF design achieves a nearly ideal ML attack resilience of 49.96 &#x0025; with a small area of 239,857 F², and demonstrates a stable operation across a wide range of supply voltage from 0.5&#x2013;1.4 V and temperature from &#x2212;40&#x2013;100 &#x00B0;C. This represents <inline-formula> <tex-math notation="LaTeX">$3\times $ </tex-math></inline-formula> improvement in area efficiency, <inline-formula> <tex-math notation="LaTeX">$2.25\times $ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$1.08\times $ </tex-math></inline-formula> improvement in operating voltage and temperature range, respectively, compared to the state-of-the-art results.

The physical unclonable functions (PUFs) have been attracted attention to prevent semiconductor counterfeits. However, the risk of machine learning attack for an arbiter PUF, which is one of the typical PUFs, has been reported. Therefore, an XOR arbiter PUF, which has a resistance against the machine learning attack, was proposed. However, in recent years, a new machine learning attack using power consumption during the operation of the PUF circuit was reported. Also, it is important that the detailed tamper resistance verification of the PUFs to consider the security of the PUFs in the future. Therefore, this study proposes a new machine learning attack using electromagnetic waveforms for the XOR arbiter PUF. Experiments by an actual device evaluate the validity of the proposed method and the security of the XOR arbiter PUF.