Abstract Convolutional neural network accelerators are increasingly used in safety–critical applications, including autonomous vehicles. Therefore, they are particularly vulnerable to hardware Trojan insertion, a security attack that takes place during the development of integrated circuits. This work presents for the first time a large-scale study of the impact of hardware Trojan insertion on convolutional neural network accelerators, focusing on those that use approximate commuting techniques, prevalent in embedded applications. We investigate three types of such networks, MobileNet V2, ShuffleNet V2, and GhostNet, trained in datasets of grayscale speed limit sign images and GTSRB. Our results show that certain parts of these architectures are more susceptible to hardware Trojan attacks, specifically a specific set of processing elements, referred to as “important” in the classification, ReLU6, and Max pooling layers, respectively. These findings are subsequently used to develop two countermeasures; the first relies on selective hardware redundancy (SHR), and the second uses a combination of hardware and time redundancy (SHTR). The proposed defenses are experimentally validated. Our results show that the SHR provides speedy recovery from an attack while incurring between 6 and 10% area overheads, whereas SHTR requires more time to detect the Trojan, and its area overhead is much smaller (~ 0.3%).

Hardware obfuscation is a well-known countermeasure against reverse engineering. For FPGA designs, obfuscation can be implemented with a small overhead by using underutilised logic cells; however, its effectiveness depends on the stealthiness of the added redundancy. In this paper, we show that it is possible to deobfuscate an SRAM FPGA design by ensuring the full controllability of each instantiated look-up table input via iterative bitstream modification. The presented algorithm works directly on bitstream and does not require the possession of a flattened netlist. The feasibility of our approach is verified on the example of an obfuscated SNOW 3G design implemented on a Xilinx 7-series FPGA.

We explore mobile device touchscreen characteristics to build a user-device (UD) biometric physical unclonable function (PUF). Human user touchscreen interaction induces dynamic capacitive differences. Sensors detect current differences which are a function of both (1) a human biometric of how a shape is traced and (2) silicon foundry process transistor-level variability embedded in the touchscreen grid. This forms a physical function with input x defining a shape and output y abstracted from the measured current value stream. We argue and establish that this physical function has PUF attributes. Moreover, it provides a robust user-device biometric-based authentication mechanism. Authentication is based on geometric shapes (challenges) drawn on the touchscreen. Users trace them. The authentication layer creates a response abstract, and validates it against a user profile. Authentication accuracy is affected by the complexity of geometric shapes as well as the validation algorithm. We consider polyline shapes (simple gestures) and complex closed geometric shapes (complex gestures). Complex gestures offer higher response entropy, but are computationally less efficient with a slightly lower validation accuracy. Complex gestures achieve 99.6% accuracy compared with 100% for simple gestures. User profiles exhibit physical unclonable function (PUF) properties. Touchscreen gestures are quantized into binary strings. Gesture hamming distance is 60+ bits for 128-bit strings for different user-device profiles; it is 0 bits for the same profile. This demonstrates variability and reproducibility respectively. Montreal TestU01 tests binary string pseudorandom characteristics; the majority of tests pass showing pseudorandom number generator (PRG) characteristics.

We propose a novel malware detection system for critical embedded and cyber-physical systems (CPS). The system exploits electromagnetic (EM) side-channel signals from the device to detect malicious activity. During training, the system models EM emanations from an uncompromised device using a neural network. These EM patterns act as fingerprints for the normal program activity. Next, we continuously monitor the target device’s EM emanations. Any deviation in the device’s activity causes a variation in the EM fingerprint, which in turn violates the trained model, and is reported as an anomalous activity. The system can monitor the target device remotely (without any physical contact), and does not require any modification to the monitored system. We evaluate the system with different malware behavior (DDoS, ransomware, and code modification) on different applications using an Altera Nios-II soft-processor. Experimental evaluation reveals that our framework can detect DDoS and ransomware with 100% accuracy (AUC = 1.0), and stealthier code modification (which is roughly a 5 μ s long attack) with an AUC ≈ 0.99, from distances up to 3 m. In addition, we execute control-flow hijack, DDoS, and ransomware on different applications using an A13-OLinuXino—a Cortex A8 ARM processor single board computer with Debian Linux OS. Furthermore, we evaluate the practicality and the robustness of our system on a medical CPS, implemented using two different devices (TS-7250 and A13-OLinuXino), while executing control-flow hijack attack. Our evaluations show that our framework can detect these attacks with perfect accuracy.

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.