Search for: All records

Spoofing a passive Hall sensor with fake magnetic fields can inject false data into the downstream of connected systems. Several works have tried to provide a defense against the intentional spoofing to different sensors over the last six years. However, they either only work on active sensors or against externally injected unwanted weak signals (e.g., EMIs, acoustics, ultrasound, etc.), which can only spoof sensor output in its linear region. However, they do not work against a strong magnetic spoofing attack that can drive the passive Hall sensor output in its saturation region. We name this as the saturation attack. In the saturation region, the output gets flattened, and no information can be retrieved, resulting in a denial-of-service attack on the sensor.Our work begins to fill this gap by providing a defense named PreMSat against the saturation attack on passive Hall sensors. The core idea behind PreMSat is that it cangenerate an internal magnetic field having the same strength but in opposite polarity to external magnetic fields injected by an attacker. Therefore, the generated internal magnetic field by PreMSat can nullify the injected external field while preventing: (i) intentional spoofing in the sensor’s linear region, and (ii) saturation attack in the saturation region. PreMSat integrates a low-resistance magnetic path to collect the injected external magnetic fields and utilizes a finely tuned PID controller to nullify the external fields in real-time. PreMSat can prevent the magnetic saturation attack having a strength up to ∼4200 A-t within a frequency range of 0 Hz–30 kHz with low cost (∼$14), whereas the existing works cannot prevent saturation attacks with any strength. Moreover, it works against saturation attacks originating from any type, such as constant, sinusoidal, and pulsating magnetic fields. We did over 300 experiments on ten different industry-used Hall sensors from four different manufacturers to prove the efficacy of PreMSat and found that the correlation coefficient between the signals before the attack and after the attack is greater than 0.94 in every test case. Moreover, we create a prototype of PreMSat and evaluate its performance in a practical system — a grid-tied solar inverter. We find that PreMSat can satisfactorily prevent the saturation attack on passive Hall sensors in real-time. 

Myocardial Infarction (MI), also known as heart attack, is a life-threatening form of heart disease that is a leading cause of death worldwide. Its recurrent and silent nature emphasizes the need for continuous monitoring through wearable devices. The wearable device solutions should provide adequate performance while being resource-constrained in terms of power and memory. This paper proposes an MI detection methodology using a Convolutional Neural Network (CNN) that outperforms the state-of-the-art works on wearable devices for two datasets - PTB and PTB-XL, while being energy and memory-efficient. Moreover, we also propose a novel Template Matching based Early Exit (TMEX) CNN architecture that further increases the energy efficiency compared to baseline architecture while maintaining similar performance. Our baseline and TMEX architecture achieve 99.33% and 99.24% accuracy on PTB dataset, whereas on PTB-XL dataset they achieve 84.36% and 84.24% accuracy, respectively. Both architectures are suitable for wearable devices requiring only 20 KB of RAM. Evaluation of real hardware shows that our baseline architecture is 0.6x to 53x more energy-efficient than the state-of-the-art works on wearable devices. Moreover, our TMEX architecture further improves the energy efficiency by 8.12% (PTB) and 6.36% (PTB-XL) while maintaining similar performance as the baseline architecture. 

In autonomous vehicles (AVs), early warning systems rely on collision prediction to ensure occupant safety. However, state-of-the-art methods using deep convolutional networks either fail at modeling collisions or are too expensive/slow, making them less suitable for deployment on AV edge hardware. To address these limitations, we propose SG2VEC, a spatio-temporal scene-graph embedding methodology that uses Graph Neural Network (GNN) and Long Short-Term Memory (LSTM) layers to predict future collisions via visual scene perception. We demonstrate that SG2VEC predicts collisions 8.11% more accurately and 39.07% earlier than the state-of-the-art method on synthesized datasets, and 29.47% more accurately on a challenging realworld collision dataset. We also show that SG2VEC is better than the state-of-the-art at transferring knowledge from synthetic datasets to real-world driving datasets. Finally, we demonstrate that SG2VEC performs inference 9.3x faster with an 88.0% smaller model, 32.4% less power, and 92.8% less energy than the state-of-the-art method on the industry-standard Nvidia DRIVE PX 2 platform, making it more suitable for implementation on the edge.