The increasing adoption of AI models has driven applications toward the use of hardware accelerators to meet high computational demands and strict performance requirements. Beyond consideration of performance and energy efficiency, explainability and reliability have emerged as pivotal requirements, particularly for critical applications such as automotive, medical, and aerospace systems. Among the various AI models, Decision Tree Ensembles (DTEs) are particularly notable for their high accuracy and explainability. Moreover, they are particularly well-suited for hardware implementations, enabling high-performance and improved energy efficiency. However, a frequently overlooked aspect of DTEs is their reliability in the presence of hardware malfunctions. While DTEs are generally regarded as robust by design, due to their redundancy and voting mechanisms, hardware faults can still have catastrophic consequences. To address this gap, we present an in-depth reliability analysis of two types of DTE hardware accelerators: classical and approximate implementations. Specifically, we conduct a comprehensive fault injection campaign, varying the number of trees involved in the classification task, the approximation technique used, and the tolerated accuracy loss, while evaluating several benchmark datasets. The results of this study demonstrate that approximation techniques have to be carefully designed, as they can significantly impact resilience. However, techniques that target the representation of features and thresholds appear to be better suited for fault tolerance.

• A modular redundancy-based approximation is proposed for decision tree ensembles. • Modular redundancy is used to select only a subset of trees for classifying each class label. • This strategy allows aggressive approximation while preserving accuracy. • The effectiveness of the solution is demonstrated and shown. The deployment of machine learning models at the edge is crucial for enabling low-latency decision-making, optimizing resource utilization, and enhancing data confidentiality. Random Forest classifiers have proven to be highly accurate while offering computationally efficient inference, making them well-suited for resource-constrained edge devices. However, as the volume of training data grows, the complexity and size of these models also increase, limiting their deployment in edge computing scenarios. In order to address this challenge, we propose a novel approximation strategy for Random Forest classifiers leveraging on the concept of modular redundancy. In particular, our approach imposes that each target class is determined by only a subset of trees in a modular redundant fashion. This allows to prune from each tree the leaves related to no-longer relevant classes, significantly reducing the size of the model. To achieve an optimal balance between accuracy and resource savings with minimal computational time, we introduce an heuristic algorithm that determine the best subset of trees for each class. We evaluate our approach on multiple UCI machine learning datasets using a hardware accelerator for tree ensembles, demonstrating its effectiveness. The result shows that, on average, a 2.5% reduction in accuracy leads to save up to 50% in hardware overhead and energy consumption.

Nowadays, predicting public transport passenger flow (PF) is essential to optimize service planning and provide information to commuters. However, while current research focuses on enhancing accuracy using advanced models, like recurrent and graph neural networks, other key aspects, such as interpretability and computational efficiency, are often neglected. To fill this gap, we propose a framework to design on-board explainable PF prediction based on eXtreme Gradient Boosting (XGBoost). This framework enhances model interpretability and reduces computational costs, making the resulting PF predictive model suitable for inference on low-end devices. The proposed framework is validated on a real-world dataset from a major Italian city, involving 25 buses. Our results show that the framework achieves performance comparable to Convolutional Neural Networks (CNNs), with only a 0.2% percentage increase in Mean Absolute Percentage Error (MAPE). Additionally, it significantly reduces both inference time and energy consumption, with percentage decrease of 63%. Finally, we present examples to illustrate the interpretability of the predictions using the SHapley Additive exPlanation (SHAP) method. • Passenger flow (PF) prediction supports transport companies and passengers. • Currently, the main objective of research on PF is maximizing accuracy. • Our XGBoost-based framework provide accurate and explainable predictions. • The resulting model can be effectively executed on low end devices.

Multiple Classifier Systems (MCSs) have been in-creasingly designed to take advantage of hardware features, such as high parallelism and computational power, to guarantee higher throughput and lower latency. Although the combination of multiple classifiers leads to high classification accuracy, the required area overhead makes the design of a hardware accelerator unfeasible, hindering the adoption of commercial configurable devices. For this reason, in this paper, we exploit the Approximate Computing (AxC) design paradigm to automatically generate approximated hardware implementations of MCSs by trading hardware area overhead off for classification accuracy. In particular, we propose an algorithm that identifies the resiliency source of the model and uses it to introduce approximation with minimum accuracy loss.

In many Internet of Things (IoT) applications, resource-constrained devices often collaborate in groups for the acquisition, transmission, and management of sensitive information. To uphold the security of these operations, symmetric encryption algorithms are commonly employed due to their efficiency and speed. Nevertheless, establishing a key management mechanism, that accommodates the distinctive features of the IoT domain, remains an ongoing challenge. This paper introduces Group-Key PHEMAP, a novel Physically Unclonable Function (PUF)-based protocol for group key management in IoT applications. The proposed protocol relies solely on lightweight operations for group key management and supports dynamic membership without leveraging additional cryptographic keys. We present a mathematical demonstration for security properties and a comprehensive analysis, regarding both computational and communication costs, as well as scalability property concerning the growing number of devices within the group. Finally, we validate the suitability of our proposal by resorting to the ns-3 network simulator, and, by implementing the protocol on devices representing typical characteristics of those used in IoT applications.