In recent years, cyber-attacks have increased significantly in both volume and sophistication, making the detection of security violations a crucial feature in computer systems. This is particularly true in the Internet of Things (IoT), where devices are vulnerable to failures and malicious attacks due to their resource-constrained nature. Given the proliferation of new security threats, anomaly-based detection approaches are essential for intrusion detection and prevention systems to effectively defend against attackers. This paper proposes an information-theoretic approach based on entropy to establish an anomaly detection model. A real case study in IoT networks based on Routing Over Low power and Lossy networks (RPL) illustrates the application of the proposed approach. Preliminary experimental results demonstrate that our method is both practical and extendable.
Currently, quality assurance in the food supply chain is predominantly conducted manually with traditional destructive methods. Best before date forecasts depend mainly on time of arrival without including weather or cold chain. Measured data within the food supply chain is not digitized and can therefore not be aggregated for analysis. We propose a new architecture for improving sustainability of the food supply by combining innovative measurement techniques with a trusted cloud edge infrastructure. Forecasts based on digital twins and a shared federated data space for freshness data serve to optimize supply chain processes and reduce the loss of fresh produce along the supply chain. In this paper, we present a suitable system architecture with independent data spaces and flexible implementations for trusted local deployments.
The proliferation of data from IoT devices has introduced challenges for efficient and accurate information retrieval in distributed Cloud Data Centers. This paper presents a Hybrid Distributed Information Retrieval (H-DIR) model optimized for IoT environments using advanced semantics and a hybrid query approach. The model aims to improve retrieval precision and efficiency by leveraging semantic concepts and contextual metadata, while also enhancing security and data privacy. Applied in a real industrial setting for gardening services, the H-DIR model demonstrates potential to increase revenue through economic quantification of services and value creation via faster preventive measures.
In IoT-based systems, managing risk effectively is crucial for maintaining operational continuity, especially when faced with attacks on physical assets such as sensors or actuators. These situations demand immediate and coordinated human responses to minimize damage. This paper explores the initial challenges in managing and predicting the evolution of human-involved workflows in such contexts, especially in the absence of comprehensive historical data for model training. We propose a novel approach leveraging digital twins, modeled using BPMN 2.0, to facilitate knowledge transfer across semantically similar workflows. This methodology allows for preliminary predictions regarding unmonitored workflows by utilizing insights from previously established ones. We present initial results from both synthetic and real-world data, which suggest the potential of our approach to enhance risk management practices in IoT settings. These findings are intended to foster discussion and further exploration within the research community.
The increasing pervasiveness of Artificial Intelligence (AI), and Convolutional Neural Networks (CNNs) in edge-computing and Internet of Things applications pose several challenges, including the hunger for computational and power resources of predictive models, and their robustness w.r.t. security threats, e.g., adversarial attacks. As for the former, the approximate computing emerged as one of the most promising solutions to lower the computational effort of AI, since the output of approximate application is usually barely distinguishable from the exact one. Nevertheless, alterations to predictive models through approximation may actually jeopardize inner characteristics of CNNs, such as their adversarial robustness, that is their ability to discern legitimate inputs from systematically crafted malicious ones. In this paper, we investigate the vulnerability of the approximate CNNs to adversarial attacks. Specifically, we target approximate CNNs while resorting to different adversarial attacks in an aversion scenario, and we empirically prove approximation may actually compromise adversarial robustness.
