Apertura Evento

Opening Keynote

Speaker

Prof. Nicola Mazzocca, Università Federico II Napoli, Direttore del Laboratorio Embedded Systems & Smart Manufacturing (ESSM)

Apertura Evento

Opening Keynote

Speaker

Prof. Ernesto Damiani, Università degli Studi di Milano, Presidente del Consorzio Interuniversitario Nazionale per l'Informatica (CINI)

Le opportunità del PNRR

PNRR Opportunities

Speaker

Prof. Alessandro Beghi, Università di Padova, Coordinatore dell'ecosistema dell'innovazione INest e presidente del centro di competenza SMACT

3A-ITALY, An Extended Partnership Opportunity for the Circular and Sustainable Made in Italy

Le opportunità del PNRR

PNRR Opportunities

Speaker

Prof. Elisa Negri, Politecnico di Milano, con la presentazione dal titolo "3A-ITALY, An Extended Partnership Opportunity for the Circular and Sustainable Made in Italy"

Dottorati Nazionali

National Doctorates

Speaker

Prof. Giorgio Cannata, Università di Genova, Referente per il Dottorato Nazionale in Robotica e Macchine Intelligenti

Dottorati Nazionali

National Doctorates

Speaker

Prof. Barbara Caputo, Politecnico di Torino, Coordinatrice del Dottorato Nazionale su Intelligenza Artificiale in Industria 4.0

Presentazione del 7° Italian Workshop on Embedded Systems, 22-23 settembre 2022 Bari

7th Italian Workshop on Embedded Systems Presentation, September 22nd-23rd 2022, Bari, Italy

Speaker

Prof. Daniela De Venuto, Politecnico di Bari

Detecting Unknown Attacks through Unsupervised Anomaly-Based Intrusion Detection

Autori

Authors

Dr. Tommaso Zoppi, Università di Firenze
Dr. Andrea Ceccarelli, Università di Firenze
Prof. Andrea Bondavalli, Università di Firenze

Abstract

Anomaly detection aims at identifying patterns in data that do not conform to the expected behaviour, relying on machine-learning algorithms that are suited for binary classification. It has been arising as one of the most promising techniques to suspect attacks or failures, as it has the potential to identify errors due to unknown faults as well as intrusions and unseen, unknown attacks. However, building, configuring, exercising, and evaluating anomaly detection algorithms is not trivial, as it may generate misleading results. Moreover, recent Meta-Learning techniques show promising results even with unsupervised algorithms, but are difficult to understand and implement, stacking up even more complexity.

This tutorial will discuss in detail attack detection through unsupervised anomaly detection, and will review the construction of an evaluation campaign through i) the identification of the attack models and datasets, ii) the selection and discussion of unsupervised algorithms, even when iii) adopting meta-learning, iv) the identification of target metrics, v) the execution of the algorithms and vi) their comparison. Attendees will also be involved in an hands-on session where algorithms will be executed on public attack datasets thanks to RELOAD, a tool for the Rapid EvaLuation Of Anomaly Detection algorithms. The tool is primarily meant to be used by non-experts that start approaching binary classification using ML algorithms, and hides details which may be misleading for beginners, executing analyses through a simple UI.

A Factory Abstraction Architecture for I4.0

Autori

Authors

Prof. Franco Fummi, Università di Verona

Abstract

Smart factories can be seen as complex cyber-physical production systems (CPPSs). Their design, implementation, management and evaluation needs some abstraction strategies to focus on the relevant aspects instead of on the low-level details. This workshop proposes an abstraction methodology, and related tools, starting from the way to build a complete model of the CPPS, based on SysML. There is also the description of protocols (like OPC-UA) to see the CPPS as a service oriented architecture, where IIoT data are collected by an ad-hoc architecture. The model allows the automatic configuration of this data collection architecture with the generation also of the digital-twin of the production line. Finally, the integration of different Manufacturing Execution Systems (MESs) produces an integrated view through the so called Meta-MES.

Edge AI on FPGA-based MPSoC devices

Autori

Authors

Prof. Alessandro Cilardo, Università Federico II Napoli
Dr. Vincenzo Maisto, Università Federico II Napoli

Abstract

Current Multi-Processor System-on-Chip (MPSoC) devices, particularly those based on hardware-reconfigurable technologies, offer unprecedented opportunities for accelerating AI workloads in edge-class systems. This tutorial will provide an in-depth presentation of the current state of MPSoC platforms featuring field-programmable gate arrays (FPGA) acceleration, including advanced hardware/software design flows and reference solutions for AI applications. The tutorial will include a demo addressing Xilinx technologies, focusing on the use of an advanced design solution targeted at CNN inference acceleration, the Deep Learning Processor Unit (DPU), and discussing a few opportunities coming from Dynamic Function eXchange (DFX) based on hardware reconfigurability.

Designing and development of a robotic loading machine for an automated warehouse for pharmaceutical products

Autori

Authors

Prof. Luigi Palopoli, Università di Trento

Abstract

The project is related to designing and developing a robotic loading machine for an automated warehouse for pharmaceutical products. The products currently available in the market have severe limitations in the loading operation's throughput and the ability to manage boxes with a cylindrical shape. UNITN and FbK have joined their efforts in developing the machine from the ground up. An innovative layout for the machine, the adoption of state-of-the-art robotic technologies, and the development of an innovative perception subsystem were the areas where the two research centres produced the most tangible results. The machine has been on display at an important fair and has become one of the most sophisticated products in the customer's portfolio.

HEPSYCODE Demo: HW/SW CO-DEsign of HEterogeneous Parallel dedicated SYstems

Autori

Authors

Prof. Luigi Pomante, Università degli Studi dell'Aquila
Dr. Vittoriano Muttillo, Università degli Studi dell'Aquila

Abstract

"In the last few years, the spread and importance of embedded systems have been even more increasing but it is still not yet possible to completely engineer their system-level design flow." The previous sentence opens the main abstract of the HEPSYCODE methodology and tool demo (www.hepsycode.com). But, what is HEPSYCODE and what are the main issues related to this methodology? To answer such a question, it is needed to know the main aspects related to HW/SW Co-Design in terms of design flow and activities within the context of the so-called "Electronic Design Automation" (EDA) domain. By defining the "design flow" as a sequence of steps that lead to the implementation of an application-specific, possibly embedded, digital electronic device, each step can be partially or fully automated using a set of SW tools (i.e. EDA Tools). In such a context, the choice of the best HW/SW implementation technologies can be very critical in terms of non-functional constraints and time to market. This problem, otherwise, has been simply described as a trade-off among various design metrics (performance, power, size, cost, etc.) and, nowadays, there is no difference between what hardware or software can implement from the functional point of view. Furthermore, today, most of the time is spent for simulation activities (using different modeling languages at a different level of abstractions) and to validate and verify the possible solutions taken into account by the designers. So, HEPSYCODE offers an innovative methodology and a toolchain to manage the main issues related to the design of complex embedded systems taking into account different metrics to produce as output a suitable solution considering different application domains such as automotive, aerospace, railway, etc.

Specifying and Testing Railway Interlocking Systems

Autori

Authors

Arturo Amendola, Rete Ferroviaria Italiana (RFI)
Alessandro Arenella, Rete Ferroviaria Italiana (RFI)
Roberto Cavada, Fondazione Bruno Kessler (FBK), Digital Industry Center
Alessandro Cimatti, Fondazione Bruno Kessler (FBK), Digital Industry Center
Mirco Franzago, Rete Ferroviaria Italiana (RFI)
Alberto Griggio, Fondazione Bruno Kessler (FBK), Digital Industry Center
Fitsum Kifetew, Fondazione Bruno Kessler (FBK), Digital Industry Center
Davide Prandi, Fondazione Bruno Kessler (FBK), Digital Industry Center
Giuseppe Scaglione, Rete Ferroviaria Italiana (RFI)
Angelo Susi, Fondazione Bruno Kessler (FBK), Digital Industry Center
Matteo Tessi, Rete Ferroviaria Italiana (RFI)

Abstract

Functional specifications of complex embedded systems are often written in natural language but can be ambiguous and subject to different interpretations. This phenomenon emerges in several domains such as avionics and automotive, and is particularly evident in the specification, verification and testing of railway Interlocking systems. Different regulations and technical specifications expressed in natural language documents must be reconciled and interpreted to design and evolve digital systems.

We propose a Model-Based, tool-supported methodology for the specification, implementation verification and testing of railway Interlocking system, to ensure product standardization, smooth specification of requirements, and automated code generation and verification of the system. The approach relies on a Controlled Natural Language (CNL) to support railway experts having deep knowledge on regulations and provisions in writing the specifications of interlocking procedures and the related test suites using their own jargon. From the CNL specifications, models in SysML and C/Python code are automatically generated, thus retaining full traceability between all the artefacts. Finally, test suites can be specified using the CNL to support railway experts during the validation of the overall interlocking system. The approach is supported by the tools AIDA and TOSCA that support the specification of the Interlocking system and its testing respectively.

Human tracking in industrial environments

Autori

Authors

Prof. Marco Cristani, Università di Verona
Dr. Federico Cunico, Università di Verona

Abstract

The demo will present the tracking application, in which a subject will be detected in the ICE lab and mapped on a 2D plant, continuously over time, following the movements within the laboratory, using the network of cameras in ICE.

Context-Aware-Privacy Policy in Industry 4.0

Autori

Authors

Prof. Elisa Quintarelli, Università di Verona
Prof. Federica Paci, Università di Verona
Dr. Stefano Centomo, Università di Verona

Abstract

In this demo, we introduce a novel model for privacy policies in Industry 4.0 that leverages the context to selectively enable the access of potentially sensitive data. As a practical case study, we show how our approach allows the use of automatic sensing for workplace safety reasons while guaranteeing the privacy of each individual in the factory, be they employees or visitors.

Extending IoT across a cluster of FPGAs: the Gluon-board

Autori

Authors

Prof. Roberto Giorgi, Università degli Studi di Siena
Dott.Ric. Marco Procaccini, Università degli Studi di Siena

Abstract

The Smart-Home and Smart-Videosurveillance represent good examples of the need to scale IoT devices from small contexts to larger ones: e.g., edge devices that monitor from a single home to a larger building or a Videosurveillance system with adaptive capabilities that depend on the complexity of the scene to be monitored. This demo shows the Gluon-Board that includes: i) a flexible choice of plug-in modules encompassing a variety of Xilinx Zynq Ultrascale+; ii) 4x20Gbit/s inexpensive interconnects to build FPGA-clusters made of up to 255 Gluon-boards (without the need of any external hardware, such as switches/routers); iii) a full software stack based on Ubuntu LTS derived from the AXIOM project (https://www.axiom-project.eu).We will demonstrate the Gluon-Board performance and fault-tolerance capabilities when hot-plug adding/removing a board (e.g., removing one for replacing a faulty board or adding one for improving the system capabilities). The Gluon-Board has the possibility of distributing threads across the boards to achieve a seamless execution even in the presence of faults or modifications of the cluster structure while using the same application binary.

Pervasive monitoring through wearable devices in industrial context

Autori

Authors

Dr. Cristian Turetta
Dr. Florenc Demrozi
Prof. Graziano Pravadelli

Abstract

Health informatics considers the acquisition of health-related data using unobtrusive sensors and wearable technologies to be a cornerstone. Sensors can be woven or incorporated into clothing, accessories, and the living environment, allowing for the acquisition of health data to be seamless and widespread in everyday life. Moreover, in industrial scenarios, wearable devices are pervasive instruments able to capture information in locations not covered by any other monitoring systems, such as video-based systems. For such purpose, a wearable-based approach is integrated into the ICE laboratory to reduce the limitations of such systems and provide ubiquitous coverage concerning users' activity in the industrial plant.

Containerization and Orchestration of Software for Autonomous Mobile Robots and its verification environment on Edge-Cloud architectures

Autori

Authors

Dr. Francesco Lumpp, Università di Verona
Dr. Samuele Germiniani, Università di Verona
Prof. Graziano Pravadelli, Università di Verona
Prof. Nicola Bombieri, Università di Verona
Prof. Franco Fummi, Università di Verona

Abstract

The demo presents a platform based on Docker and Kubernetes that enables containerization and orchestration of ROS-based robotic SW applications and the corresponding verification environment for heterogeneous and hierarchical Edge-Cloud architectures. The results of the platform will be presented through a dashboard, which will show, in real-time, information of the computational resources (CPU, memory, network bandwdith, etc.), the actual orchestration of the software and verification checkers across the cluster nodes, and the corresponding assertions status for a mobile robot (Robotnik Kairos) in working conditions.

Interactive mixed reality applications for Industry 4.0

Autori

Authors

Dr. Marco Emporio, Università di Verona
Dr. Ariel Caputo, Università di Verona
Dr. Thomas de Marchi, Università di Verona
Dr. Alessandro Corradi, Università di Verona
Prof. Andrea Giachetti, Università di Verona

Abstract

Prof. Andrea Giachetti will present two demos:

The first demo will introduce one or more mixed reality applications developed within the ICE lab or in ongoing collaborative projects, describing and motivating the specific interaction design and demonstrating potentialities and limits;

The second demo combines the human pose coordinates provided by BeFine with human activity information extrapolated through wearable devices. As a result, the wearable device worn by the human subject enables his/her identity recognition through the simultaneous detection of the specific pose (captured by the BeFine cameras) and of the activity (through the wearable device).

Offloading Real-time Multi-camera 3D Human Pose Estimation on Edge Computing Devices

Autori

Authors

Mirco De Marchi, Università di Verona
Michele Boldo, Università di Verona
Enrico Martini, Università di Verona
Prof. Nicola Bombieri, Università di Verona
Prof. Franco Fummi, Università di Verona

Abstract

BeFine is a real-time human pose estimation platform in which the 3D poses are estimated through a network of edge computing devices, each one composed by an RGB-D camera and a 3D HPE application. A centralized aggregator collects the multiple data flows through a standard protocol (i.e., MQTT) and merges them, in real time, through a pipeline of filtering, clustering and association algorithms. It addresses network communication issues (e.g., delay and bandwidth variability) through a two-levels synchronization, and supports both single and multi-person pose estimation. The human poses will be published on a Kafka topic. A visualizer will subscribe to this topic and will show poses overlaid on the real scene.

Data driven anomaly detection for mobile robots in Industry 4.0

Autori

Authors

Dr. Alberto Castellini, Università di Verona
Dr. Cristian Morasso, Università di Verona
Dr. Francesco Trotti, Università di Verona
Prof. Alessandro Farinelli, Università di Verona

Abstract

We apply an anomaly detection method based on Hidden Markov Models to identify anomalous behaviours of a Kairos mobile robot moving in the ICE lab, a research laboratory that faithfully represents a modern production chain based on Industry 4.0. The model of the nominal behaviour of the robot has been generated in a data-driven way, namely, learning the parameters of an HMM from data acquired during the normal functioning of the robot. Anomalies are then detected comparing the distributions of new data acquired from robot sensors (e.g., position, velocity, orientation, lidar signals) with the distributions of the emission probabilities of the HMM. In particular, the Hellinger distance is used as a measure of dissimilarity between distributions. The approach was applied to detect an anomaly due to wrong observations of the lidar when the robot is close to a window. A video shows in real-time how the anomaly has been detected by the proposed approach.

A High-interaction Physics-aware Honeynet for Industrial Control Systems

Autori

Authors

Dr. Marco Lucchese, Università di Verona
Prof. Massimo Merro, Università di Verona
Prof. Federica Paci, Università di Verona

Abstract

Industrial control systems (ICSs) automate industrial processes in many critical domains such as electric power distribution, nuclearpower production, and water supply. The increasing connectivity andintegration of ICSs with organizations corporate networks have madethem vulnerable to cyber-attacks that aim to introduce perturbationsin the controlled physical processes. In this context, honeypots can bea solution to defend against attacks targeting ICSs and to discover newattack strategies. Actually, honeypots for ICSs have done several progresses in the last years, but they still have limitations when dealing withsophisticated attack scenarios. In this paper, we propose HoneyICS, anhigh-interaction, physics-aware, scalable, reconfigurable and extensiblehoneynet for ICSs that can be deployed both in research environmentsto analyze new attacks, and in production environments to detect andpossibly stop attacks to real ICS/SCADA devices. We have conducted aseries of experiments to show the effectiveness of HoneyICS in deceivingattackers and collecting information about their attack strategies.

Towards the acceleration of AI on MPSoC embedded systems and open platforms

Autori

Authors

Dr. Maisto Vincenzo, Università Federico II Napoli,
Dr. Mercogliano Stefano, Università Federico II Napoli,
Dr. Rocco di Torrepadula Franca, Università Federico II Napoli

Abstract

Deep Neural Networks in edge applications demand several inference operations. FPGA-based MPSoC platforms may play a competitive role in the accelerations of these applications because of higher energy efficiency, with respect to GPU devices. Furthermore, their hardware reconfiguration capabilities offer the possibility to provide diversity in the pool of available accelerators.

FPGA fabric in MPSoCs platform can be further leveraged to integrate open-source RISC-V architectures. Indeed, the great momentum of the RISC-V ISA is driving both academia and industry to shift towards a more flexible and economic solution. This ISA provides custom instruction extensions for special purpose computation, enabling cutting-edge security architecture solutions and effective deep-learning datasets protection.

Digital Twins: innovative applications, architectural aspects and open challenges

Autori

Authors

Dr. Alessandra Somma, Università Federico II Napoli,
Dr. Francesco Vitale, Università Federico II Napoli

Abstract

Based on the Industrie 4.0 technologies and primarily used by industrial and engineering companies, the Digital Twin (DT) is a dynamic and self-evolving virtual replica of a physical object or process, characterized by a bi-directional seamless communication that allows real-time data sharing between physical and digital worlds. During the last decade, several DT applications in many different domains have been proposed, ranging from manufacturing to healtchare (e.g., Human Digital Twin (HDT)), from security analysis and monitoring to anomaly detection in Industrial Internet of Things (IIoT).

Despite their remarkable industrial adoption and their strategic relevance, there are still several open issues and challenges related to the set-up and engineering of Digital Twins (DTs) that must be further investigated. Among them, the lack of a reference architecture for DT implementation and the need for adequate solutions able to cope with the inherent security threats to which DTs are subject urge particular attention, in addition to the issues related to the creation of realistic and reliable models of DTs, the management of ethical issues raised by the exchange of data, and the cost of development in terms of necessary capital and human resources.

Concerning the above mentioned open issues, our research activities have been primarily focused on investigating the architectural aspects of DT implementation. In this regard, we have proposed a layered software architecture integrating all relevant DT functionalities and services derived from the analysis of the state of art. The effectiveness of the proposed architecture has been demonstrated by means of a real-life industrial case study developed by Hitachi Rail that provides social distance monitoring via a DT-based approach. The proposed architecture has been further developed by mapping each service onto the autonomic computing MAPE-K (Monitor-Analyze-Plan-Execute over a shared Knowledge) feedback loop steps, in order to enable decentralization and self-adaptiveness. Finally, in order to address security and trustworthiness of virtual replicas, we are exploring the integration of Distributed Ledger Technologies (DLTs) and Physical Unclonable Functions (PUFs) into the proposed DT architecture.

Anomaly detection in IIoT through machine learning and process mining

Autori

Authors

Dr. Fabrizio De Vita, Università Federico II Napoli,
Dr. Francesco Vitale, Università Federico II Napoli

Abstract

Industrial Internet of Things (IIoT) applications are playing a key role in the Industry 4.0 context due to their deployment leading to increased context awareness, low response times, and the opportunity to collect and process data originating from heterogeneous environments.

Analysing data collected from devices deployed within these application drive insightful activities aiming at discovering anomalous behaviour and applying timely recovery routines for steering the application state to acceptable conditions. The proposed methodology aims at highlighting anomalies out of IIoT time-series data through the joint use of machine learning dimensionality reduction algorithms and process mining routines. Results have shown good detection performances, with accuracy figures as high as 96.4%.

Enhancing the Trustworthiness of Deep Neural Networks on Cyber-physical Systems

Autori

Authors

Dr. Giulio Rossolini, Scuola Superiore Sant'Anna,
Prof. Alessandro Biondi, Scuola Superiore Sant'Anna,
Prof. Giorgio Buttazzo, Scuola Superiore Sant'Anna

Abstract

Nowadays, deep neural networks yield impressive performance in computer vision tasks such as semantic segmentation and object detection. These remarkable results have encouraged the use of deep learning models also in safety-critical cyber-physical systems (CPS), such as autonomous robots and cars. However, the trustworthiness of neural networks is often questioned by the existence of adversarial attacks, especially those performed in the physical world, which are most relevant to CPS. Such attacks are usually performed by means of adversarial objects, most often in the form of patches or posters, which are capable of corrupting the model outcome when processed as a part of the input image.

This research study tackles two important concerns in the field of trustworthy AI, that is, the evaluation and the robustness of DDNs against adversarial attacks performed in the real world.

Bio-inspired reinforcement learning for autonomous UAV: Comparing Spiking versus Artificial Neural Networks in an accelerated training environment

Autori

Authors

Dr. Luca Zanatta, Università di Bologna,
Dr. Alberto Musa, Università di Bologna,
Dr. Francesco Barchi, Università di Bologna,
Dr. Andrea Bartolini, Università di Bologna,
Prof. Andrea Acquaviva, Università di Bologna

Abstract

Spiking Neural Networks (SNN) are gaining more interest from the scientific community thanks to the promise of greater energy-efficient and greater computational power. This poses several challenges as today's SNN training for RL is based on Artificial Neural Network (ANN) training and then conversion from ANN to SNN, which does not leverage SNN event-based processing inherent capabilities. The work compares an ANN and an SNN in an event-camera-based obstacle avoidance task, trained with Reinforcement Learning (RL) using the Deep Q-Learning (DQL) algorithm. We create an experimental setup composed of Unreal Engine 4, AirSim, and an event camera that simulates a real-world obstacle avoidance environment. Additionally, we train an SNN with a gradient-based training method enabling the use of all their expressiveness even in the training phase, showing comparable performance between the ANN and the SNN. To the large training time due to the real-time simulation in AirSim we developed an acceleration method which leads to an effective trade-off between speed-up and accuracy of the training.

Model-Based Design of Service-oriented Manufacturing Systems

Autori

Authors

Dr. Sebastiano Gaiardelli, Università di Verona,
Dr. Stefano Spellini, Università di Verona,
Dr. Michele Lora, Università di Verona

Abstract

Industry 4.0 transitions current production lines to "smart" manufacturing systems, named in the literature "Cyber-Physical Production Systems (CPPSs)". Such systems are capable of communicating with each other, acquiring and transmitting real-time production data. The aim of this work is to propose a design framework, to support the creation of such a class of systems. The proposed framework is based on the SysML language, which acts as a unified specification tool for the different viewpoints of a CPPS: automation, topology, network, production processes and, ultimately, software architectures. Specifically to the software level, the framework supports the design of architectures compliant with the Service-oriented Manufacturing (SOM) principle. In particular, the functionality provided by machines is categorized in "services", that are exposed to entire plant using a specific network protocol (i.e., OPC-UA). This pattern facilitates the design of the functionality of production systems, since the knowledge of the implemented production can be structured, abstracted and manipulated according to fixed building blocks. Different tools have been developed within the framework to exploit such models and principles, to carry out verification, analysis and synthesis of implementations. Furthermore, a model-based scheduling approach for SOM-based production architectures has been developed, to estimate and effectively implement real production cycles.

Hierarchical Digital-twin for CPPS Augmented with Different Classes of Faults

Autori

Authors

Dr. Francesco Tosoni, Università di Verona,
Dr. Sadia Azam, Università di Verona,
Dr. Nicola Dall'Ora, Università di Verona,
Dr. Enrico Fraccaroli, Università di Verona

Abstract

In the context of Industry 4.0, a factory becomes a complex and heterogeneous ecosystem, where many technologies, systems and workers cooperate to compose a smart factory. This smart factory can be modelled as a Cyber-Physical Production System (CPPS). This research proposes methodologies, algorithms, and techniques to further push functional safety in a CPPS by considering in a holist way its cyber and physical parts. The project supports the design of a hierarchical digital twin of a CPPS and it automatizes its creation through different solutions. Moreover, all the levels of the digital twin will be validated with different classes of faults covering both cyber and physical parts.

Using ensemble models for emergency prevention in hydroelectric power plants with dam

Autori

Authors

Dr. Stéfano Frizzo Stefenon, Fondazione Bruno Kessler (FBK)

Abstract

In this assessment, real data from a hydroelectric power plant during a flood in southern Brazil are used. Floods are serious problems in Brazil and it is necessary to anticipate these events to better manage the consequences of these conditions. To use the data, we had meetings with the plant's management and received information from the plant's hydraulic control. For a global evaluation, several ensemble learning methods are compared. The ensemble models are superior to models based on deep learning, being more efficient (we will demonstrate this in the results).

IoT and Blockchain for Certified and Sustainable Safety-Critical Continuous Monitoring Applications

Autori

Authors

Dr. Nicola Elia, Università di Bologna,
Dr. Emanuele Parisi, Università di Bologna,
Dr. Francesco Barchi, Università di Bologna,
Dr. Andrea Bartolini, Università di Bologna,
Dr. Livio Pompianu, Università di Bologna,
Dr. Salvatore Carta, Università di Bologna,
Prof. Andrea Acquaviva, Università di Bologna

Abstract

Continuous Monitoring applications are increasingly important in various context such as predictive maintenance and real-time anomaly detection in safety-critical infrastructures. Those kind of applications often require data to be stored for large amounts of time, making cloud storage requirements become poorly sustainable. This work introduces a blockchain-based framework for certified removal of IoT data in safety-critical databases. It allows for the deployment of data-evaluation policies, in a smart contract fashion, to identify redundant/outdated measurements flowing in the database and, therefore, mark them as eligible for removal. Using a blockchain ensures that data evaluation is tamper-proof and transparent to system stakeholders. The effectiveness of the proposed framework is tested within a real-world case study producing accelerometer data of a bridge monitoring application.