In the automotive industry, Electronic Control Units (ECUs) equipped with internal communication protocols have revolutionized vehicle functionality, allowing for the management of various operations, starting from basic tasks such as controlling windshield wipers to more intricate safety-related functions. However, an advanced system such as this also introduces vulnerabilities to cyber-attacks. Such attacks often exploit software weaknesses within the communication protocol components the ECUs utilize. Recent research has also underscored the susceptibility of modern vehicles to these cyber threats. It is crucial to note that the methodologies and techniques proposed by automotive researchers primarily address only one of the three essential phases of cybersecurity: Detection, Response, and Prevention, highlighting the necessity for developing new techniques and methodologies that encompass these phases and enhance the security of modern vehicles. The literature review in this regard discovered diverse critical concerns. Firstly, there is a need to compare identification techniques between traditional and quantum algorithms regarding detection time and model accuracy. Secondly, there is a notable absence of comprehensive risk assessment methodologies utilizing international standards or reports. Consequently, the absence of comprehensive methodologies or frameworks for managing cybersecurity in the automotive context renders this imperative. In response to these critical issues, this thesis introduces ANDURIL (Automotive Network Defense Unified Response Intrusion Limitation), a model distinguished by five dimensions and three operational units. Within each dimension, various techniques and methodologies have been proposed, considering the three phases of cybersecurity: Detection, Response, and Prevention. This approach enables automotive organizations to select one or more techniques from each domain to carry out detection, response, and prevention activities. These techniques involve the use of Machine Learning (ML) algorithms, a novel approach based on the Security Operation Center (SOC), and the use of Quantum ML algorithm to identify threats on the Controller Area Network (CAN) bus. In the response and prevention phase, the new approaches include a pattern designed to assess the risk of attack and propose security controls and a web application to support automotive companies in the event of an attack. However, it is important to note that the experiments conducted so far have been in vitro, and the proposed techniques need to be validated on real-world automotive embedded components (in vivo). Moreover, validating such techniques with automotive companies is a critical area that needs future exploration.
Nell’industria automobilistica, la presenza di Electronic Control Unit (ECU) dotate di protocolli di comunicazione interni ha rivoluzionato la funzionalità dei veicoli, permettendo la gestione di varie operazioni, dalle attività di base come il controllo dei tergicristalli a funzioni di sicurezza più complesse. Tuttavia, questo sistema avanzato introduce anche vulnerabilità agli attacchi informatici, che spesso sfruttano le debolezze del software all’interno delle componenti o dei protocolli di comunicazione utilizzati dalle ECU. Studi recenti hanno evidenziato problematiche dei veicoli moderni a queste minacce informatiche. È fondamentale notare che le metodologie e le tecniche esistenti proposte dai ricercatori automobilistici affrontano principalmente solo una delle tre fasi essenziali della sicurezza informatica, ovvero Rilevamento, Risposta e Prevenzione. Questo mette in evidenza la necessità di sviluppare nuove tecniche e metodologie che abbraccino tutte e tre le fasi e possano essere integrate per migliorare la sicurezza dei veicoli moderni. Inoltre, dalla revisione della letteratura emergono preoccupazioni critiche. In primo luogo, è necessario confrontare le tecniche di identificazione tra algoritmi tradizionali e algoritmi quantistici in termini di tempo di rilevamento e accuratezza del modello. In secondo luogo, si nota una notevole assenza di metodologie di valutazione del rischio complete che utilizzino standard internazionali o report. Di conseguenza, diventa imperativo colmare l’assenza di metodologie o framework completi per gestire la sicurezza informatica nel contesto automobilistico. In risposta a queste questioni critiche, questa tesi introduce ANDURIL (Automotive Network Defense Unified Response Intrusion Limitation), un modello caratterizzato da cinque dimensioni e tre unità operative. All’interno di ciascuna dimensione, sono state proposte varie tecniche e metodologie, tenendo conto delle tre fasi della sicurezza informatica: Rilevamento, Risposta e Prevenzione. Questo approccio consente alle organizzazioni automobilistiche di selezionare una o più tecniche da ciascun dominio per eseguire attività di rilevamento, risposta e prevenzione. Queste tecniche includono l’uso di algoritmi di Machine Learning (ML), un approccio innovativo basato sul Security Operation Center (SOC) e l’uso di algoritmi di Quantum ML per identificare minacce sul bus Controller Area Network (CAN). Nella fase di risposta e prevenzione, nuovi approcci includono un modello progettato per valutare il rischio di attacco e proporre controlli di sicurezza, oltre a un’applicazione web che può essere utilizzata per supportare le aziende automobilistiche in caso di attacco. È importante notare, tuttavia, che gli esperimenti condotti sono stati in vitro, ed è necessaria una validazione delle tecniche proposte su componenti embedded reali (in vivo). Inoltre, la validazione di tali tecniche con aziende automobilistiche, rappresenta un’area critica che necessita future esplorazioni.
Security Management in Automotive Environment
DE VINCENTIIS, MIRKO
2025
Abstract
In the automotive industry, Electronic Control Units (ECUs) equipped with internal communication protocols have revolutionized vehicle functionality, allowing for the management of various operations, starting from basic tasks such as controlling windshield wipers to more intricate safety-related functions. However, an advanced system such as this also introduces vulnerabilities to cyber-attacks. Such attacks often exploit software weaknesses within the communication protocol components the ECUs utilize. Recent research has also underscored the susceptibility of modern vehicles to these cyber threats. It is crucial to note that the methodologies and techniques proposed by automotive researchers primarily address only one of the three essential phases of cybersecurity: Detection, Response, and Prevention, highlighting the necessity for developing new techniques and methodologies that encompass these phases and enhance the security of modern vehicles. The literature review in this regard discovered diverse critical concerns. Firstly, there is a need to compare identification techniques between traditional and quantum algorithms regarding detection time and model accuracy. Secondly, there is a notable absence of comprehensive risk assessment methodologies utilizing international standards or reports. Consequently, the absence of comprehensive methodologies or frameworks for managing cybersecurity in the automotive context renders this imperative. In response to these critical issues, this thesis introduces ANDURIL (Automotive Network Defense Unified Response Intrusion Limitation), a model distinguished by five dimensions and three operational units. Within each dimension, various techniques and methodologies have been proposed, considering the three phases of cybersecurity: Detection, Response, and Prevention. This approach enables automotive organizations to select one or more techniques from each domain to carry out detection, response, and prevention activities. These techniques involve the use of Machine Learning (ML) algorithms, a novel approach based on the Security Operation Center (SOC), and the use of Quantum ML algorithm to identify threats on the Controller Area Network (CAN) bus. In the response and prevention phase, the new approaches include a pattern designed to assess the risk of attack and propose security controls and a web application to support automotive companies in the event of an attack. However, it is important to note that the experiments conducted so far have been in vitro, and the proposed techniques need to be validated on real-world automotive embedded components (in vivo). Moreover, validating such techniques with automotive companies is a critical area that needs future exploration.File | Dimensione | Formato | |
---|---|---|---|
Tesi_De_Vincentiis_Mirko_final.pdf
accesso aperto
Dimensione
8.98 MB
Formato
Adobe PDF
|
8.98 MB | Adobe PDF | Visualizza/Apri |
Tesi_De_Vincentiis_Mirko_final_1.pdf
accesso aperto
Dimensione
8.98 MB
Formato
Adobe PDF
|
8.98 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/210169
URN:NBN:IT:UNIBA-210169