The automotive industry is rapidly evolving, with modern vehicles capable of perceiving their surroundings and assisting the driver in an increasing amount of situations. Car manufacturers and research groups are starting to experiment with self-driving prototypes and soon, cars will be able to drive themselves with little to no human input. To achieve self-driving capabilities, these vehicles rely on complex systems of sensors, software, and computers to navigate roads and make decisions in real-time. These systems have complete control of the vehicle's steering, braking and acceleration. Any compromise of those systems by an external attacker could have devastating consequences, including loss of control, accidents, and even fatalities. Therefore, we must provide a clear picture of the assets, attack surfaces and potential attackers of autonomous vehicles in order to define appropriate countermeasures and minimize risks of cyber-attacks. This process is known as threat modeling. This thesis focuses on threat modeling of autonomous vehicles, with a specific emphasis on the self-driving prototype developed by the Artificial Intelligence Driving Autonomous (AIDA) group, a research team within Politecnico di Milano. We review existing threat modeling frameworks and apply them to the AIDA prototype, identifying potential assets, attack surfaces, and threats. We conduct experimental validation of cyber-attacks targeting the prototype's Light Detection And Ranging (LiDAR) sensors, including sniffing, denial of service, and spoofing. These experiments validate our assumptions on a specific category of threats and their respective countermeasures. Our analysis on the AIDA prototype provide valuable insights for the development of secure and reliable autonomous vehicles, which is critical for their widespread adoption and deployment.
L'industria automobilistica è in rapida evoluzione: i veicoli moderni sono in grado di percepire l'ambiente circostante e di assistere il conducente in un numero crescente di situazioni, le case automobilistiche e i gruppi di ricerca stanno iniziando la sperimentazione su prototipi di auto a guida autonoma. Presto, le auto saranno in grado di guidarsi da sole senza l'intervento umano. Per potersi guidare da soli, questi veicoli richiedono complessi sistemi composti da sensori, software e computer e delegano loro il controllo completo del veicolo. Un'eventuale compromissione da parte di un aggressore esterno avrebbe conseguenze devastanti, tra cui perdita di controllo e potenziali incidenti stradali. Per questo motivo, è di vitale importanza fornire un quadro chiaro delle risorse da proteggere, delle superfici di attacco e dei potenziali aggressori dei veicoli a guida autonoma, al fine di definire contromisure adeguate e ridurre al minimo i rischi di attacchi informatici. Questo processo è noto come valutazione delle minacce. Questa tesi si concentra sulla valutazione delle minacce dei veicoli a guida autonoma, con un'attenzione particolare rivolta al prototipo di guida autonoma sviluppato dal gruppo di ricerca Artificial Intelligence Driving Autonomous (AIDA) del Politecnico di Milano. Si esaminano le tecniche di valutazione delle minacce esistenti e le si applicano al prototipo sviluppato da AIDA. Si conduce una valutazione sperimentale focalizzata sugli attacchi informatici rivolti ai sensori Light Detection And Ranging (LiDAR) del prototipo, tra cui sniffing, denial of service e spoofing. Questi esperimenti confermano le nostre ipotesi su questa specifica categoria di minacce e le rispettive contromisure. I risultati delle analisi sul prototipo sviluppato da AIDA forniscono indicazioni rilevanti per la progettazione di veicoli a guida autonoma sicuri e affidabili, condizione fondamentale per la loro adozione e diffusione su larga scala.
Threat modeling of autonomous vehicle security: a case study on the AIDA self-driving prototype
Boccia, Giuseppe
2023/2024
Abstract
The automotive industry is rapidly evolving, with modern vehicles capable of perceiving their surroundings and assisting the driver in an increasing amount of situations. Car manufacturers and research groups are starting to experiment with self-driving prototypes and soon, cars will be able to drive themselves with little to no human input. To achieve self-driving capabilities, these vehicles rely on complex systems of sensors, software, and computers to navigate roads and make decisions in real-time. These systems have complete control of the vehicle's steering, braking and acceleration. Any compromise of those systems by an external attacker could have devastating consequences, including loss of control, accidents, and even fatalities. Therefore, we must provide a clear picture of the assets, attack surfaces and potential attackers of autonomous vehicles in order to define appropriate countermeasures and minimize risks of cyber-attacks. This process is known as threat modeling. This thesis focuses on threat modeling of autonomous vehicles, with a specific emphasis on the self-driving prototype developed by the Artificial Intelligence Driving Autonomous (AIDA) group, a research team within Politecnico di Milano. We review existing threat modeling frameworks and apply them to the AIDA prototype, identifying potential assets, attack surfaces, and threats. We conduct experimental validation of cyber-attacks targeting the prototype's Light Detection And Ranging (LiDAR) sensors, including sniffing, denial of service, and spoofing. These experiments validate our assumptions on a specific category of threats and their respective countermeasures. Our analysis on the AIDA prototype provide valuable insights for the development of secure and reliable autonomous vehicles, which is critical for their widespread adoption and deployment.File | Dimensione | Formato | |
---|---|---|---|
2024_12_Boccia_Thesis.pdf
non accessibile
Descrizione: Thesis
Dimensione
11.1 MB
Formato
Adobe PDF
|
11.1 MB | Adobe PDF | Visualizza/Apri |
2024_12_Boccia_Executive_Summary.pdf
non accessibile
Descrizione: Executive Summary
Dimensione
3.38 MB
Formato
Adobe PDF
|
3.38 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/230468