The automotive industry has undergone significant evolution and expansion in recent years, this has lead to the creation of progressively intricate in-vehicle networks and an increased quantity of on-board Electronic Control Units (ECUs). The controller area network protocol, used for on-board communication in vehicles, is generally accepted as the industry standard. The CAN protocol and its newer version, Controller Area Network with Flexible Data-rate (CAN FD) protocol, are still chosen due to their dependable and effective real-time transmission capabilities. The degree of connectivity that cars have with other internal and external devices such as cellular connections, WiFi, Bluetooth, and GPS (Global Positioning System) connections has significantly increased in the automotive industry. Furthermore, tests and development are now underway for autonomous self-driving automobiles. Security was either completely ignored or given minimal thought during the initial creation of the CAN and other in-vehicle network protocols. In particular, there is no authentication for communications in CAN, so attackers can send spoofed messages across the bus. Though these messages can be detected by the current Intrusion Detection Systems (IDSs), an attacker can circumvent this security by first disconnecting the target Electronic Control Unit (ECU) and then spoofing the communications. Bit injection attacks are a vulnerability in the Controller Area Network (CAN) protocol that let an intruder disconnect devices which are connected to the bus by inserting bits at certain times. These attacks have the potential to interfere with a vehicle’s routine operation or, more broadly, to target safety components. Considering the attacker’s potential to modify the bus at the bit level, bit injection attacks are more subtle than other forms of attacks. In this work, we propose an Intrusion Detection system (IDS) that provide defence against attacks in which the attacker inserts an error flag composed of six dominant bits (0s) to force the victim ECU to discard its own frame, eventually disabling itself due to the fault handling behavior of CAN. Our approach detects such attacks with an accuracy of up to 97 percent, without requiring any modification to already existing ECUs or to the network architecture.
Negli ultimi tempi l’industria automobilistica ha subito una significativa evoluzione ed espansione, ciò ha portato alla creazione di reti di bordo progressivamente complesse e ad un aumento della quantità di unità di controllo elettroniche (ECU) di bordo. Il protocollo di rete utilizzato per la comunicazione a bordo dei veicoli, generalmente accettato come lo standard del settore, è il protocollo CAN e la sua versione più recente, Controller Area Network Flexible Data rate. Questi vengono ancora scelti per la loro affidabilità e efficacia di trasmissione in tempo reale. Il grado di connettività che le auto hanno con altri dispositivi interni ed esterni come connessioni cellulari, connessioni WiFi, Bluetooth e GPS (Global Positioning System) è notevolmente aumentato nel settore automobilistico. Inoltre, test e sviluppo sono attualmente in corso per le automobili autonome a guida autonoma. Il problema della sicurezza è stato completamente ignorato durante la creazione iniziale del CAN e altri protocolli di rete a bordo del veicolo. In particolare, non è prevista alcuna autenticazione per le comunicazioni in CAN, abilitando gli aggressori a inviare messaggi contraffatti sul bus. Sebbene questi messaggi possano essere rilevati dagli attuali sistemi di rilevamento delle intrusioni (IDS), un utente malintenzionato può aggirare questa misura scollegando prima l’unità di controllo elettronica di destinazione (ECU) e quindi falsificando le comunicazioni. Gli attacchi bit injection rappresentano una vulnerabilità nel protocollo Controller Area Network (CAN) che consentono a un intruso di disconnettere i dispositivi collegati al bus inserendo bit in specifici momenti. Questi attacchi possono potenzialmente interferire con il normale funzionamento del veicolo o mirare ai componenti di safety. In questo lavoro proponiamo un sistema di rilevamento delle intrusioni (IDS) che fornisce una misura di difesa contro attacchi in cui l’aggressore genera un messaggio di errore, costringendo la ECU vittima a scartare il proprio frame, eventualmente disabilitandosi a causa del comportamento di gestione dei guasti del protocollo CAN. Il nostro approccio rileva tali attacchi con una precisione fino a 97 percento, senza richiedere alcuna modifica alle ECU già esistenti o all’architettura di rete.
CANPak: an intrusion detection system against stealthier attacks for controller area network
ABBASI, SIKANDAR MEHMOOD
2023/2024
Abstract
The automotive industry has undergone significant evolution and expansion in recent years, this has lead to the creation of progressively intricate in-vehicle networks and an increased quantity of on-board Electronic Control Units (ECUs). The controller area network protocol, used for on-board communication in vehicles, is generally accepted as the industry standard. The CAN protocol and its newer version, Controller Area Network with Flexible Data-rate (CAN FD) protocol, are still chosen due to their dependable and effective real-time transmission capabilities. The degree of connectivity that cars have with other internal and external devices such as cellular connections, WiFi, Bluetooth, and GPS (Global Positioning System) connections has significantly increased in the automotive industry. Furthermore, tests and development are now underway for autonomous self-driving automobiles. Security was either completely ignored or given minimal thought during the initial creation of the CAN and other in-vehicle network protocols. In particular, there is no authentication for communications in CAN, so attackers can send spoofed messages across the bus. Though these messages can be detected by the current Intrusion Detection Systems (IDSs), an attacker can circumvent this security by first disconnecting the target Electronic Control Unit (ECU) and then spoofing the communications. Bit injection attacks are a vulnerability in the Controller Area Network (CAN) protocol that let an intruder disconnect devices which are connected to the bus by inserting bits at certain times. These attacks have the potential to interfere with a vehicle’s routine operation or, more broadly, to target safety components. Considering the attacker’s potential to modify the bus at the bit level, bit injection attacks are more subtle than other forms of attacks. In this work, we propose an Intrusion Detection system (IDS) that provide defence against attacks in which the attacker inserts an error flag composed of six dominant bits (0s) to force the victim ECU to discard its own frame, eventually disabling itself due to the fault handling behavior of CAN. Our approach detects such attacks with an accuracy of up to 97 percent, without requiring any modification to already existing ECUs or to the network architecture.| File | Dimensione | Formato | |
|---|---|---|---|
|
2024_04_Abbasi_Thesis.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Bit injection attacks are a vulnerability in the Controller Area Network (CAN) protocol that let an adversary disable devices connected to the bus by inserting bits at certain times. These attacks can be used to interfere with a vehicle's routine operation or, more broadly, to target safety components. In this work, we propose an Intrusion Detection system (IDS) that provide defence against attacks in which the attacker inserts an error flag composed of six dominant bits (0s) to force the victim ECU to discard its own frame, eventually disabling itself due to the fault handling behavior of CAN, even though the actual frame of victim was correct and was without errors. Our device detects such sorts of attack, marked them as intrusive/ attack packet.
Dimensione
2.74 MB
Formato
Adobe PDF
|
2.74 MB | Adobe PDF | Visualizza/Apri |
|
2024_04_Abbasi_Executive_Summary.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Executive Summary of the Thesis
Dimensione
547.73 kB
Formato
Adobe PDF
|
547.73 kB | 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/10589/219051