Vehicle CAN ECU Hacking Detection Using Learned Message Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Controller Area Network (CAN) in vehicles is vulnerable to hacking, which can lead to manipulation of vehicle functions, posing risks such as theft, collisions, and fatalities, with existing mechanisms failing to identify legitimate versus compromised sources of messages.
Innovation Solution
Implementing hacking-detection software modules in ECUs that monitor parameters and generate pattern-detection mechanisms using machine-learning algorithms to identify malicious activity, encrypt and decrypt messages based on identified sources, and respond with alerts or operational restrictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing CAN message transmission mechanisms are used, then communication between ECUs is simple and efficient, but the system is vulnerable to hacking and cannot identify legitimate versus compromised sources of messages
Solution Approach 1:
The patent applies preliminary action by having ECUs pre-register their CAN identifiers with neighboring ECUs before actual message transmission. This creates a预先 established trust relationship that enables later verification of message sources without adding complex runtime verification mechanisms. The registration phase establishes the foundation for security while keeping the actual message handling simple.
Solution Approach 2:
The patent implements feedback by having receiving ECUs monitor incoming CAN messages and compare the source identifiers against their registered neighboring ECU identifiers. When a message source is not recognized, the system generates an alert and can restrict operations. This feedback loop enables automatic detection and response to potential hacking attempts without requiring complex centralized verification.
2Measurement precision
If machine-learning algorithms are implemented for pattern detection, then malicious activity can be identified, but the processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by implementing a two-tier detection approach: first, a simple rule-based verification checks if message sources are in the registered neighboring ECU list (fast, low-computation); second, machine-learning pattern detection is only activated when anomalies are detected or at lower priority. This ensures most messages are processed quickly while maintaining high detection accuracy for malicious activity.
Solution Approach 2:
The system performs preliminary registration of normal message patterns and neighboring ECU identifiers during system initialization or stable operation periods. This pre-learning phase allows the machine-learning models to be pre-trained on legitimate traffic patterns, so that during actual operation, the system only needs to detect deviations from these patterns rather than analyzing every message from scratch, significantly reducing real-time processing requirements.
3Reliability
If encryption and decryption of messages are implemented, then message integrity is protected, but the communication speed and energy consumption increase
Solution Approach 1:
The patent applies local quality by implementing encryption selectively rather than universally. Specifically, encryption and decryption are applied only to messages that originate from untrusted or remote sources, while messages from registered neighboring ECUs use simple verification. This localized application of security measures protects message integrity where needed while minimizing the energy overhead associated with cryptographic operations.
Solution Approach 2:
The system implements partial encryption by using full cryptographic verification only when pattern detection flags a message as suspicious or when communicating with non-neighboring ECUs. For routine communications between registered neighbors, the system uses lighter-weight verification mechanisms. This partial approach maintains security for critical communications while reducing overall energy consumption across the network.
Data Source
AI summary
Cybersecurity on a Controller Area Network (CAN) in a vehicle. In an embodiment, electronic control units (ECUs), connected to a CAN bus, each comprise a hacking detection system, which, during an initialization stage, transmits a message comprising a CAN identifier, used by the respective ECU, to at least one other hacking detection system, receives a message comprising a CAN identifier, used by at least one other ECU, from the other hacking detection system, monitors one or more parameters, including at least one parameter of CAN messages received by the respective ECU and transmitted by the respective ECU, and generates a pattern-detection mechanism based on the monitored one or more parameters. Then, during a detection stage, each hacking detection system monitors the one or more parameters, and detects malicious activity based on the generated pattern-detection mechanism and the one or more parameters monitored during the detection stage.