V2X Anomaly Detection Model Using Surrounding Node Logs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced vehicle-to-everything (V2X) networks are susceptible to cyber attacks due to increased data exchange, and existing anomaly detection models do not consider the impact of surrounding nodes on the transmission source node, leading to decreased accuracy in anomaly determination.
Innovation Solution
A server-based system that creates an anomaly detection model using log information from V2X networks, which includes time-stamped sensor and control data, first anomaly information from the target node, and second anomaly information from surrounding nodes, to select training data for supervised learning, thereby accounting for potential adverse effects from surrounding nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing anomaly detection models are used without considering surrounding nodes, then the system complexity is low, but the anomaly detection accuracy decreases due to false negatives
Solution Approach 1:
The patent segments the anomaly detection problem into multiple components: local anomaly detection at each node and global anomaly detection considering surrounding nodes. The model separately processes local status information and surrounding node information, then combines them to make comprehensive anomaly determination, thereby improving accuracy without creating an intractably complex unified model
Solution Approach 2:
The patent adds a new dimension to the anomaly detection model by incorporating surrounding node information alongside local node status. This dimensional expansion from single-node to multi-node perspective enables the model to detect anomalies that would otherwise be missed, improving measurement precision while maintaining manageable complexity through structured data organization
2Adaptability or versatility
If more V2X communication channels are implemented, then the system functionality is enhanced, but the susceptibility to cyber attacks increases
Solution Approach 1:
The patent implements feedback mechanisms where nodes continuously monitor their own status and the status of surrounding nodes, then use this feedback information to update anomaly detection models. This ongoing feedback loop enables the system to adapt to changing communication conditions and detect cyber attacks in real-time, allowing enhanced functionality while maintaining security
Solution Approach 2:
The patent introduces anomaly detection models as intermediary components between the V2X communication channels and the decision-making system. These models act as mediators that filter and interpret communication data, identifying potential cyber attacks before they affect system functionality, thus enabling enhanced communication while protecting against harmful factors
3Reliability
If training data selection does not account for surrounding node anomalies, then the data processing is simpler, but the anomaly detection reliability is reduced due to false negatives
Solution Approach 1:
The patent applies preliminary action by pre-processing and organizing training data according to the relationships between nodes and surrounding nodes before the actual anomaly detection occurs. The data is structured in advance with clear identifiers for local and surrounding node information, enabling reliable anomaly detection while maintaining efficient processing through pre-organized data structures
Solution Approach 2:
The patent applies local quality by treating local node status information and surrounding node information as distinct quality categories. The training data selection process differentiates between these local qualities, selectively incorporating surrounding node anomaly information only when relevant, thereby improving reliability without unnecessarily complicating the entire data processing pipeline
Data Source
AI summary
A server includes memory and a processor that obtains a log created by a first node monitored on a V2X network, selects predetermined information therefrom, and performs supervised machine learning using the selected predetermined information as training data to create a model for a monitoring node that monitors the first node. The log associatively includes a time, and at the time: first node status information as indicated by at least one of sensor or control data on the first node; first anomaly information indicating whether an anomaly occurred on the first node; and second anomaly information indicating whether an anomaly occurred on a second, surrounding node of the first node that communicates with the first node. When the first anomaly information indicates an anomaly, or when the first and second anomaly information indicate no anomaly and an anomaly, respectively, the associated status information is selected as the predetermined information.