Vehicle Travel Data Manipulation Detection Using Route-Aware Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting manipulation of vehicle travel data, such as those used in electronic tachographs, are susceptible to tampering, leading to inaccurate records of driving and break times, which can compromise safety by masking driver fatigue and potential accidents.
Innovation Solution
A method and system utilizing a machine learning algorithm, trained on patterns from multiple vehicles' data, including position, environmental, and control unit data, to detect manipulation of travel data by comparing actual movement profiles with expected profiles for a given route, using a device that collects and processes data from the vehicle's control unit and determines a classification result indicating potential tampering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional error detection routines are used in the control unit, then manipulation of distance signals can be detected, but manipulations can still succeed if the firmware is directly tampered with and error messages are deleted
Solution Approach 1:
The patent transitions from analyzing single-vehicle data to comparing movement profiles across multiple vehicles and routes. By adding the dimension of comparative analysis across different data sources, the system detects manipulations that would be invisible in isolated single-vehicle error detection routines.
Solution Approach 2:
The patent introduces a server as an intermediary that collects, stores, and analyzes movement profile data from multiple vehicles. This intermediary enables comparative analysis across different vehicles and routes, allowing detection of manipulated data through pattern recognition that would be impossible with local error detection alone.
2Loss of information
If firmware manipulation is performed to record fictitious movement states, then accurate-looking travel data can be generated, but machine learning analysis of movement patterns can reveal inconsistencies
Solution Approach 1:
The system implements feedback by continuously comparing actual movement profiles against expected profiles derived from multiple data sources (route data, environmental data, other vehicles). This feedback loop identifies inconsistencies in fictitious movement states, revealing manipulations that preserve basic data integrity.
Solution Approach 2:
The patent performs preliminary analysis by collecting and storing movement profile data from multiple vehicles and routes before detecting manipulations. This preliminary data gathering enables the machine learning algorithm to establish baseline patterns and detect deviations, making manipulation detection easier before it becomes too late.
3Measurement precision
If comprehensive data collection from multiple sources is performed, then manipulation detection accuracy improves, but data processing requirements and system complexity increase
Solution Approach 1:
The patent creates simplified representations (copies) of complex movement profiles by extracting key characteristics and patterns. The server stores and analyzes these profile copies rather than raw data, enabling precise manipulation detection while reducing data processing requirements and system complexity.
Data Source
Figure 1
Figure 2A~2D
Figure 3~4
AI summary
The invention relates to a method for detecting manipulation of driving data (7) of a control device (3) of a vehicle (2), comprising: obtaining the driving data (7), wherein the driving data (7) represent a movement profile (31) of the vehicle (2) on a driven route; obtaining at least one position (9) of the vehicle (2), wherein the at least one position (9) of the vehicle (2) is indicative of at least one section of the driven route; and inputting the obtained driving data (7) and the obtained at least one position (9) of the vehicle (2) into a machine learning algorithm (14), wherein the machine learning algorithm (14) is configured to determine, based on the movement profile (31) of the vehicle (2) and the obtained at least one position (9) of the vehicle (2), whether the obtained driving data (7) has been manipulated.