A method for data integrity verification and tamper location of autonomous vehicles

Abstract

This invention discloses a data integrity verification and tamper location method for autonomous vehicles, comprising the following steps: S1: Watermark embedding: Obtaining raw LiDAR point cloud data from a LiDAR sensor, and embedding a preset binary watermark sequence into the raw LiDAR point cloud data based on a three-dimensional quantization index modulation method to generate watermarked LiDAR point cloud data; S2: Data transmission: Transmitting the watermarked LiDAR point cloud data to an advanced driver assistance system in the vehicle network; S3: Parallel processing; S4: Tampering determination and location: Based on the comparison results of the integrity verification process, determining whether the LiDAR point cloud data has been tampered with. If it is determined to have been tampered with, the tampered LiDAR point cloud data area is located based on the data point positions where the watermark information does not match. This invention can accurately locate tampering, controllable distortion, avoid decryption delays in encryption methods, meet the real-time requirements of autonomous driving, has broad application prospects, and is conducive to widespread application.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method for verifying the data integrity and locating tampering in autonomous vehicles. Background Technology

[0002] Autonomous vehicles rely on various sensors, such as LiDAR, radar, and cameras, to perceive their surroundings. This sensor data is transmitted via a network between electronic control units within the vehicle. As vehicle connectivity increases, various communication interfaces have emerged, which also introduce potential attack surfaces. Attackers could exploit these interfaces to tamper with sensor data, such as inserting fake objects or deleting real targets, causing the autonomous driving system to make incorrect decisions, such as unnecessary braking or collisions with obstacles, seriously threatening driving safety.

[0003] Existing technologies typically employ encryption methods for data integrity verification. However, encryption requires decryption before data processing, introducing additional latency and potentially becoming a performance bottleneck for handling high-frequency, high-bandwidth LiDAR point cloud data. Furthermore, encryption methods generally cannot directly pinpoint the exact location of data tampering. Therefore, there is an urgent need to develop a data integrity verification and tampering location method for autonomous vehicles to address these technical challenges.

[0004] In view of this, the present invention is hereby proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a data integrity verification and tamper location method for autonomous vehicles. It is specifically designed for the characteristics of LiDAR point cloud data and typical attack models in autonomous driving. It can accurately locate tampering, control the distortion, avoid the decryption delay in encryption methods, meet the real-time requirements of autonomous driving, has broad application prospects, and is conducive to its widespread application.

[0006] To achieve the above objectives, the present invention provides a data integrity verification and tamper location method for autonomous vehicles, comprising the following steps: S1: Watermark embedding: Obtain raw LiDAR point cloud data from the LiDAR sensor, and embed a preset binary watermark sequence into the raw LiDAR point cloud data based on the three-dimensional quantization index modulation method to generate watermarked LiDAR point cloud data. S2: Data transmission: Transmit watermarked LiDAR point cloud data to the advanced driver assistance system in the vehicle network; S3: Parallel Processing: Receive watermarked LiDAR point cloud data from the advanced driver assistance system and perform the following operations in parallel: SS1: Object recognition processing: Directly perform object detection, classification, or tracking processing on watermarked LiDAR point cloud data; SS2: Integrity verification processing: Extract watermark information from the received LiDAR point cloud data and compare the extracted watermark information with the expected watermark information; S4: Tampering Detection and Location: Based on the comparison results of the integrity verification process, determine whether the LiDAR point cloud data has been tampered with. If it is determined to have been tampered with, locate the tampered LiDAR point cloud data area according to the data point positions where the watermark information does not match.

[0007] Preferably, in step S1, embedding the watermark based on the three-dimensional quantization index modulation method includes the following steps: SS1: Performs voxelization on the raw LiDAR point cloud data, dividing the point cloud into three-dimensional voxels of fixed size; SS2: For each voxel, determine its representative vertex; SS3: Based on the watermark bits to be embedded, use the corresponding 3D quantizer to quantize and modulate the 3D coordinates of the representative vertex, and move the representative vertex to one of the eight vertices in a cube with a side length of Δ / 2 centered on the original quantization position, where Δ is the quantization step size.

[0008] Preferably, in S1, the watermark sequence is a periodically repeating sequence, and each watermark symbol corresponds to a three-dimensional bit tuple, which is used to control the displacement direction of the representative vertex in three-dimensional space.

[0009] Preferably, in SS3 of S1, the quantization step size Δ is configured according to the target application's requirements for LiDAR data accuracy, in order to balance the impact of watermark robustness and embedding distortion on the performance of the object recognition algorithm.

[0010] Preferably, in SS2 of S3, the integrity verification process adopts a blind watermark extraction mechanism, which does not require the original LiDAR point cloud data to participate in the watermark extraction process.

[0011] Preferably, in step S4, the method for locating the tampered LiDAR point cloud data region is as follows: identify the data point indices that do not match the watermark information, and determine the bounding box of the tampered region based on the spatial coordinates corresponding to these indices.

[0012] Preferably, the tampering attack types detected by the method include spoofed object insertion and target object deletion.

[0013] The present invention provides a method for verifying data integrity and tampering location of autonomous vehicles, which has the following beneficial effects.

[0014] 1. Real-time performance and low latency: Watermark verification and object recognition are processed in parallel, avoiding the decryption delay in encryption methods, which is suitable for the real-time requirements of autonomous driving.

[0015] 2. Precise Tampering Location: It can not only detect the occurrence of tampering, but also pinpoint the specific area in the point cloud that has been modified.

[0016] 3. Controllable distortion: The impact of watermark embedding on the original data can be flexibly controlled by the quantization step size Δ, ensuring that the performance of the object recognition algorithm is not significantly affected.

[0017] 4. Highly targeted: It is specifically designed for the characteristics of LiDAR point cloud data and typical attack models in autonomous driving. Detailed Implementation

[0018] The present invention will be further described below with reference to specific embodiments to help understand the content of the present invention.

[0019] This invention provides a method for verifying data integrity and locating tampering in autonomous vehicles, comprising the following steps: S1: Watermark Embedding: Raw LiDAR point cloud data is acquired from the LiDAR sensor. Based on a three-dimensional quantization index modulation method, a preset binary watermark sequence is embedded into the raw LiDAR point cloud data to generate watermarked LiDAR point cloud data. The watermark sequence is a periodically repeating sequence, and each watermark symbol corresponds to a three-dimensional bit tuple, used to control the displacement direction of the represented vertex in three-dimensional space. Watermark embedding based on the three-dimensional quantization index modulation method includes the following steps: SS1: Performs voxelization on the raw LiDAR point cloud data, dividing the point cloud into three-dimensional voxels of fixed size; SS2: For each voxel, determine its representative vertex; SS3: Based on the watermark bits to be embedded, the 3D coordinates of the representative vertex are quantized and modulated using the corresponding 3D quantizer. The representative vertex is then moved to one of the eight vertices within a cube centered at the original quantized position with a side length of Δ / 2, where Δ is the quantization step size. The quantization step size Δ is configured according to the target application's requirements for LiDAR data accuracy to balance the impact of watermark robustness and embedding distortion on the performance of the object recognition algorithm.

[0020] S2: Data transmission: Transmit watermarked LiDAR point cloud data to the advanced driver assistance system in the vehicle network; S3: Parallel Processing: Receive watermarked LiDAR point cloud data from the advanced driver assistance system and perform the following operations in parallel: SS1: Object recognition processing: Directly perform object detection, classification, or tracking processing on watermarked LiDAR point cloud data; SS2: Integrity verification processing: Extract watermark information from the received LiDAR point cloud data and compare the extracted watermark information with the expected watermark information; the integrity verification processing adopts a blind watermark extraction mechanism, which does not require the original LiDAR point cloud data to participate in the watermark extraction process.

[0021] S4: Tampering Detection and Location: Based on the comparison results of the integrity verification process, determine whether the LiDAR point cloud data has been tampered with. If it is determined to have been tampered with, locate the tampered LiDAR point cloud data region based on the location of data points with mismatched watermark information. The method for locating the tampered LiDAR point cloud data region is as follows: identify the indices of data points with mismatched watermark information, and determine the bounding box of the tampered region based on the spatial coordinates corresponding to these indices.

[0022] The tampering attack types detected by the above method include spoofed object insertion and target object deletion.

[0023] This invention is specifically designed for the characteristics of LiDAR point cloud data and typical attack models in autonomous driving. Watermark verification and object recognition are processed in parallel, avoiding decryption delays inherent in encryption methods and making it suitable for the real-time requirements of autonomous driving. This invention can not only detect tampering but also pinpoint the specific modified area within the point cloud. By using a quantization step size Δ, the impact of watermark embedding on the original data can be flexibly controlled, ensuring that the performance of the object recognition algorithm is not significantly affected.

[0024] This article uses specific examples to illustrate the inventive concept in detail. The description of the above embodiments is only for the purpose of helping to understand the core idea of ​​the present invention. It should be noted that any obvious modifications, equivalent substitutions or other improvements made by those skilled in the art without departing from the inventive concept should be included within the protection scope of the present invention.

Claims

1. A method for verifying data integrity and locating tampering in autonomous vehicles, characterized in that, Includes the following steps: S1: Watermark embedding: Obtain raw LiDAR point cloud data from the LiDAR sensor, and embed a preset binary watermark sequence into the raw LiDAR point cloud data based on the three-dimensional quantization index modulation method to generate watermarked LiDAR point cloud data. S2: Data transmission: Transmit watermarked LiDAR point cloud data to the advanced driver assistance system in the vehicle network; S3: Parallel Processing: Receive watermarked LiDAR point cloud data from the advanced driver assistance system and perform the following operations in parallel: SS1: Object recognition processing: Directly perform object detection, classification, or tracking processing on watermarked LiDAR point cloud data; SS2: Integrity verification processing: Extract watermark information from the received LiDAR point cloud data and compare the extracted watermark information with the expected watermark information; S4: Tampering Detection and Location: Based on the comparison results of the integrity verification process, determine whether the LiDAR point cloud data has been tampered with. If it is determined to have been tampered with, locate the tampered LiDAR point cloud data area according to the data point positions where the watermark information does not match.

2. The data integrity verification and tamper location method for an autonomous vehicle according to claim 1, characterized in that, In step S1, embedding a watermark based on the three-dimensional quantization index modulation method includes the following steps: SS1: Performs voxelization on the raw LiDAR point cloud data, dividing the point cloud into three-dimensional voxels of fixed size; SS2: For each voxel, determine its representative vertex; SS3: Based on the watermark bits to be embedded, use the corresponding 3D quantizer to quantize and modulate the 3D coordinates of the representative vertex, and move the representative vertex to one of the eight vertices in a cube with a side length of Δ / 2 centered on the original quantization position, where Δ is the quantization step size.

3. The data integrity verification and tamper location method for an autonomous vehicle according to claim 2, characterized in that, In S1, the watermark sequence is a periodically repeating sequence, and each watermark symbol corresponds to a three-dimensional bit tuple, which is used to control the displacement direction of the representative vertex in three-dimensional space.

4. The data integrity verification and tamper location method for an autonomous vehicle according to claim 3, characterized in that, In SS3 of S1, the quantization step size Δ is configured according to the target application's requirements for LiDAR data accuracy in order to balance the impact of watermark robustness and embedding distortion on the performance of the object recognition algorithm.

5. The data integrity verification and tamper location method for an autonomous vehicle according to claim 4, characterized in that, In SS2 of S3, the integrity verification process adopts a blind watermark extraction mechanism, which does not require the original LiDAR point cloud data to participate in the watermark extraction process.

6. The data integrity verification and tamper location method for an autonomous vehicle according to claim 5, characterized in that, In step S4, the method for locating the tampered LiDAR point cloud data region is as follows: identify the data point indices that do not match the watermark information, and determine the bounding box of the tampered region based on the spatial coordinates corresponding to these indices.

7. The data integrity verification and tamper location method for an autonomous vehicle according to claim 6, characterized in that, The tampering attack types detected by the method include spoofed object insertion and target object deletion.