Objectives High-definition maps are regarded as key data resources in the development of digital transportation, making their data integrity protection critically important. However, conventional hash algorithms used for file verification are not capable of inferring the location of tampered content from a single checksum anomaly. Therefore, an algorithm that can not only verify the integrity of high-definition map data but also accurately locate tampered regions is essential for mitigating network security risks in digital transportation and safeguarding public safety. To address this issue, this paper proposes a fragile watermarking algorithm specifically designed for high-definition maps in the OpenDRIVE format.

Methods The proposed algorithm is designed based on unicode zero-width characters and message digest algorithm 5 hash algorithm. The proposed algorithm uses zero-width character sequences to embed watermark information, with extensible markup language nodes serving as the basic units for watermark generation, embedding, extraction, and verification. The watermark information corresponding to each node is composed of a combination of a tree watermark and a road feature watermark. To avoid affecting the normal use of map data, the watermark sequence is embedded at the end of the line corresponding to the starting tag or the unique tag of each node. During data distribution, the administrator can use this algorithm to generate fragile watermarks,embed them into high-definition map data and verify whether the watermarks are successfully embedded in the file. When the user needs to verify the integrity of the file, this algorithm can be used to extract and verify the watermark. If the two are identical, the file is considered intact, otherwise, the algorithm determines the tampered location based on the detected anomalies.

Results Experimental results demonstrate that embedding watermarks into high-definition map documents does not cause any visible abnormal display. The watermark embedding process introduces no significant increase in file size, and the ratio of the size increment to the original file size can be controlled within 5%. In addition, the proposed algorithm exhibits high sensitivity to both node-level and watermark-level attacks and is capable of accurately locating tampered positions. In cases of whole-element deletion, the geometric information of the deleted element can still be inferred.

Conclusions The proposed algorithm is suitable for the integrity protection of high-definition map data, and can also be applied to the integrity protection of other structurally similar data such as hypertext markup language and cascading style sheets, as long as cancelling some preprocessing requirements, reconfirming the appropriate watermark embedding position and clarifying the definition of features and the sorting rules between features.