Abstract: Aiming at the demand of data leakage traceability in the process of high-precision measurement data sharing in the power grid data middle platform, the traditional LSB database watermarking method is difficult to maintain data accuracy and watermark robustness at the same time. To solve this problem, this paper proposes a collaborative optimization LSB watermarking method with precision constraints for power grid data leakage traceability. Firstly, the embeddable bit layer is dynamically determined according to the allowable error of the service, and the watermark embedding distortion is controlled under the premise of meeting the service accuracy constraint. Secondly, in order to solve the problem of reduction of embedding capacity caused by accuracy constraints, the pseudo-random decentralized embedding and BCH error correction coding mechanism are combined to improve the watermark anti-deletion, anti-rearrangement and anti-format conversion capabilities. Finally, an auxiliary traceability mechanism based on virtual primary key is introduced to improve the positioning stability in complex data flow environment. Experiments conducted in the real environment of the power grid data middle platform show that, compared with the conventional LSB watermarking method, the proposed approach reduces the root mean square error change rate to 0.15%, maintains strong robustness across various attack scenarios, and achieves a minimum watermark bit error rate of 5.9%. The traceability accuracy reaches 100%, demonstrating good engineering applicability for power grid data sharing scenarios.

Key words: power grid data middle platform, database watermarking, LSB optimization, leakage traceability, precision constraint, robustness