Abstract: Data leakage is one of the biggest challenges for big data applications.Digital watermarking is an effective way for data tracking and copyright protection.However,the current digital watermarking method is mainly focus on multimedia file,such as images,audio and video files.There are little digital watermarking methods for data protection in the big data environment.Therefore,this paper proposed a plain text watermarking method based on orthogonal co-ding for big data.First,the plain text watermark is converted into a binary byte stream by coding.The orthogonal watermarking method based on row hash value and row-sequence permutation are designed.The binary watermark string is divided into segments and numbers.The watermark segment number to be embedded is calculated according to the hash value of each line of content,and the corresponding watermark segment is converted into an invisible string which is embedded to the end of line.Then,the line order is adjusted so that the hash value of each line corresponds to the binary watermark string with the flag added,which achieves the embedding of the watermark.Watermark extraction method is the inverse process of the embedding method.It can resist the destruction of watermark by operations such as replacement operation for row order in big data environment,and achieve the effect of text tampering detection by embedding fragile watermarks at the same time.Based on the proposed method,a big data watermarking system was designed and implemented.Spark was adopted to solve the problem of watermark embedding and extraction performance of massive texts,which can quickly trace the source of data leakage and improve the security of big data.Experimental and theoretical analysis prove that the proposed method has better watermark capacity performance and good concealment.At the same time,it has strong robustness since it can resist multiple content attacks and format attacks.

Key words: Big data, Digital watermark, Orthogonal, Plain text, Traceability