An efficient format-independent watermarking framework for large-scale data sets

In this paper, we propose an efficient distortion-free watermarking of large-scale data sets in various formats by exploiting the power of parallel and distributed computing environment. In particular, we adapt MapReduce, Pig and Hive paradigms for the data in CSV, XML and JSON formats by identifying key computational steps involved in the sequential watermarking algorithms. Following this, we design a middleware which allows watermark generation and verification (under any computing paradigm of user's choice) of large-scale data sets (in any suitable format of user's interest) and their conversion without affecting the watermark. The experimental evaluation on large-scale benchmark data sets shows a significant reduction of watermark generation and verification times. Interestingly, in case of XML and JSON formats, Pig and Hive outperform the MapReduce paradigm, whereas MapReduce shows better performance in case of CSV format. To the best of our knowledge, this is the first proposal towards large-scale data sets watermarking, considering popular distributed computing paradigms and data formats.

Automated summary

This paper proposes an efficient distortion-free watermarking technique for large-scale datasets in various formats using parallel and distributed computing environment. Experimental evaluation shows performance depends on data format and chosen computing paradigm.