Information Hiding Used In Database And Study On Steganalysis Based On Ensemble Learning

Since its emergency in the 1990s, information hiding has aroused great concern to the researchers in the field of information security. Information hiding is a technique that can embed specific information (such as copyright, secret message, etc.) into image, audio, video or text files, and it can be used for copyright protection of digital media or covert communication. This paper mainly discusses two issues, the application of information hiding in relational database and steganalysis.Most watermarking relational database algorithms need to modify the attribute value of relational database a little bit in order to embed watermarking, and this will undermine the authenticity and validity of the data. By using information hiding, our watermarking method does not need to modify the data of the database, while it will modify the cover of the information hiding. So this method can avoid the damage to the source data. Our algorithm not only applies to the relational database with numeric attributes, but also applies to the relational database with attributes of non-number type. Experiments show that our technique is resilient to tuple deletion, alteration and insertion attacks.Information hiding can be used both for legal and illegal purpose, so the research on steganalysis is extremely necessary. This paper uses ensemble learning, which will combine the weak classifiers to a strong classifier, to improve the accuracy of steganalysis. At present, steganalysis can detect precisely when the cover media carries much secret information, however, the performance is not good enough when the cover media carries only little secret information. So by using ensemble learning technique, we can improve the performance. Experiments show that the effect of ensemble based on fisher classifier is not very obvious, and the performance of ensemble based on neural network classifier improves markedly.At the end, the paper discusses some research topics and directions about watermarking relational database and steganalysis based on ensemble learning in the future.