Dynamic prediction based watermarking for temporal relational databases

Temporal databases maintain time-varying information. So the probability of database tampering is also increased. The need of security and proper control system are obvious, to find intruders who do the data modification, the modified data and recovery of original data. This work addresses the problem of database authentication for forensic analysis of temporal databases. Existing approaches used for database tampering detection and recovery have limited success due to the complexity of database authentication and analysis. This paper proposes a novel approach to tamper detection, analysis and recovering original data back by adding a set of bits to the original data that acts as a watermark. The proposed system provides a strong validation and recovery scheme to maintain data security and integrity. Various attacks like insertion, deletion and modification are then performed to check the robustness of the watermark. The performance of the proposed watermarking method is analyzed through Mean Square Error (MSE) and normalized correlation (NC). Extensive analysis shows that the proposed method is robust against various forms of database attacks, including insertion, deletion and modification. From the experimentation, it shows that the proposed scheme is well suited for content protection (numerical and non-numerical) in temporal database watermarking.