A zero-distortion fragile watermarking scheme to detect and localize malicious modifications in textual database relations

• Main Authors: Alfagi, Abdalrahman, Abd. Manaf, Azizah, Hamida, Belal Ahmed, Olanrewaju, Rashidah Funke
• Format: Article
• Language: English
• Published: Islamabad : Little Lion Scientific 2016
• Subjects: Q Science (General)
• Online Access: http://irep.iium.edu.my/41221/7/41221_a%20zero_distortion%20fragile%20watermarking.pdf; http://irep.iium.edu.my/41221/; http://crawl.prod.proquest.com.s3.amazonaws.com/fpcache/708f5ba1f2063c036a388eb60a17112d.pdf?AWSAccessKeyId=AKIAJF7V7KNV2KKY2NUQ&Expires=1478655933&Signature=PnGGsG2wjgKP2mPApb48XRoNs3M%3D

In this paper, we present a new zero-distortion fragile watermarking scheme to detect and localize malicious modifications in textual database relations. Most existing fragile watermarking schemes introduce errors or permanent distortion into the original database content. These distortions violate the integrity of the database consequently the database quality and usability are degraded. Although, some fragile schemes are able to authenticate the database integrity and detect the malicious modifications made on the database but they are either tuples or attributes ordering based and unable to characterize the attack, identify the type of attack, identify the tampered data and locating the tampered tuples. In addition, most existing fragile schemes are based on LSB or MSB in generating the watermark unlike to this scheme, which is based on local characteristics of the relation itself such as frequencies of characters and text length in generating the watermark. This scheme is serviceable for sensitive and insensitive textual relational database since it does not introduce any error into the original contents. In addition, this scheme overcomes the weaknesses of data integrity violate, data usability and data quality degradation. The experimental results show the ability of proposed scheme in authenticating the database relation as well as the ability of characterize the attack, identifying the changed data and locating the text that affected by the malicious modification without depend on tuples and attributes ordering.