International Core Journal of Engineering 2020-26 | Page 145

C. The watermark extracted
The database is partitioned and restored by the saved
intermediate data, and the watermark position is embedded
and assigned to the watermark sequence, and the final data is
sequentially subtracted from the intermediate data. If the
number is positive, the record is 1, and if it is negative, the
record is 0. Until the extracted 8-bit binary number is
compared with the saved watermark sequence, the data block
is embedded in the watermark and recorded as 1, otherwise 0
is recorded until all data blocks are traversed.
TABLE I. M EAN VARIANCE VARIANCE BEFORE AND AFTER DATABASE
WATERMARK EMBEDDING
Through the analysis of the experimental data, it can be
seen that the mean value and variance change of the
watermark before and after the watermark embedding are
small, and have little effect on the data, and will not impair
the usability of the data. This shows that the reversible
database watermarking technology based on differential
evolution algorithm proposed in this paper has good
invisibility.
In the process of watermark extraction, according to
whether the data block is embedded in the watermark to
obtain a one-dimensional binary array, the one-dimensional
array is processed into a two-dimensional array, and the two-
dimensional array is traversed to determine the width and
height of the image. If the array value is 1, the pixel
representing the current position is White, that is, the RGB
value is 255, and vice versa, to obtain a binary image of the
copyright image. The database copyright can be judged
based on the edge information of the binary image.
In this paper, a simulation attack experiment is carried
out to test the robustness of the watermarking algorithm,
including the attack of adding subset, deleting subset and
modifying subset.
IV. E XPERIMENTAL RESULTS AND ANALYSIS
As shown in table Ⅱ, the extraction rate of watermarks
was calculated under the attack of increasing different
subsets of the database. As can be seen from the table, the
increase of subset has no effect on watermark extraction, and
the watermark information can be extracted completely.
The experimental environment is Windows 10 operating
system, Intel (R) Core (TM) i7-8550U processor, 8GB RAM.
The copyright image uses the mathematical modeling icon of
Guilin University of Technology as shown in Figure 3. The
edge image is preprocessed using the edge detection operator
for the copyright image. The binary image obtained by
extracting the edge feature points of the image is shown in
Fig. 4. The database uses currency for external transaction
data. Block height: 200*7, dimension percentage: 0.4,
watermark sequence: 1010101011, number of iterations: 100,
population: 100濁The database watermark embedding and
extraction program uses Java programming, and uses JDBC
technology to connect to the MySQL database to perform the
most basic operations on the database.
TABLE II. S UBSET ADDITION ATTACK
As shown in table Ⅲ, the extraction rate of watermark
under different modification ratio was calculated by
simulating the subset modification attack, and the results
showed that the algorithm had strong robustness under the
subset attack of different intensity.
TABLE III. S UBSET MODIFICATION ATTACK
Fig 3. Copyright image
As shown in table Ⅳ, subset deletion attacks with
different strengths are simulated, and experiments show that
the algorithm can resist deletion attacks with different
strengths.
Fig 4. Edge detection imag
By using the edge detection operator for the image, the
watermark embedding amount is greatly reduced. The
invisibility of the watermarking algorithm is improved for
the low-reliability numerical relational database. The
introduction of the differential evolution algorithm improves
the robustness of the watermarking algorithm. The ratio of
the variance of the mean value of the database before and
after the statistical watermark embedding is analyzed, and
the invisibility of the algorithm is analyzed, as shown in the
following table:
TABLE IV. S UBSET DELETION ATTACK
From the above experimental results, the algorithm is
robust.
V. C ONCLUSIONS
In this paper, we propose a reversible database
watermarking method based on differential evolution
algorithm. By extracting the edge feature points of the
123