| CPC G06F 21/16 (2013.01) [G06N 3/0464 (2023.01); G06N 3/08 (2013.01); H04N 19/124 (2014.11); H04N 19/467 (2014.11); H04N 19/625 (2014.11)] | 7 Claims |
|
1. A neural network-based security defense method for encrypted multimedia data, comprising:
adding implicit digital watermark information to original encrypted multimedia traffic to obtain watermark encrypted multimedia traffic;
determining a first feature sequence of the watermark encrypted multimedia traffic, wherein the watermark encrypted multimedia traffic is divided into a plurality of segmented traffic according to a specified time interval, and the first feature sequence comprises a first feature vector of each of the plurality of segmented traffic; and based on the first feature sequence and a second feature sequence of the original encrypted multimedia traffic, determining a correlation coefficient between the first feature sequence and the second feature sequence, wherein the original encrypted multimedia traffic is divided into a plurality of segmented traffic according to the specified time interval, and the second feature sequence comprises a second feature vector of each of the plurality of segmented traffic; and
if the correlation coefficient indicates that a negative correlation between the first feature sequence and the second feature sequence satisfies a preset condition, performing training based on the first feature sequence and the second feature sequence to obtain a target neural network, wherein the target neural network is configured to identify whether there is implicit digital watermark information in encrypted multimedia traffic to be detected,
wherein adding the implicit digital watermark information to the original encrypted multimedia traffic to obtain the watermark encrypted multimedia traffic comprises:
randomly selecting the implicit digital watermark information from a plurality pieces of implicit digital watermark information in an implicit watermark information base, wherein the implicit watermark information base is configured to store implicit digital watermark information of different watermark types;
quantizing the selected implicit digital watermark information to obtain quantized
implicit digital watermark information;
encoding the quantized implicit digital watermark information to obtain encoded
implicit digital watermark information; and
embedding the encoded implicit digital watermark information into the original
encrypted multimedia traffic to obtain the watermark encrypted multimedia traffic corresponding to the original encrypted multimedia traffic, and
wherein embedding the encoded implicit digital watermark information into the original encrypted multimedia traffic comprises:
determining an embedding location for whole multimedia from the original encrypted multimedia traffic, and embedding the encoded implicit digital watermark information into the original encrypted multimedia traffic based on the embedding location; or
determining an embedding location for a plurality of consecutive frames from the
original encrypted multimedia traffic, and embedding the encoded implicit digital watermark information into the original encrypted multimedia traffic based on the embedding location; or
determining an embedding location for each frame from the original encrypted
multimedia traffic, and embedding the encoded implicit digital watermark information into the original encrypted multimedia traffic based on the embedding location.
|