Research On Image Data Hiding Methods In Network Environment

Reversible data hiding is widely used in the feld of image privacy protection due to its nature of lossless recovery of carrier images.However,the existing research work cannot meet the users’ demands for protecting images according to diferent privacy levels when facing real scenarios in social platforms.In addition,most social platforms compress user images for transmission,resulting in existing privacy protection methods often fail to recover the compressed protected images.To solve the above problems,a hierarchical privacy protection method that can recover images by region and an anti-JPEG compression privacy protection method that can resist real channel compression attacks are proposed in a targeted manner.In order to clarify the current status of the research on reversible data hiding,the existing typical algorithms and their derivatives are reviewed,and the performance advantages and disadvantages of each algorithm in DCT domain embedding are analyzed by combining the embedding capacity and embedding distortion,and it is pointed out that the diference expansion method is suitable for carriers with small values while the histogram translation is suitable for carriers with concentrated values.By analyzing the whole process of JPEG compression,it is pointed out that the medium frequency region in the DCT block is more suitable for embedding message than the high frequency region,and the fle increment introduced after message embedding is smaller.In order to achieve the purpose of image hierarchical protection for social platform users,an image hierarchical privacy protection method based on reversible data hiding is proposed.First,the specifed image is partitioned and assigned a key according to the set privacy protection policy;the DCT block is divided into three regions: low-frequency,mid-frequency and highfrequency,and the auxiliary information generated by adjusting the low-frequency coeffcients is encrypted with a key and embedded in the mid-frequency coeffcients,and the mid-frequency coeffcients are dislocated by the key;fnally,the location information is embedded in the image using histogram translation.Experiments with two face datasets,fam2 a and fam4 a,show that the proposed method can achieve both sub-regional recovery and lossless recovery of the original image,and the introduction of the key further enhances the security.In addition,the proposed method reduces the fle increment with similar privacy preserving efect as the previous method,which can signifcantly save storage space in hierarchical privacy preserving scenarios.Since social platforms compress user images,a privacy-preserving method against JPEG compression is proposed for smooth recovery of protected images after compression.First,the image is simulated to be compressed and the DC coeffcients of the compressed image are adjusted,and the auxiliary information is embedded in the AC coeffcients using an optimized embedding algorithm so that the message can be extracted without errors within a certain distortion range for the encrypted image.Then,the coeffcients of the original image before compression are adjusted according to the processed compressed image,and fnally an intermediate image with distortion within a certain range after compression is generated.The validation under Fam2 a and Fam4 a data sets shows that the generated intermediate image can be recovered to the compressed version of the original image without distortion after channel compression.At the same time,the security and privacy preserving efects of the proposed method are at the leading level in both lossy and lossless channel transmission.