How can I edit the bytes code of a JPG file preserving still its header and footer? (in Java)

I already have successfully coded my Steganography program in a PNG file using Java. My program works very well in both PNG and BMP files. But when I tried running my program in a JPG file, the revealed data is not the same as the original data. Certainly, the headers of each file type isn't the same. And so now I wonder; Do the data structures of PNG and JPG files aren't t开发者_如何学编程he same? I need to know exactly how to manipulate the bytes of a JPG file without affecting its header and the footer.

Thanks.

---

First of all you need to tell the exact method you are using for image steganography e.g. hiding the secret data in the lsb's of the pixels of the image, reading the file in a binary format etc.

If working with lsb's is your procedure then I hope the following answer satisfies your query-

'PNG' and 'BMP' are actually lossless file formats. After manipulating the bits of pixels of these formats when you create the new image no data is lost. This is the reason you are able to retrieve all the hidden data.

'JPG' formats however use a lossy compression technique due to which the data hidden in the pixels is lost. Even I faced this problem and the solution to this exists in handling the image in transform domain. You need to use the Direct Cosine Transform method for its implementation.

The transform domain involves manipulation of the algorithms and image transforms such as discrete cosine transformation (DCT) and wavelet transformation. These methods can hide information in more significant areas of the image and may also manipulate the properties of the image like luminance. These kinds of techniques are more effective than image domain bit-wise steganographic methods. The transform domain techniques can be applied to image of any format. Also, the conversion between lossless and lossly formats may survive.

How DCT works in steganography?

The image is broken down into 8x8 blocks of pixels. DCT is applied to each block from left to right, top to bottom. The quantization table compresses each block to scale the DCT coefficients and the message is embedded in the scaled DCT coefficients.

A lot of reasearch is still required in this method. I am working on its code and will post it ASAP.

It will be a pleasure to hear of other methods or different efficient techniques from other developers.