Design and Development of a Steganographic System based on a Modern, Steganalysis Resistant LSB Algorithm on FPGA

Abstract

Steganography is the science and art of covert communication, without letting anyone know that there is any message hidden in the cover object. Practically all sorts of digital multimedia objects can be utilised as cover objects for steganography. But digital images have been the most lucrative option for cover media. In images, spatial domain algorithms are most suitable for hardware implementation due to the less complex nature of operations and ability to provide better performance. Hardware implementations of steganographic algorithms are identified as a major research gap in this work. This thesis work makes an attempt to implement a modern spatial domain steganographic algorithm, which is resistant to many recent steganalysis attacks, on FPGA based hardware. Many operations and elements in original one third probability algorithm have been found either not fully optimised or unsupported for hardware implementation. This thesis has optimised some of operations. Some of unsupported functionality has been replaced with their synthesizable equivalents. Some basic functions like major change, minor change and LSB functions have been refined for FPGA implementation. Elements like two dimensional packed arrays, floor function and intermediate floating point values have been avoided and their suitable replacements have been reinvented wherever possible.Especially, the structure of change matrix ‘C’ has been thoroughly redesigned. A system level design consisting of image storing RAMs and A FSM for embedding operations has been developed. This design based on modified algorithm has been implemented on Xilinx SP605 board (Spartan 6 series XC6SLX45T FPGA). The design is found out to be working correctly with minimum period of 9.526ns, which translates to the maximum frequency of 104.971 MHz. Resultant stego images, produced by implemented design, have been analysed for their qualitative correctness and quantitative parameters, both of which have been found at par with or marginally betterthan original algorithm.