A trio of Indian researchers have proposed a method of steganography which hides messages in by using non-random distribution of letters with or without straight lines.
Steganography is a group of techniques for hiding messages in plain sight. Microdots, tiny text written inside a full stop and only legible when magnified, are one steganography technique. Steganography is hard to detect and decrypt, so much so that this paper from the US National Science and Technology Council (PDF) wrings its hands about it’s potential use by terrorists.
The new method for steganography has been outlined by Shraddha Dulera and Devesh Jinwala, both from the Department of Computer Engineering at the S V National Institute of Technology in Surat, India, and Aroop Dasgupta of Gandhinagar ‘s Bhaskaracharya Institute for Space Applications and Geo-Informatics. The trio’s paper, Experimenting with the Novel Approaches in Text Steganography, suggests that the low signal-to-noise ratio required by many current steganography techniques makes for slow decipherment.
The trio’s alternative is a system based on the characteristics of letters in the Hindu-Arabic alphabet, which they group into those possessing straight lines and those possessing curved lines. Each group is assigned a value of either zero or one as the basis for a binary code.
One method for using this scheme is to “ … generate a random string that contains the single letters (from alphabet) as the cover text. Subsequently, whenever we want to hide a ‘0’ bit in the input text file, we use the letters from the group A amongst the letters generated; whereas whenever we wish to hide a ‘1’ bit, we use the letters from the group B amongst the letters generated.”
A second scheme sees curved or straight letters capitalised at the start of sentences, so that the sentence “All birds can fly. Ostrich is a bird. Ostrich can also fly” yields a binary value of 100.
The trio’s third scheme proposes to further divide the alphabet into letters with:
- A straight horizontal middle line;
- One vertical straight line;
- A diagonal line.
By doing so, it becomes possible to create a code in which capital letters can have a binary value of 0, 1, 10 or 11.
The trio assert that “Our analysis reveals that our approaches impart increased randomness and because of randomness, these approaches are noticeable but it cannot be decoded until a user is not aware about these approaches. In addition, the proposed approaches are also immune to retyping and reformatting of text.”
But they also warn that “… one of the weaknesses of the proposed approaches is that once their applicability is known, they can easily be attacked. Hence, it is essential to keep the application of a particular approach to a particular data set secret, while using them.”