Text steganography, the art of concealing a secret text inside another innocuous text called the cover, is usually performed by insertion of whitespace, punctuation marks, misspelling words, or by arbitrarily capitalizing words or inserting synonyms, changing font-sizes & colors, etc. All of these have the disadvantage that they either arouse suspicion or are easily noticeable; and even lost if manually copied, i.e. handwritten. Furthermore, they are easily detectable by automated checkers. Still there are other methods which require a stego-key in order to decrypt the message. In covert intelligence operations, transmission of the stego-key may not be possible at all, more so when the message is urgent. Digital communications and Internet connectivity may also be lacking in certain situations, and the only mode of message passing available may be the exchange of handwritten text on paper; which effectively rules out text modifications like font-changes, whitespace insertion, etc. or any form of digital steganography like image/audio steganography. Finally, in almost all text-steganographic techniques, there is no provision to for the receiver to detect whether or not there is indeed any message embedded. This is very important in intelligence operations where a number of decoy text need to be sent with only one concealing the actual message. In this paper, we propose a new tool called STEGASSIST that can help the sender in generating the stego-text manually. It is a low-tech form of steganography that is especially suited to covert operations like espionage or under-cover journalism. In this method, the generated cover and the stego-text are identical, or in other words, there is no cover-text. Moreover, decryption does not require a stego-key, and the stego-text may be printed or even hand-written and sent via unreliable messengers, or published, without arousing any suspicion. Finally, the received stego-text can be checked by the receiver to detect whether or not there is any actual message embedded in it.

Clustering is the technique of finding useful patterns in a dataset by effectively grouping similar data items. It is an intense research area with many algorithms currently available, but practically most algorithms do not deal very efficiently with noise. Most real-world data are prone to containing noise due to many factors, and most algorithms, even those which claim to deal with noise, are able to detect only large deviations as noise. In this paper, we present a data-clustering method named SIDNAC, which can efficiently detect clusters of arbitrary shapes, and is almost immune to noise – a much desired feature in clustering applications. Another important feature of this algorithm is that it does not require apriori knowledge of the number of clusters – something which is seldom available.

This paper introduces a simple and fast lossless compression algorithm, called CAD, for the compression of protein sequences. The proposed algorithm is specially suited for compressing proteomes, which are the collection of all proteins expressed by an organism. Maintaining a changing dictionary of actively used amino-acid residues, the algorithm uses the adaptive dictionary together with Huffman coding to achieve an average compression rate of 3.25 bits per symbol, better than most other existing protein-compression and general-purpose compression algorithms known to us. With an average compression ratio of 2.46:1 and an average compression rate of 1.32M residues/sec, our algorithm outperforms every other compression algorithm for compressing protein sequences in terms of the balance in compression-time and compression rate.

The subset sum problem is to decide whether for a given set of integers A and an integer S, a possible subset of A exists such that the sum of its elements is equal to S. The problem of determining whether such a subset exists is NP-complete; which is the basis for cryptosystems of knapsack type. In this paper a fast heuristic algorithm is proposed for solving subset sum problems in pseudo-polynomial time. Extensive computational evidence suggests that the algorithm almost always finds a solution to the problem when one exists. The runtime performance of the algorithm is also analyzed.