Sonal Ayyappan

Framework for design of Graybot in social network

The increased use of Internet in all areas of life has resulted in different types of Internet threats that has transformed from attacks which disable infrastructure to those which affect people and organizations. Botnet is a major internet threat and the command and control architecture of botnet differentiates it from other types of malwares. Botnet is a collection of compromised computers in a network under the command and control of attacker called botmaster. This paper proposes a botnet that uses Twitter for its command and control. The botmaster uses the fake account to post tweets which includes the encrypted commands. The infected machines called botclients obtains these tweets, decrypts the encrypted command in the tweet and launch attacks in the infected machines. The key used for encrypting the communication between the botserver and botclients is obtained using Diffie Hellman Key Exchange. The functionality and implementation of proposed botnet helps to understand about how social networking sites can be used for botnet command and control and about the destructive nature of botnet attacks.

Off-line handwritten character recognition using Hidden Markov Model.

In this paper, we are presenting a method for the recognition of Malayalam handwritten vowels using Hidden Markov Model (HMM). OCR is a method to detect characters in different sources. The goal of OCR is to classify optical patterns in an image to the corresponding characters. Recognition of handwritten Malayalam vowels is proposed in this paper. Images of the characters written by eighteen subjects are used for this experiment. Training and recognition are performed using Hidden Markov Model Toolkit. Recognition process involves several steps including image acquisition, dataset preparation, pre-processing, feature extraction, training and recognition. An average accuracy of about 81.38% has been obtained.

Investigation of effectiveness of ensemble features for visual lip reading.

Features used for classification play essential role in the performance of system. In the field of Lip reading, features appear in large number which has to be solved by selection of subset of features. Work covered in this paper validates the performance of individual visual features such as lip height, lip width, area of lip region, angles at corners and then combine them to create a new subset feature that improves the classification accuracy of certain weak features when combined with significant attributes. Each feature provides different level of representing classification characteristics for words. Area feature provided highest independent accuracy of 75.70%. Ensemble feature area-h4 produced highest combined accuracy of 71.25%.

A novel technique for LiDAR data segmentation and three-dimensional space projection

LiDAR (Light Detection And Sensing) data is a huge 3-dimensional data which is difficult to explore information from the data. It is necessary to segment this LiDAR data. In this paper, segmentation is done using quad-tree structure. In each segment, a point cloud is obtained. Subsequently we create a Triangulated Irregular Network (TIN) for each point cloud. Later TIN is merged to obtain structural lines for a given LiDAR data. The initial result helps us understand and reveal relevant information about the objects/shapes from the raw benchmark LiDAR dataset, i.e., ALS data of Toronto.

A reversible crypto-watermarking system for secure medical image transmission

Health care institution demands exchange of medical images of number of patients to sought opinions from different experts. In order to reduce storage and secure transmission of the medical images, the paper presents a reversible crypto-watermarking technique. The system is based on stream cipher encryption of the image using a secret key and the secret key ciphering is done by public-private key crypto system-ELGamal. The secret key which is encrypted using ELGamal and is watermarked in Discrete Cosine Transform (DCT) domain using the spread spectrum coding algorithm. The Digital Watermarking is the process of embedding data to multimedia content. This can be done in spatial as well as frequency domain of the cover image to be transmitted. The robustness against attacks is ensured while embedding the encrypted secret key in transform domain. But if capacity is the major concern then spatial pixel manipulations based on prediction errors can be effective media for transferring the data. The experimental results reveals and analyzes the reversible crypto-watermarking method in terms of PSNR value, entropy requirements and histogram analysis.

Image watermarking scheme based on MD5 and simplified DES

Digital watermarking is one of the prominent method to prevent illegal copying and duplication of digital media. It is the process of hiding digital data in any form of multimedia data such as image, audio, video, etc. This paper proposes a new block based watermarking technique in transform domain which uses cryptographic methods for embedding the watermark This method uses the MD5 hash of the cover image and the key generated from Simplified-DES to find out the bits to which the watermark is to be embedded. The hash of the cover image is encrypted using SDES and watermark embedding bits are selected. Experimental results shows that the method is robust against different geometrical attacks.