Associate Professor | Department of Computer Science
Contact (Off.): 9810154611
Email Address : veenu[dot]bhasin[at]pgdav[dot]du[dot]ac[dot]in
Information Security, Steganalysis, Feature Selection, Machine Learning, Deep Learning and its use in Healthcare
Dr. Veenu Bhasin, an alumnus of University of Delhi, has been with the Department of Computer Science, PGDAV College since July 2008. She started her career in 1996 as Software Consultant with HCL Infosystems Limited. She has been teaching various undergraduate and postgraduate courses in University of Delhi since August 2000. She is currently working as an Associate Professor.
She completed her Doctorate in Computer Science from Department of Computer Science, University of Delhi in 2018 and the Ph.D. thesis was titled “Feature Selection based Multi-class Image Steganalysis using Soft Computing Techniques”. She has published more than dozen research papers in various peer-reviewed international conferences and journals. She has presented several papers in various international conferences including IEEE SMC’2013, held at Manchester, UK.
She had been a resource person for various workshops at CPDHE and at Centre for Science Education and Communications, University of Delhi. She was the coordinator of e-lesson creation team for Discrete structures, a paper in the curriculum of B.Sc.(Hons.) Computer Science, and authored a chapters in the same; these e-lessons are available on virtual learning environment of University of Delhi.
Subjects (For sessions after 2020):
Title: Feature Selection based Multi-class Image Steganalysis using Soft Computing Techniques
Abstract: Multi-class steganalysis categorizes an image into either as non-stego or into classes which correspond to different steganography methods. The steganalysis process analyzes features calculated from images and classification decision is based on this analysis. This work was an attempt to improve the multi-class steganalysis process for JPEG images.
ELM a multi-class fast learning classifier is used as a classifier in the multi-class steganalysis process in the work, making the process very fast and useful for real-time usage. The feature set comprised of Markov features and calibrated Markov features.
Two strategies had been used in the work to circumvent the enormity of the dimensionality of the feature sets– Multiclass Steganalysis process using ensemble of ELMs and Feature Selection using Swarm Intelligence techniques.
Stochastic Diffusion Search (SDS) is adapted for two-class steganalytic feature selection. SDSFS (filter type) and FS-SDS (wrapper type) were implemented. For multiclass steganalysis, feature selection needs to base the selection of features on fitness criteria which involves multiclass aspect and thus the classification results given by ELM is used as fitness criteria. Glowworm Swarm optimization, ABC, PSO and Harmony Search have also been adapted to select optimized feature set for multi-class steganalysis. A potential solution corresponds to a feature subset. Fitness criteria computation involves F1-score and size of the feature subset.
As culmination of the work in the thesis, the design of StegTrack a novel proactive antivirus-like memory-resident steganalysis tool with GUI was proposed. StegTrack tracks the traffic of images on a system and performs multi-class steganalysis on images. StegTrack gives user the flexibility of choosing the feature extractor, feature selection & classifier. This tool introduces cleaning; stego-image is rendered unfit for extracting hidden material from it. For testing the utility and feasibility of this tool, JPEG version of tool Stego-Tracker was implemented in Java & MATLAB.
Book Chapter:
Paper published in Book Series:
List of Papers presented and published in proceedings of International conferences
List of Papers published in proceedings of International conferences
