Computer Science Focus on Research - University of Houston
Skip to main content

Computer Science Focus on Research

When: Monday, April 1, 2019
Where: PGH 563
Time: 11:00 AM


Generating Coherent and Targeted Emails by Leveraging Deep Neural Learners

Avisha Das, PhD Student

Social engineering attacks like phishing, email masquerading, where a perpetrator impersonates as a legitimate entity, have always been a threat to cybersecurity researchers. However, despite having a higher probability of success, executing such an attack is costly in terms of time and manual labor. With the advancements in machine learning and natural language processing techniques, the attackers can now use more sophisticated methods to evade detection. In a proactive scenario, we presume that attackers would resort to automated methods of attack vector generation. Deep neural learners are capable of natural text generation when trained on huge amounts of written textual content by an individual. However, the application of neural text generation methods in email generation is fairly challenging owing to the presence of noise in emails and the diversity in email writing style. First, we build a simple, baseline neural model for generation and then improve upon it by adding a parallel layer for sentence selection to increase coherency. This also helps the defender reinforce their systems by automatic generation newer attacks.

Avisha Das received her Bachelor's degree in Electronics and Communication Engineering from West Bengal University of Technology, India in 2014. She is currently a Ph.D. student in Computer Science working under Dr. Verma in the ReDAS Lab. Her research interests include natural language processing, language generation, deep learning, information retrieval, and security analytics.

 

Visual Content Analysis of Lecture Videos

Mohammad Rajiur Rahman, PhD Student

VideoPoints improves access to lecture videos by creating topical segments and also extracting searchable content from these videos. It is important to represent each of the segments by a frame that contains most important visual content.  We propose to develop a summarization technique that extracts the unique visual objects (like images, diagrams, graphs etc) from a video segment and presents the most important ones in a single frame. We need to solve two problems to achieve this; the first is to  measure the importance of each visual object present in the particular segment, and the second is to organize the identified important visual objects to generate a new frame.

Mohammad Rajiur Rahman received his undergraduate degree in Computer Science and Engineering from Bangladesh University of Engineering and Technology in 2007. He is presently a PhD student and Teaching Assistant in Department of Computer Science at University of Houston. His research interests include Image Processing and Applied Machine Learning.

 

Assessing the Impact of Video Compression on Foreground Detection

Poonam Beniwal, PhD Student

Automated video surveillance is a research topic of growing interest. Video cameras are deployed in large numbers, resulting in a dramatic increase in the number of videos. Automated analysis is done on videos of different video qualities and resolution. It is important to know the reliability of these automated analysis systems in different scenarios and on different video qualities. We evaluated different algorithms on videos with different compression levels.

Poonam Beniwal is a third-year Ph.D. student working with Dr. Shishir Shah. Her research interests include Computer Vision and Deep Learning with the application to Public Safety and Security.