When: Wednesday, March 7, 2018
Where: PGH 563
Time: 11:00 AM

Experiments with Neural Networks for Small and Large-Scale Authorship Verification

Speaker: Marjan Hosseinia

We study a special case of authorship verification problem for both small and large-scale datasets. Given two documents we want to verify whether they are written by the same author. The underlying small-scale problem has two main challenges: First, the authors of the documents are unknown to us because no previous writing samples are available. Second, the two documents are short (a few hundred to a few thousand words) and may differ considerably in the genre and/or topic. Two models are proposed: a document transformation technique and a simple parallel recurrent neural network. We evaluate our methods on various types of datasets including Authorship Identification datasets of PAN competition, Amazon reviews and machine learning articles.

Bio:

Marjan Hosseinia is a third-year Ph.D. student advised by professor Arjun Mukherjee. Her research interests are Text Mining, Natural Language Processing and Machine Learning.

Solutions Using Microsoft HoloLens Augmented Reality (AR) Apps

Speaker: Mohammed Alshair

Individuals with Autism Spectrum Disorder (ASD) face many challenges in their social interactions. One of these challenges includes their difficulty to recognize the emotions of interlocutors. The struggle to understand how others are responding in conversation can cause the patient to be frustrated and confused which can lead to persistent social anxiety. Therefore, we developed an AR app that can be utilized by the users in their daily lives to assist with their social interactions. This app has demonstrated its ability to recognize the interlocutors' expressions and provide the information to the user in near real-time. Additionally, we developed a second AR app for therapists to use in the treatment of patients with social anxiety disorder. The app allows therapists to trigger their patients’ anxiety by controlling and communicating through a virtual person projected into the real environment from their own PCs. In doing this, the patient interacts with the person projected so the therapist can train the patient on how to deal with their social anxiety. Furthermore, officials, such as first responders, police, and medical providers, are overwhelmed and cannot help everyone in a timely manner during natural disasters. Consequently, there are efforts by many organizations to help prepare the public for these disasters. One of these organizations is the Community Emergency Response Team (CERT). Hence, we are developing a third AR app to be used in addition to the traditional training approach. This app will assist CERT to train the public in scenarios that normally are not possible to create in a safe environment and/or expensive to have continuous access to.

Bio:

Mohammed Alshair is a fourth-year PhD candidate at the University of Houston. His advisors are Dr. Chang Yun and Dr. Jaspal Subhlok. He finds it beneficial to use innovative technologies such as augmented and virtual realities in all sectors, for example: medicine, education, engineering, etc. And that is why his main interest is solving problems using these different realities regarding training, treatment, assisting, and education, to name a few.

Scaling and Effectiveness of Email Masquerade Attacks: Exploiting Natural Language Generation

Speaker: Shahryar Baki

Email-based attacks are a rich field with well-publicized consequences. We show how current Natural Language Generation (NLG) technology allows an attacker to generate masquerade attacks on scale, and study their effectiveness with a within-subjects study. We also gather insights on what parts of an email do users focus on and how users identify attacks in this realm, by planting signals and also by asking them for their reasoning. The insights gathered and the tools and techniques employed could help defenders in: (i) implementing new, customized anti-phishing solutions for Internet users including training next-generation email filters that go beyond vanilla spam fi lters and capable of addressing masquerade, (ii) more effectively training and upgrading the skills of email users, and (iii) understanding the dynamics of this novel attack and its ability to trick humans.

Bio:

Shahryar Baki is a third year Ph.D. student of Computer Science at University of Houston. He is working with Professor Rakesh Verma for his research. Shahryar's main research focuses on utilizing Natural Language Processing tools to improve cyber-security techniques (phishing email/website specifically).