| Mentor: Professor Ernst L. Leiss |
| Project Title: |
Data-Centric Privacy, Various Projects (Code: EL1) |
| Mentor: |
Professor Ernst Leiss |
| Description: |
There is greatly increased concern about privacy. Some of it is driven by legal requirements (HIPAA, FERPA). In many cases, the concern is directly related to technological capabilities that threaten privacy. This overarching topic (really a family of topics) looks at some threat to privacy and explores technological means for averting this threat. This may be inference control in statistical databases, obscuring location information in cell-phone communications (something like TOR for voice), preserving privacy in data mining operations, or exploring techniques that prevent the use of certain biometric measurements for identification of individuals (e. g., face recognition, gait recognition, voice recognition, or keystroke dynamics). In each case, a student starts by researching the specific threat, then examines the technologies that are typically used to threaten privacy within the context of the scenario, and finally develops technologies that disable or impede the use of these technologies and verifies rigorously that these techniques improve the privacy of the subjects involved. |
| Research Objective: |
Enable individuals to preserve their privacy in specific circumstances. |
| Student Learning: |
Become familiar with privacy and technologies that threaten it. Explore technologies that counteract these threats. |
| Mentor: Professor Ioannis A. Kakadiaris |
| Project Title: |
Video Mining: Detecting Abandoned Luggage Items in a Surveillance Scenario (Code: IK1) |
| Mentor: |
Professor Ioannis Kakadiaris and Shishir Shah |
| Description: |
In recent years there have been a number of incidents where terrorist organizations have planted explosive devices in ordinary baggage to cause immense disruption in mass transportation networks and other areas of critical infrastructure. This project will investigate how to automatically determine which pieces of luggage have been abandoned by their owners. |
| Research Objective: |
Develop a technique for automated abandoned luggage detection and evaluate its performance on different datasets. |
| Student Learning: |
Student will become familiar with computer vision techniques for feature extraction, tracking and object detection. |
| Project Title: |
Video Mining: An Annotation tool for Surveillance Scenarios (Code: IK2) |
| Mentor: |
Professor Ioannis Kakadiaris and Shishir Shah |
| Description: |
Dataset annotation is a time consuming and expensive task to perform. This project focuses on the development of an annotation tool for surveillance scenarios, that will help annotators speed up the process of labeling for human appearance, attributes, and behavior. |
| Research Objective: |
Develop a tool that will speed up the annotation process for both single images, as well as video sequences, by combining manual and automated capabilities for detecting, tracking and labeling pedestrians and their attributes, as well as detecting and labeling events. |
| Student Learning: |
Student will become familiar with standard computer vision techniques for feature extraction, as well as, person detection and tracking. |
| Project Title: |
Towards real-time soft biometric prediction in videos (Code: IK3) |
| Mentor: |
Professor Ioannis Kakadiaris and Shishir Shah |
| Research Objective: |
Implement a human detection and tracking algorithm that will also add in the video frames in real time the soft biometric attributes such as gender, height, age etc. |
| Student Learning: |
Student will learn how to track humans in videos and depict their soft biometrics on top of the bounding box in real time. The prediction part of the soft biometrics will be done with existing deep learning techniques. |
| Mentor: Professor Ioannis Pavlidis |
| Project Title: |
Bio and Behavioral Feedback (B2-Feedback) in Transitional Vehicle Technology (Code: IP1) |
| Mentor: |
Professor Ioannis Pavlidis |
| Description: |
There have been tremendous developments in the area of automobile assistive systems, which aim to make driving safer. Currently, such systems range from auto-braking to auto-parking. In this research project, we aim to develop the next generation of assistive systems that will quantify the stress and performance levels of drivers, determining if their apparent and hidden states are within safe margins. If these states found to be unsafe, then the system will notify the drivers, who will have to correct their behaviors (e.g., by disengaging from a texting distraction), in order for the complex system to return to homeostasis. This research draws on tera-bytes of simulator and field experimental data that the UH Computational Physiology Lab collects in collaboration with the Texas A&M Transportation Institute under separate federal funding (http://cpl.uh.edu/projects/stress-studies/dds/). |
| Research Objective: |
In an increasingly computer-assisted driving world, this research aims to bring harmony between the states of the computerized car and the driver, which are now merging into a complex cyber-physical-human system. |
| Student Learning: |
Students will engage in sophisticated multi-modal data analytics and biofeedback algorithm development, drawing on the massive and unique UH-A&M experimental dataset. |
| Project Title: |
Hidden Patterns and Drivers of Academic Scholarship (Code: IP2) |
| Mentor: |
Professor Ioannis Pavlidis |
| Description: |
The UH Computational Physiology Lab in collaboration with Northwestern’s Kellogg School of Management have been amassing a comprehensive biographic, bibliographic, and funding dataset for academic researchers in the United States. The dataset is publicly visualized via the Scholar Plot interface (http://www.scholarplot.com). This dataset provides unique capabilities to computational researchers in the science of science area. |
| Research Objective: |
Investigate patterns of cross-disciplinary collaboration and hidden drivers of academic success that may inform the science policy of the future. |
| Student Learning: |
Students will engage in social network data analytics and multiple linear regressions. |
| Mentor: Professor Rakesh Verma |
| Project Title: |
Classification Techniques for Unbalanced Datasets (Code: RV1) |
| Mentor: |
Professor Rakesh Verma |
| Description: |
Many real-world applications must deal with unbalanced datasets. For example, a large majority of the websites on Internet are benign, but there is a significant minority of malicious websites. Many automatic classification techniques cannot handle unbalanced datasets well. Even traditional metrics such as recall, precision and F1-measure are not suitable for measuring the performance of classifiers on unbalanced datasets. |
| Research Objective: |
To design classification techniques that can work well for unbalanced datasets. |
| Student Learning: |
Data mining, machine learning, cost-sensitive learning, and performance analysis. |
| Project Title: |
Adaptive Malicious Email Detection (Code: RV2) |
| Mentor: |
Professor Rakesh Verma |
| Description: |
The Internet society and economy are currently plagued by many different kinds of attacks. One serious problem is emails containing malware or social engineering attacks. There has been considerable research in designing automatic methods, but the vast majority of them are oblivious to the user profile. |
| Research Objective: |
To design adaptive automatic classification techniques based on machine learning and natural language processing techniques. |
| Student Learning: |
Machine learning, natural language processing, human-computer interaction and the email ecosystem. |
| Mentor: Professor Stephen Huang |
| Project Title: |
A Study of Cyber Behavior of Intruders (Code: SH1) |
| Mentor: |
Professor Stephen Huang |
| Description: |
By cyber behavior, we mean the intruder’s interaction with the computer and the indirect effects caused by the interaction. These behaviors/interactions can be between human (hacker) and system (target machine), between two computer systems, or between two humans. We have observed that user activities in the cyber space always leaves some information that can be mined and used for identification. Hackers sometimes try to hide their presence but their evasion techniques may lead to other strategies for intruder detection. |
| Research Objective: |
To better understand and model intruders’ cyber behavior for the purpose of preventing security breaches. |
| Student Learning: |
In this project, a student will learn computer security, cyber behavior, and modeling of behavior. Students will also be involved in design experiments and collect data to validate the model. |
| Project Title: |
Real-Time Intrusion Detection Monitoring System to Prevent Data Breach (Code: SH2) |
| Mentor: |
Professor Stephen Huang |
| Description: |
In order to avoid being detected, computer hackers typically go through a long chain of computers to break into a target machine. This can be achieved by using a chain of stepping-stones hosts or through Tor. We are interested in real-time algorithms that can detect such intruders effectively. The project involves the integration of algorithm design, network protocol, and computer security techniques into a system. |
| Research Objective: |
To better understand and detect intruders into computer systems in real-time. |
| Student Learning: |
In this project, a student will learn computer security, stepping-stone detection, network protocols, and various algorithms. Students will also be involved in design experiments and collect network packets for analysis. |
| Mentor: Professor Nouhad Rizk |
| Project Title: |
Using Data Mining to Identify Academic Factors Which Influence Student Retention (Code: NR1) |
| Mentor: |
Professor Nouhad Rizk |
| Description: |
STEM fields are some of the most challenging fields of study, and within these fields, it is not uncommon for a quarter or even a half of the class to drop out of certain classes and retake them a second semester. The purpose of this project is seeking to build an early warning system to identify students who are at risk of failing a course. This project centers on the use of data mining techniques to identify factors which can be utilized by this early warning system to identify such at-risk students. The process starts by collecting data on various academic factors and analyze the collected data for correlations with grades. Next step will be utilizing a data analytic approach such as Decision Tree, Random Forest, SVM, and Deep Neural Networks to construct a model predicting the student’s success in a course. The last step is analyzing the model to identify influential factors and to determine its effectiveness at predicting student outcomes. |
| Research Objective: |
Improve student retention and help students graduate on time. |
| Student Learning: |
Become familiar with data mining techniques specifically prediction methods. |
| Project Title: |
Crime Data Analytics (Code: NR2) |
| Mentor: |
Professor Nouhad Rizk |
| Description: |
The research explores many factors that motivates offenses, since in every era the way of committing a crime has drastically changed due to widespread technological evolution. Studies of this sort helps professionals in a lot of educational and governmental fields like schools or law societies to identify which factor contributes the most in encouraging an individual to commit a misdeed. By detecting which factor(s) contribute the most to criminal behavior, certain law enforcing agencies can back track and re-evaluate where the government has gone wrong or where we have failed as a society. Some preliminary research shows that social factors like child abuse, anti-social behavior or acceptance among peers or economic reasons like poverty, huge family responsibilities, status hierarchy or perhaps political reasons like racial supremacy or partisanship gives people the confidence that they can get away from laws could contribute to delinquent behavior. Data science strategies in python like linear regression or Principal Component Analysis will be used to analyze the crime rate by designing a predictor model based on criminal factors discovered though out the study. |
| Research Objective: |
Build a model highlights the important factors contributing to the overall behavior of a criminal and what social, economic or perhaps political behavior motivates them to commit something illegal through Data Mining techniques. |
| Student Learning: |
Learn how to make an effective predictor model by exploring Data Mining techniques like Decision Forests, Clustering, Regression, and Classification algorithms. |