Summer Intern Projects

CIRI and the Information Trust Initiative had seven interns this summer conducting fantastic research. Read more about their projects below.

Using Natural Language Processing for Analyzing Disaster Management

Samira Zad, Florida International University

Abstract:

Providing timely and effective response to disasters is a challenge for emergency responders and related agencies. The complex and non-routine nature of disaster management requires a substantial amount of planning, coordination and mobilization of resources among a network of organizations with different functions. Humanitarian assistance and disaster relief (HADR) efforts can benefit from computational methods that aim to support the coordination of tasks between relief agents more rapidly and efficiently, and the collection and analysis of accurate and reliable information. We first developed a conceptual and operational definition of what a good response is. We then brought this work into an application context, namely the 2010 Haiti earthquake relief operations by the United States Coast Guard’s (USCG) and related organizations. We conducted semi-structured interviews with the captains of the three first USCG cutters on-site at Haiti to construct a ground truth for the timeline and steps of the Haiti response, along with measures of a response’s success. We then collected data from multiple sources (social media, news media, blogs, scientific literature, and governmental/public agency data), and applied computational methods from text mining and natural language processing to these data to understand differences in the perception of an event depending on the data type and analysis methods. For each source and stage of the HADR process, we identify the situational awareness, and compare the results per source type and against the ground truth. This work helps to identify which sources can be combined to obtain a comprehensive and reliable picture of a situation, and which would be redundant; allowing for more scalable and efficient analytical support for HADR efforts.

Anomaly Detection for Raven II Surgical Robot

Uchenna Ezeobi, Howard University

Abstract:

In this project, we demonstrate targeted attacks on teleoperated surgical robots, Raven II. These attacks exploit vulnerabilities in the software of the robot. We implemented a fault injection engine that inserts faults into different parts of the robot software at runtime. We present an approach that uses the z-score algorithm to detect this attack in real time by monitoring various parameters of the robot. We further compare this approach with the dynamic model-based detection. Our current experiment shows that the z-score algorithm can detect malicious attack before it manifests in the simulated RAVEN II robot.

Community Threat Analysis on Twitter: Combining Sentiment, Agreement, and Size

Johnny Carter Jr., Howard University

Abstract:

Twitter is used for broadcasting information. It allows someone to share information about themselves and their surroundings. By having a large volume of information posted on social media offer a new approach into the physical world through a social network. Methods: By use Clustering featured with sentiment analysis, we should be able to identify large social groups who sentiment exhibit. A positive or negative views on issues that conflict with national security. By extracting and labeling the text from tweets using a scientifically accepted with clustering along with sentiment database. Per tweet, the classifier from the set that shows the greatest confidence.

Findings and Conclusion:

Sentiment analysis in combination with Clustering shows improvement in result.

Castle Keep - An Interactive Demonstration of Network Security Concepts

Alexander Monaco, Florida International University

Abstract:

The main goal of this project was to develop a testable improvised network representation that would demonstrate network security concepts and provide a framework for educational experimentation. After doing research on various examples of recognizable models that could be used as an alternative visualization to a network topology, the program, “Castle Keep”, was developed. Castle Keep is a tower defense game that is meant to introduce, teach, and reinforce security concepts through gameplay. Within the game, users can visualize and identify causation through visual cues, learn from experience and perform tasks to protect their Castle and the Kingdom surrounding it, similar to setting up firewall rules and intrusion detection software to protect their critical hosts.

Biometric Security: The Vulnerability of Facial Recognition to Face Morphing

Cynthia Jules, Howard University

Abstract:

Computers and smartphones have become an increasingly vital part of our lives as they have taken the place of traditional security methods for storing sensitive information. As people become more reliant on this technology and the risk of cyber crimes increase, we have to find new ways to protect our devices from being compromised. Biometric data has been used as a relatively secure method of verifying an individual's identity. Facial recognition (FR) has become a tool employed by many companies to secure our devices, and to protect against presentation attacks they have begun to build more robust systems. However, they have neglected to design systems impervious to morphed facial images. Morphed face images are artificially generated images, which blend the facial images of two or more different subjects into one. This allows two people to use one ID to gain access to passport and banking systems or allows an attacker to present the morphed image, in conjunction with other spoofing methods, to gain access to facial recognition systems. This research evaluates an experiment on the vulnerabilities of facial recognition systems to morphed images, as well as, determines what the minimum threshold of morphing and editing is needed to gain access into the system.

A Conceptual Resilience Framework for Transportation Infrastructure

Jorge Navarrete, University of Texas, El Paso

Abstract:

The increase in hydrometeorological events has amplified severity of natural hazards such as: floods, droughts, storms, among others. These phenomenon changes are distressing the transportation assets around the globe. To mitigate the influence of hydrometeorological events there is a nascent emphasis on enhancing resiliency of critical transportation infrastructure. The purpose of this study is to propose a conceptual framework that specifically focuses in the technical dimension out of the four resiliency dimensions mainly technical, organizational, social and economic. This framework can be ameliorated in future research to quantitatively assess resiliency in targeted transportation assets based on indicators and stakeholders input. To elaborate in resiliency assessments, a methodology to quantify the level of risk in terms of “low, medium, high or very high” is also performed. A case study of a bridge shows the applicability of this methodology to quantify the level of risk due to extreme hydrometeorological events by calculating the probability of occurrence and severity. Additionally, a mitigating plan is summarized based on the Federal Highway Administration (FHWA) and National Oceanic and Atmospheric Administration (NOAA) guidance.

Measuring Illinois Pavement's Performance in Response to Hydrometeorological Changes

Angel Rodarte, University of Texas, El Paso

Abstract:

In recent years, the transportation infrastructure assets have been continuously badgered by the hydrometeorological events leading to unexpected maintenance and rehabilitation. For instance, the road’s ability to withstand stressors is degraded with the increase in temperature and precipitation. To evaluate the influence of hydrometeorological events, six pavement designs from cities around the state of Illinois were analyzed for 20-years using Pavement ME Design Software and future climate data from North American Regional Climate Change Assessment Program (NARCCAP). The performance results were analyzed in terms of International Roughness Index (IRI) and asphalt concrete (AC) rutting. The performance data identified that there is an increase in IRI while influence on AC rutting was minimal. With IRI of 100 in./mile triggering maintenance, some pavements will require maintenance as early as one and a half years. Therefore, the impact of hydrometeorological event needs to be accounted for while designing pavements to enhance resilience.