DISSERTATION DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Jun Zhou Advisor: Jijun Tang Date : April 6th Time: 2:40 – 4:00 pm Place : 3A75, Swearingen Abstract Over millions of year of evolutionary history, the order and content of the genomes got changed by rearrangements, duplications and losses. There is always a consistent passion to find out what happened and what can happen in the evolutionary process. Due to the great development of various technology, the information about genomes is exponentially increasing, which make it possible figure the problem out. The problem has been shown so interesting that a great number of algorithms have been developed rigorously over the past decades in attempts to tackle these problems following different kind of principles. However, difficulties and limits in performance and capacity, and also low consistency largely prevent us from confidently statement that the problem is solved. To know the detailed evolutionary history, we need to infer the phylogeny of the evolutionary history (Big Phylogeny Problem) and also infer the internal nodes information (Small Phylogeny Problem). The work presented in this thesis focuses on assessing methods designed for attacking Small Phylogeny Problem and algorithms and models design for genome evolution history inference from FISH data for cancer data. During the recent decades, a number of evolutionary models and related algorithms have been designed to infer ancestral genome sequences or gene orders. Due to the difficulty of knowing the true scenario of the ancestral genomes, there must be some tools used to test the robustness of the adjacencies found by various methods. When it comes to methods for Big Phylogeny Problem, to test the confidence rate of the inferred branches, previous work has tested bootstrapping, jackknifing, and isolating and found them good resampling tools to corresponding phylogenetic inference methods. However, till now there is still no system work done to try and tackle this problem for small phylogeny. We tested the earlier resampling schemes and a new method inversion on different ancestral genome reconstruction methods and showed different resampling methods are appropriate for their corresponding methods. Cancer is famous for its heterogeneity, which is developed by an evolutionary process driven by mutations in tumor cells. Rapid, simultaneous linear and branching evolution has been observed and analyzed by earlier research. Such process can be modeled by a phylogenetic tree using different methods. Previous phylogenetic research used various kinds of dataset, such as FISH data, genome sequence, and gene order. FISH data is quite clean for the reason that it comes from single cells and shown to be enough to infer evolutionary process for cancer development. RSMT was shown to be a good model for phylogenetic analysis by using FISH cell count pattern data, but it need efficient heuristics because it is a NP-hard problem. To attack this problem, we proposed an iterative approach to approximate solutions to the steiner tree in the small phylogeny tree. It is shown to give better results comparing to earlier method on both real and simulation data. In this thesis, we continued the investigation on designing new method to better approximate evolutionary process of tumor and applying our method to other kinds of data such as information using high-throughput technology. Our thesis work can be divided into two parts. First, we designed new algorithms which can give the same parsimony tree as exact method in most situation and modified it to be a general phylogeny building tool. Second, we applied our methods to different kinds data such as copy number variation information inferred form next generation sequencing technology and predict key changes during evolution.

Thesis Defense Author : Ahmed Al-Qari Advisor : Dr. John Rose Abstract Proteomics has made major progress in recent years after the sequencing of the genomes of a substantial number of organisms. A typical method for identifying peptides uses a database of peptides identified using tandem mass spectrometry (MS/MS). The profile of accurate mass and elution time (AMT) for peptides that need to be identified will be compared with this database. Restricting the search to those peptides detectable by MS will reduce processing time and more importantly increase accuracy. In addition, there are significant impacts for clinical studies. Proteotypic peptides are those peptides in a protein sequence that are most likely to be confidently observed by current MS-based proteomics methods. There has been rapid improvement in the prediction of proteotypic peptides for AMT studies based on amino acid properties such as amino acid content, polarity, charge and hydrophobicity using a support vector machine (SVM) classification approach. Our goal is to improve proteotypic peptide prediction. We describe the development of a classifier that considers amino acid usage that has achieved a classification sensitivity of 90% and specificity 81% on the Yersinia pestis proteome (using 3-AAU). Using Ordered Amino Acid Usage (AAU) feature, we were able to identify a different set of peptides that was not identified by the 35 peptides features that STEP (Webb-Robertson, 2010)[2] have used. This means that Ordered Amino Acid Usage (AAU) feature could complement other features used by STEP to improve identification accuracy. Building on this success, we used STEP (Webb-Robertson, 2010)[2] 35 amino acids features to complement Ordered Amino Acid Usage (AAU) feature in order to enhance the overall accuracy.

THESIS DEFENSE Author : Shannon Hood Advisor : Dr. Ioannis Rekleitis Abstract This paper proposes a solution to the problem of cooperative exploration using an Unmanned Ground Vehicle (UGV) and an Unmanned Aerial Vehicle (UAV). More specifically, the UGV navigates through the free space, and the UAV provides enhanced situational awareness via its higher vantage point. The motivating application is search and rescue in a damaged building. A camera atop the UGV is used to track a fiducial tag on the underside of the UAV, allowing the UAV to maintain a fixed pose relative to the UGV. Furthermore, the UAV uses its front facing camera to provide a birds-eye-view to the remote operator, allowing for observation beyond obstacles that obscure the UGV's sensors. The proposed approach has been tested using a TurtleBot 2 equipped with a Hokuyo laser ranger finder and a Parrot Bebop 2. Experimental results demonstrate the feasibility of this approach. This work is based on several open source packages and the generated code will be available online.

Thesis Defense Author : Nicholas Weidner Advisor : Dr. Ioannis Rekleitis Abstract The proposed work presents a systematic approach for 3-D mapping and reconstruction of underwater caves. Exploration of underwater caves is very important for furthering our understanding of hydrogeology, managing efficiently water resources, and advancing our knowledge in marine archaeology. Underwater cave exploration by human divers however, is a tedious, labor intensive, extremely dangerous operation, and requires highly skilled people. As such, it is an excellent fit for robotic technology, which has never before been addressed. The proposed solution employs a stereo camera and a video-light. The approach utilizes the intersection of the cone of video-light with the cave boundaries resulting in the construction of a wire frame outline of the cave. Successive frames produce a scalable accurate point cloud which, through the use of adapted 3-D geometry reconstruction techniques, creates a fully replicated model of the cave system.

People know when they need to hire a dentist or an accountant, but they rarely know when they need to hire a linguist. This talk focuses on the professional opportunities available to people with traditional linguistics and computational linguistics training, from undergraduate to graduate degrees. Laura Walsh Dickey shares her experience transitioning from academia to industry. As part of the talk, she discusses specific problems she’s worked on and the kinds of interesting challenges that linguists might find themselves working on in industry. She talks about how to spot jobs that might be appropriate for linguists and gives practical tips about finding them, applying for them, and deciding what’s right for you. Laura Walsh Dickey is a Program Manager at Google, focusing on machine learning and language technology. She joined Google in 2013 with a PhD in Linguistics from the University of Massachusetts, Amherst and 25 years of experience in academia, consulting, and industry. Her research at the Max Planck Institute for Psycholinguistics and Northwestern University focused on the phonology of liquid consonants, speech perception, and speech production. Her forays into the consulting world opened up a new area of linguistic problems which needed to be solved, from drug name confusability to teaching foreign language pronunciation to understanding what people mean when they type in that Google search box.

IIT Faculty Candidate Seminar Sponsored by Department of Integrated Information Technology Yoris Au Department of Information Systems and Cyber Security College of Business The University of Texas at San Antonio Abstract: This study investigates the effect of observational learning in the crowdsourcing market as an attempt to identify appropriate mechanism(s) for sustaining this increasingly popular business model. Observational learning occurs when crowdsourcing participating agents obtain knowledge from signals they observe in the marketplace and incorporate such knowledge into subsequent actions to improve their participation outcomes. This form of learning is examined in the context of the two-sided crowdsourcing platform in which participating customers’ and professionals’ decisions interact with and influence each other. Two structural models are constructed to capture customer and professional’s probability of success in the presence of various constantly changing market signals. A third model is developed to capture factors that influence market outcomes such as level of participation by professionals. These models will be estimated using the Bayesian approach on a longitudinal dataset that consists of seven years of transaction data in four product categories from a leading crowdsourcing site. We expect to observe learning effect in this crowdsourcing market and to identify various factors that influence the probability of a professional (agent) submitting a bid to a crowdsourcing project and the probability of a customer (principal) selecting a winner through observational learning.

IIT Faculty Candidate Seminar Sponsored by Department of Integrated Information Technology Martin Stytz Chief Research & Technology Officer Calculated Insight Abstract: Healthcare decision-support is a new, crucial, vibrant area of research motivated by the rapid pace of advances in medical information technology and the vast amount of data that a healthcare provider will have to comprehend and prioritize when technology allows a complete patient medical record to be made available. We believe that the amount of data that the healthcare provider will confront will be overwhelming. In a large-scale medical data environment, preventing data overload and protecting the integrity of the data will be important. Clearly, the healthcare provider requires decision-support tools that will aid in retrieving, identifying, displaying, and analyzing the relevant medical data. Our goal is to develop simulation technologies that can be used to build advanced medical decision-support tools that can exploit the large-scale amounts of medical data that will be available. However, as noted the data must be trustworthy. A cyber-attack upon medical data can disrupt information, sow confusion, thwart situational awareness, increase decision time, and delay reaction to events. Because of the seriousness of the consequences of a cyber-attack, we contend that medical decision-makers must be prepared to operate within environments where information is compromised. A safe method for preparing for the cyber-attacks is to acclimate medical decision-makers to information compromised environments using simulation systems. The cyber-attack simulation environments can cause the information uncertainty and confusion that cyber-attacks produce. These same cyber simulation environments can be used to develop intelligent cyber defense systems that react to preserve the medical information environment. Additionally, the cyber-attack simulation environment can be used to develop and test cyber defense strategies and technologies. In the talk, we will discuss simulation to improve healthcare delivery through the development of better decision-support tools and the use of simulation to improve cyber security, and medical infrastructure cyber resilience. Biography: Dr. Martin Stytz received his PhD form University of Michigan in 1989. His research interests encompass secure systems, secure software development, cybersecurity, high-confidence data analysis, and cyber situational awareness.. Dr. Stytz has published 28 journal articles, over 300 technical articles and holds two patents. Dr. Stytz has conducted $12.8 Million ($8 Million as PI) in research for the US Government.

COLLOQUIUM Selcuk Uluagac Abstract Cyber space is expanding fast with the introduction of new Internet of Things (IoT) and CPS devices. Wearables, smart watches, glasses, fitness trackers, medical devices, Internet-connected house appliances and vehicles have grown exponentially in a short period of time. Our everyday lives will be dominated by billions of connected smart devices by the end of this decade. Similarly, our nation's critical infrastructure (e.g., Smart Grid) also deploys a myriad of CPS and IoT equipment. Given the increasingly critical nature of the cyberspace of these CPS and IoT devices and the CPS infrastructure, it is imperative that they are secured against malicious activities. In this talk, I will briefly introduce three different current research topics related to the security of CPS and IoT devices and the CPS infrastructure: (1) The first topic will introduce the sensory channel threats to CPS and IoT systems. I will discuss how using sensory channels (e.g., light, temperature, infrared), an adversary can successfully attack IoT/CPS applications and devices. (2) The second topic will introduce the design of a novel IoT device fingerprinting and identification framework (IFF) to complement existing security solutions (e.g., authentication and access control) in identifying CPS and IoT devices (i.e., ensuring the devices are actually who they are). Finally, (3) The third topic will focus on the threat of counterfeit smart grid devices (e.g., PMUs, IEDs). Such devices with corrupted hardware components may exist in the deployment region without a priori knowledge and may leak important information to malicious entities. Dr. Selcuk Uluagac is currently an Assistant Professor in the Department of Electrical and Computer Engineering (ECE) at Florida International University (FIU). Before joining FIU, he was a Senior Research Engineer in the School of Electrical and Computer Engineering (ECE) at Georgia Institute of Technology. Prior to Georgia Tech, he was a Senior Research Engineer at Symantec. He earned his Ph.D. with a concentration in information security and networking from the School of ECE, Georgia Tech in 2010. He also received an M.Sc. in Information Security from the School of Computer Science, Georgia Tech and an M.Sc. in ECE from Carnegie Mellon University in 2009 and 2002, respectively. He obtained his BS in Computer Science and Engineering and BA in Naval Science from the Turkish Naval Academy in 1997. The focus of his research is on cyber security topics with an emphasis on its practical and applied aspects. He is interested in and currently working on problems pertinent to the security of Internet of Things and Cyber-Physical Systems. In 2015, he received a Faculty Early Career Development (CAREER) Award from the US National Science Foundation (NSF). In 2015, he was also selected to receive fellowship from the US Air Force Office of Sponsored Research (AFOSR)?s 2015 Summer Faculty Fellowship Program. In 2016, he received the Summer Faculty Fellowship from the University of Padova, Italy. In 2007, he received the ?Outstanding ECE Graduate Teaching Assistant Award? from the School of ECE, Georgia Tech. He is an active member of IEEE (senior grade), ACM, USENIX, and ASEE and a regular contributor to national panels and leading journals and conferences in the field. Currently, he is the area editor of Elsevier Journal of Network and Computer Applications and serves on the editorial board of the IEEE Communication Surveys and Tutorials. More information can be obtained from: http://web.eng.fiu.edu/selcuk.

COLLOQUIUM Xueqing Huang Mobile and wireless systems are bracing for a massive penetration of Internet of Things (IoT) devices and experiencing an exponential growth in wireless applications. To achieve the expected service requirements, the networking resources are being pushed to the edge, such that each edge node can function as a standalone local unit, which has its own green energy harvester, cache and computing resources. Since resources available at the edge nodes are limited and dynamic, network cooperation is critical to guarantee the smooth operation and security of the wireless access networks. In this talk, I will present the network cooperation framework that allows the ?connected? edge nodes to share their networking resources. The first part of this talk concentrates on secure cooperative data transmission. By exploring the broadcasting nature of wireless links, the radio resources in terms of energy can be shared by allowing base stations to transmit data to the same user. I will present cooperative data transmission schemes to improve the network performance in terms of energy efficiency and data confidentiality. The second part of this talk focuses on secure data crowdsourcing. By leveraging multiple data paths at the mobile edge, distributed storage resources can be used to facilitate data sharing among a crowd of users. Data privacy is paramount in assuring users in crowdsourcing; I will present a multi-party data transmission scheme to improve the data sharing latency and data privacy. Xueqing Huang received the B.E. degree from the Hefei University of Technology in 2009, and the M.E. degree from the Beijing University of Posts and Telecommunications in 2012. She is currently a Ph.D. candidate from the New Jersey Institute of Technology. Her research interests are in internet of things, physical layer and network security, and cooperative mobile edge.

COLLOQUIUM Department of Computer Science and Engineering University of South Carolina Damindra Bandara Abstract Safety objectives of Positive Train Control (PTC) are to avoid train to train collisions, train derailments and ensure railroad worker safety. Under published specifications of Interoperable Electronic Train Management System (I-ETMS), the on-board PTC controller communicates with two networks; the Signaling Network and the Wayside Interface Unit (WIU) network to gather navigational information such as the positions of other trains, the status of critical infrastructure and any hazardous conditions along the train path. PTC systems are predicated on having a reliable radio communication network. Secure Intelligent Radio for Trains (SIRT) is an intelligent radio that is customized for train operations with the aim of improving the reliability and security of the radio communication network. SIRT system can (1) operate in areas with high train congestion, different noise levels and interference conditions, (2) withstand jamming attacks, (3) improve data throughput and (4) detect threats and improve communication security. My work includes (1) Analyzing the PTC system to identify communication constraints and vulnerabilities, (2) Designing SIRT to overcome them, (3) Developing a prototype of SIRT using Software Defined Radios and (4) Testing it under varying channel conditions, noise levels and attackers. My experiments show that SIRT dynamically chooses the best modulation schemes based on the channel noise level and switches channels in response to channel jamming. Also, it changes cryptographic key values using a scheme like Lamport scheme and detects replay and forgery attacks with an accuracy more than 93%. Damindra Bandara is a Ph.D. candidate at George Mason University, Fairfax, Virginia. She will defend her Ph.D. by the end of March 2017. She received her Bachelor's degree in Electrical and Electronic Engineering from University of Peradeniya, Sri Lanka and Master degree in Information Security and Assurance from George Mason University. Previously she has worked as a Wireless Quality Assurance intern at Time Warner Cable, Herndon Virginia and as a lecturer in Department of Electrical and Electronic Engineering, University of Peradeniya, Sri Lanka. Her research interests are network security, wireless controlled trains and Software Defined Radio.