THESIS DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Chandrahas Raj G. Venkat Advisor : Dr. Rose ABSTRACT Hidden Markov Models are class of statistical models used in various disciplines for understanding speech, finding different types of genes responsible for cancer and many more. In this thesis, Hidden Markov Models are used to obtain hidden states that can correlate the flow changes in the Wakulla Spring Cave. Sensors installed in the tunnels of Wakulla Spring Cave recorded huge correlated changes in the water flows at numerous tunnels. Assuming the correlated flow changes are a consequence of system being in a set of discrete states, a Hidden Markov Model is calculated. This model comprising all the sensors installed in these conduits can help understand the correlations among the flows at each sensor and estimate the hidden states. In this thesis, using the Baum - Welch algorithm and observations from the sensors hidden states are calculated for the model. The generated model can help identify the set of discrete states for the quantized flow rates at each sensor. The hidden states can predict the correlated flow changes. This document further validates the assumption of the system being in a set of discrete states.

THESIS DEFENSE Department Of Computer Science and Engineering University of South Carolina Sourav Das ABSTRACT Time-series analysis is used heavily in modelling and forecasting weather, economics, medical data as well as in various other fields. Change point detection (CPD) means finding abrupt changes in the time-series when the statistical property of a certain part of it starts to differ. CPD has attracted a lot of attention in the artificial intelligence, machine learning and data mining communities. In this thesis, a novel CPD algorithm is introduced for segmenting multivariate time-series data. The proposed algorithm is a general pipeline to process any high dimensional multivariate time-series data using non-linear non-parametric dynamic system. It consists of manifold learning technique for dimensionality reduction, Gaussian process regression to model the non-linear dynamics of the data and predict the next possible time-step, as well as outlier detection based on Mahalanobis distance to determine the change points. The performance of the new CPD algorithm is assessed on synthetic as well as real-world data for validation. The pipeline is used on economic data to predict recession. Finally, functional magnetic resonance imaging (fMRI) data of larval zebrafish is used to segment regions of homogeneous brain activity.

Speaker: Jack Beasley (Managing Director) Affiliation: USC/Columbia Technology Incubator Location: SWGN 2A31 When: Friday, November 4th @ 2:20 - 3:10 PM Abstract: Join us Friday, November 4th at 2:20PM in SWGN 2A31 to learn about Startup Weekend. Startup Weekend is a global, grassroots movement designed to give participants a taste of the startup life and teaches the basics of launching successful ventures. The USC/Columbia Technology Incubator will be here to give us the details, and talk with us about how they help startups turn their ideas into sustainable businesses.

Speaker: Jack Beasley (Managing Director) Affiliation: USC/Columbia Technology Incubator Title: Turning Your Skills into a Business in 54 Hours Location: SWGN 2A31 When: Friday, November 4th @ 2:20 - 3:10 PM Join us Friday to learn about Startup Weekend. Startup Weekend is a global, grassroots movement designed to give participants a taste of the startup life and teaches the basics of launching successful ventures. The USC/Columbia Technology Incubator will be here to give us the details, and talk with us about how they help startups turn their ideas into sustainable businesses.

ACM will be having it's second meeting of the semester on Thursday, Novemeber 3 at 7:00PM in Amoco Hall. November 8 is election day so Dr. Duncan Buell will be speaking about voting machines and their relevance to the upcoming election. There will also be pizza for those who attend. More Details and RSVP

Speaker: Juan Caicedo Affiliation: Department of Civil and Environmental Engineering, USC Location: SWGN 2A31 Time: 2:20 - 3:10 PM

DISSERTATION DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Muhammad Nazmus Sakib Advisor: DR. Chin-Tser Huang Date : Oct 26th 2016 Time : 8:30 am Place : 3A75 Swearingen ABSTRACT In this age of ubiquitous communication in which we can stay constantly connected with the rest of the world, for most of the part, we have to be grateful for one particular invention - the Internet. But as the popularity of Internet connectivity grows, it has become a very dangerous place where objects of malicious content and intent can be hidden in plain sight. In this dissertation, we investigate different ways to detect and capture these malicious contents hidden in the Internet. First, we propose an automated system that mimics high-risk browsing activities such as clicking on suspicious online ads, and as a result collects malicious executable files for further analysis and diagnosis. Using our system we crawled over the Internet and collected a considerable amount of malicious executables with very limited resources. Malvertising has been one of the major recent threats against cyber security. Malvertisers apply a variety of evasion techniques to evade detection, whereas the ad networks apply inspection techniques to reveal the malicious ads. However, both the malvertiser and the ad network are under the constraints of resource and time. In the second part of this dissertation, we propose a game theoretic approach to formulate the problem of inspecting the malware inserted by the malvertisers into the Web-based advertising system. During malware collection, we used the online multi-AV scanning service VirusTotal to scan and analyze the samples, which can only generate an aggregation of antivirus scan reports. We need a multi-scanner solution that can accurately determine the maliciousness of a given sample. In the third part of this dissertation, we introduce three theoretical models, which enable us to predict the accuracy levels of different combination of scanners and determine the optimum configuration of a multi-scanner detection system to achieve maximum accuracy. Malicious communication generated by malware also can reveal the presence of it. In the case of botnets, their command and control (C&C) communication is a good candidate for it. Among the widely used C&C protocols, HTTP is becoming the most preferred one. However, detecting HTTP-based C&C packets that constitute a minuscule portion of everyday HTTP traffic is a formidable task. In the final part of this dissertation, we present an anomaly detection based approach to detect HTTP-based C&C traffic using statistical features based on client generated HTTP request packets and DNS server generated response packets.

Abstract: For many complex physical and biological models, the computational cost of high-fidelity simulation codes precludes their direct use for Bayesian model calibration and uncertainty propagation. Furthermore, the models often have tens to thousands of inputs--comprised of parameters, initial conditions, or boundary conditions--many of which are unidentifiable in the sense that they cannot be uniquely determined using measured responses. In this presentation, we will discuss techniques to isolate influential inputs and employ surrogate models when computational budgets are limited. For input selection, we will discuss the use of global sensitivity analysis methods to isolate influential inputs and active subspace construction for linearly related parameters. We will also discuss the manner in which Bayesian calibration on active subspaces can be used to quantify uncertainties in physical parameters. These techniques will be illustrated for models arising in nuclear power plant design and HIV characterization and treatment. Biosketch: Ralph Smith received his PhD in Applied Mathematics from Montana State University in 1990. Following a three-year postdoctoral position at the Institute for Computer Applications in Science and Engineering (ICASE) at NASA Langley Research Center, he was an Assistant Professor in the Department of Mathematics at Iowa State University. He joined the North Carolina State University faculty in 1998, where he is presently a Distinguished Professor of Mathematics. He is Editor-in-Chief of the SIAM book series on Advances in Design and Control and is on the editorial boards of the SIAM/ASA Journal on Uncertainty Quantification and the Journal of Intelligent Material Systems and Structures. He is co-author of the research monograph Smart Material Structures: Modeling, Estimation and Control and author of the books Smart Material Systems: Model Development and Uncertainty Quantification: Theory, Implementation, and Applications. His research areas include mathematical modeling of smart material systems, numerical analysis and methods for physical systems, Bayesian model calibration, sensitivity analysis, control, and uncertainty quantification.

Google representatives will be hosting a presentation on opportunities at Google for interested students on Wednesday, October 5th at 5:00PM in Amoco Hall on the 1st floor of Swearingen.