Speaker: Nabor Mendonça Affiliation: University of Fortaleza, Brazil Location: Innovation Center, Room 2277 Time: Friday 7/26/2019 (2 - 3pm) Host: Pooyan Jamshidi (Please contact me if you want to meet with the speaker) Abstract: In this talk I'll give a brief overview of how the concept of software architecture has evolved over the last 50 years, from centralized monolithic systems to today's highly distributed cloud-based applications. Then I'll discuss some of the key challenges facing the development of cloud native applications, with a focus on the microservice architectural style and its supporting practices and technologies. Bio: Nabor Mendonça is a full professor in applied informatics at the University of Fortaleza, Brazil. From 2017 to 2018 he was a visiting researcher at the Institute for Software Research, Carnegie Mellon University, working in David Garlan's ABLE group. His main research areas are software engineering, distributed systems, self-adaptive systems, and cloud computing. Prof. Mendonça has a Ph.D. in computing from Imperial College London.

DISSERTATION DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Emad Alsuwat Advisor : Dr. Farkas and Dr. Valtorta Date : June 12th, 2019 Time : 2:00 pm Place : Meeting Room 2265, Innovation Center Abstract In this research, we address the impact of data integrity on machine learning algorithms. We study how an adversary could corrupt Bayesian network structure learning algorithms by inserting contaminated data items. We investigate the resilience of Bayesian network structure learning algorithms, namely the PC and LCD algorithms, against data poisoning attacks that aim to corrupt the learned Bayesian network model. Data poisoning attacks are one of the most important emerging security threats against machine learning systems. These attacks aim to corrupt machine learning models by contaminating datasets in the training phase. The lack of resilience of Bayesian network structure learning algorithms against such attacks leads to inaccuracies of the learned network structure. In this dissertation, we propose two subclasses of data poisoning attacks against Bayesian networks structure learning algorithms: (1) Model invalidation attacks when an adversary poisons the training dataset such that the Bayesian model will be invalid, and (2) Targeted change attacks when an adversary poisons the training dataset to achieve a specific change in the structure. We also define a novel measure of the strengths of links between variables in discrete Bayesian networks. We use this measure to find vulnerable sub-structure of the Bayesian network model. We use our link strength measure to find the easiest links to break and the most believable links to add to the Bayesian network model. In addition to one-step attacks, we define long-duration (multi-step) data poisoning attacks when a malicious attacker attempts to send contaminated cases over a period of time. We propose to use the distance measure between Bayesian network models and the value of data conflict to detect data poisoning attacks. We propose a 2-layered framework that detects both traditional one-step and sophisticated long-duration data poisoning attacks. Layer 1 enforces “reject on negative impacts” detection; i.e., input that changes the Bayesian network model is labeled potentially malicious. Layer 2 aims to detect long-duration attacks; i.e., observations in the incoming data that conflict with the original Bayesian model. Our empirical results show that Bayesian networks are not robust against data poisoning attacks. However, our framework can be used to detect and mitigate such threats.

DISSERTATION DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Jason Moulton Advisor : Dr. Ioannis Rekleitis Date : May 3rd, 2019 Time : 9:00 am Place : Meeting Room 2267, Innovation Center Abstract This dissertation has four main contributions. The first contribution is the design and build of a fleet of long\hyp range, medium\hyp duration deployable autonomous surface vehicles (ASV). The second is the development, implementation and testing of inexpensive sensors to accurately measure wind, current, and depth environmental variables. The third leverages the first two contributions, and is modeling the effects of environmental variables on an ASV, finally leading to the development of a dynamic controller enabling deployment in more uncertain conditions. The motivation for designing and building a new ASV comes from the lack of availability of a flexible and modular platform capable of long\hyp range deployment in current state of the art. We present a design of an autonomous surface vehicle (ASV) with the power to cover large areas, the payload capacity to carry sufficient power and sensor equipment, and enough fuel to remain on task for extended periods. An analysis of the design, lessons learned during build and deployments, as well as a comprehensive build tutorial is provided in this thesis. The contributions from developing an inexpensive environmental sensor suite are multi-faceted. The ability to monitor, collect and build models of depth, wind and current in environmental applications proves to be valuable and challenging, where we illustrate our capability to provide an efficient, accurate, and inexpensive data collection platform for the communities use. More selfishly, in order to enable our end\hyp state goal of deploying our ASV in adverse environments, we realize the requirement to measure the same environmental characteristics in real\hyp time and provide them as inputs to our effects model and dynamic controller. We present our methods for calibrating the sensors and the experimental results of measurement maps and prediction maps from a total of 70 field trials. Finally, we seek to inculcate our measured environmental variables along with previously available odometry information to increase the viability of the ASV to maneuver in highly dynamic wind and current environments. We present experimental results in differing conditions, augmenting the trajectory tracking performance of the original way\hyp point navigation controller with our external forces feed\hyp forward algorithm.

Dr. Xuetao Weifrom the School of Information Technology at the University of Cincinnati will give a talk on Monday April 29 at 10:15 - 11:15, in Storey Innovation Center (Room 2277). Abstract: Cyberspace is a constantly changing landscape. Not only security attacks have become more and more stealthy, but also a myriad of opaque security applications have emerged. Navigating the increasingly complex cyber threat landscape has become overwhelming. Thus, it is essential to profile and understand emerging security attacks and applications. In this talk, I will first present a novel approach and tool, which illuminates in-memory injection attacks via provenance-based whole-system dynamic information flow tracking. Then, I will present a framework to enable behavior-based profiling for smart contracts, which could enhance users' understanding and control of contract behavior and assess performance and security implications. Finally, I will briefly discuss current ongoing research and future directions. Bio: Xuetao Wei is a tenure-track assistant professor in the School of Information Technology at the University of Cincinnati. He received his Ph.D. in Computer Science from the University of California, Riverside. His research interests span the areas of cybersecurity, blockchain, and measurements. His current research is supported by federal and state agencies, including NSF, DARPA, and Ohio Cyber Range. He particularly enjoys solving problems and developing innovative solutions based on interdisciplinary perspectives.

Dr. Shuo Li from the Department of Medical Imaging and Medical Biophysics at the University of Western Ontario will give a talk on Wednesday April 24 at 10:15 - 11:15, in Storey Innovation Center (Room 2277). Medical data analysis is going through great changes with tremendous new opportunities showing up. The rise of machine learning and the rise of big data analytics, have brought wonderful opportunities to invent the new generation of machine learning tools for medical data analytics, not only to solve new problems appearing, but also to solve many years challenges in conventional medical image analysis and computer vision with much more satisfactory solutions. This talk will share our experience in developing the next generation of image analytics tools with newly invented machine learning tools to help physicians, hospital administrative to analyze the growing medical data and help them to make the right decision and early decision at the right timing. Dr. Shuo Li is an associate professor in the department of medical imaging and medical biophysics at the University of Western Ontario and scientist at Lawson Health Research Institute. Before this position he was a research scientist and project manager in General Electric (GE) Healthcare for 9 years. He founded the Digital Imaging Group of London (http://digitalimaginggroup.ca/) in 2006, which is a very dynamic and highly multiple disciplinary collaboration group. He received his Ph.D. degree in computer science from Concordia University 2006, where his PhD thesis won the doctoral prize giving to the most deserving graduating student in the faculty of engineering and computer science. He has published over 100 publications; He is the recipient of several awards from GE, institutes and international organizations; He serves as guest editors and associate editor in several prestigious journals in the field; He serves as program committee members in highly influential conferences; He is the editors of six Springer books; He serves on the board of directors in prestigious MICCAI society. He will be the general chair for MICCAI 2022 conference. His current interest is development intelligent analytic tools to help physicians and hospital administrators to handle the big medical data, centred with medical images.

At Backers & Hackers, EclubSC is excited to share with you all the new apps that students from all majors have developed this semester. Join us for a night of networking, showcasing, and hearing from Laura Boccanfuso. Laura Boccanfuso is the CEO of Vän Robotics. A company that provides students K-8 a robot-assisted tutoring to help enhance the students' learning experience. Hear how she got involved in the world of technology and education, and her valuable insights of her journey to getting where she is today! More Info

Dr. Nihat Altiparmak from the Department of Computer Engineering and Computer Science at the University of Louisville, will give a talk on Friday, April 12, 2019, in the Storey Innovation Center (Room 2277) from 10:15 am - 11:15 am. Abstract Today’s most critical applications, including genome analysis, climate simulations, drug discovery, space observation, and numerical simulations in computational chemistry and high-energy physics, are all data intensive in nature. Storage performance bottlenecks are major threats limiting the performance and scalability of data intensive applications. A common way to address storage I/O bottlenecks is using parallel storage systems and utilizing concurrent operation of independent storage components; however, achieving a consistently high parallel I/O performance is challenging due to static configurations. Modern parallel storage systems, especially in the cloud, enterprise data centers, and scientific clusters are commonly shared by various applications generating dynamic and coexisting data access patterns. Nonetheless, these systems generally utilize one-layout-fits-all data placement strategy frequently resulting in suboptimal I/O parallelism. Guided by association rule mining, graph coloring, bin packing, and network flow techniques, this talk demonstrates a general framework for self-optimizing parallel storage systems that can adaptively alleviate storage performance bottlenecks and continuously provide a high-degree of I/O parallelism. The framework can be applied to a wide range of parallel storage architectures including storage arrays, key-value stores, parallel/distributed file systems, and internal parallelism of solid-state drives. In addition, this talk briefly covers efficient storage, retrieval, and processing strategies for Big Data, and recent advancements in non-volatile memory technology by identifying upcoming challenges in computer systems research to utilize new solid-state storage devices to their full potential. Bio: Dr. Nihat Altiparmak earned his B.S. degree in Computer Engineering from Bilkent University, Ankara, Turkey in May 2007, and his combined M.S. and Ph.D. degrees in Computer Science from the University of Texas at San Antonio in May 2013. He joined the Department of Computer Engineering and Computer Science at the University of Louisville as a tenure-track Assistant Professor in August 2013, and his tenure and promotion to Associate Professor is currently pending approval by the Board of Trustees of the University of Louisville. His research interests lie in the area of computer systems, specifically focusing on data storage systems, parallel and distributed systems, cloud computing, high performance computing, and computer networks. He is particularly interested in researching solid-state storage systems based on new generation non-volatile memory technologies, as well as investigating efficient storage, retrieval, and processing strategies for Big Data using high performance, distributed, and cloud architectures. His recent research findings have appeared in top-tier international journals, including IEEE Transactions on Computers, IEEE Transactions on Parallel and Distributed Systems, ACM Transactions on Storage, and ACM Transactions on Sensor Networks, as well as prestigious conferences with competitive acceptance rates. He received multiple grants from the National Science Foundation (NSF) in the PI role, including a prestigious young investigator award (NSF CRII) in 2017 and an NSF MRI award in 2018. He is a senior member of the IEEE and the founding director of the Computer Systems Laboratory at the University of Louisville.

THESIS DEFENSE Department of Computer Science and Engineering University of South Carolina Author : Kimberly Redmond Advisor : Dr. Lisa Luo Date : April 10th , 2019 Time : 11:00 am Place : Meeting room 2267, Bert Storey Innovation Center Abstract Binary code analysis is important for understanding programs without access to the original source code, which is common with proprietary software. Analyzing binaries can be challenging given their high variability: due to growth in tech manufacturers, source code is now frequently compiled for multiple instruction set architectures (ISAs); however, there is no formal dictionary that translates between their assembly languages. The difficulty of analysis is further compounded by different compiler optimizations and obfuscated malware signatures. Such minutiae means that some vulnerabilities may only be detectable on a fine-grained level. Recent strides in machine learning---particularly in Natural Language Processing (NLP)---may provide a solution: deep learning models can process large texts and encode the semantics of individual words into vectors called word embeddings, which are convenient for processing and analyzing text. By treating assembly as a language and instructions as words, we leverage NLP ideas in order to generate individual instruction embeddings. Specifically, we choose to improve upon current models that are only single-architecture, or that suffer from performance issues when handling multiple architectures. This research presents a cross-architecture instruction embedding model that jointly encodes instruction semantics from multiple ISAs, where similar instructions within and across architectures embed closely together. Results show that our model is accurate in extracting semantics from binaries alone, and our embeddings capture semantic equivalences across multiple architectures. When combined, these instruction embeddings can represent the meaning of functions or basic blocks; thus, this model may prove useful for cross-architecture bug, malware, and plagiarism detection.

Monday, April 8, 2019, in the Storey Innovation Center (Room 2277) from 10:15 am - 11:15 am. Abstract: There are two fundamental strategies for finding effective policies to solve stochastic optimization problems and these are policy search and look-ahead policies. Look-ahead policies are mainly used in the context of sequential optimization problems in which the current decision impacts the future ones. There are several types of look-ahead policies depending on the type of the forecast, the problem structure and its dimensionality. In this talk, I will discuss two forms of look-ahead policies, direct look-aheads and value function approximations. In the context of direct look-aheads, we propose a new technique called Primal-Dual Monte Carlo Trees Search that utilizes sampled information relaxation upper bounds on potential actions, creating the possibility of “ignoring" parts of the tree that stem from highly suboptimal choices. This allows us to prove that despite converging to a partial decision tree in the limit, the recommended action from Primal-Dual MCTS is optimal. Then, I will discuss an approximate dynamic programming approach in the context of ride-sharing systems. We extract and prove important properties about the problem structure such as monotonicity and spatial correlation that provide efficient value function approximations. Biography: Lina Al-Kanj is an Associate Research Scholar at the Operations Research and Financial Engineering Department at Princeton University. She received her PhD in Electrical and Computer Engineering from the American University of Beirut. Her research interests include optimal resource allocation and scheduling, stochastic optimization, dynamic programming and optimal learning with applications to energy, communication and transportation systems.