Final Examination Schedule
PLEASE JOIN US AS THE FOLLOWING CANDIDATES PRESENT THEIR CULMINATING WORK.
Tuesday, December 3
Chair: Dr. Geethapriya Thamilarasu
Candidate: Master of Science in Cybersecurity Engineering
3:30 P.M.; DISC 464
Federated Learning for Intrusion Detection in the Cloud
The economic benefits of cloud computing continue to motivate organizations across private and public sectors to transition their information technology to the cloud. Healthcare organizations find it challenging to achieve the desired cost reduction without compromising on security, privacy, or compliance. Existing security controls such as encryption or static firewalls are not a catch-all for ensuring security and privacy. In order to identify attack behavior based on more versatile patterns, Intrusion Detection Systems have been developed utilizing machine learning. While they detect many attacks, the aggregation of all network data for training negatively impacts privacy. Federated learning is a machine learning approach that honors data privacy by enabling training to be distributed across individual clients. It has been used broadly in mobile computing. In a federated setting, the resulting model is shared among the participants while the training data is not. We propose Defense Guild, an intrusion detection system that relies on federated learning to detect attacks within partitioned cloud healthcare record systems. We apply federated learning techniques from mobile computing to maintain privacy boundaries between healthcare providers in the cloud. Defense Guild delivers attack detection rates and false positive rates comparable to traditional systems while adding these privacy improvements.
Thursday, December 5
Chair: Dr. Min Chen
Candidate: Master of Science in Computer Science & Software Engineering
11:00 A.M.; DISC 464
IMPACT of sentiment analysis of STOCKTWITS data ON STOCK PRICE
Understanding stock market movements and predicting them is a well-known problem of interest. In general, it is believed stock markets are driven by public sentiments. The usage of social media has grown rapidly over past few years. The amount of new content generated online is increasing exponentially and a significant portion of this content is in the form users opinions and sentiments. Stock market prediction on the basis of public sentiments expressed on social media especially microblogging websites like twitter has been an intriguing field of research. In this project, we focus on a relatively newer microblogging website called stocktwits, which encourages financial discussions and is becoming increasingly popular where users share their sentiments about stocks, financial markets and related things multiple times a day. More specifically, we analyze the text content of stocktwits tweets and extract financial sentiment (bullish or bearish), study correlation between the aggregated daily sentiment and daily stock price movement, and finally use the sentiment information as an additional signal to improve the accuracy of stock price movements. We used stocktwits data for 9 months for 5 companies and developed a finance specific sentiment analyzer by experimenting with various combinations of text featurization and machine learning algorithms. We used daily stock data for same companies for the same duration and establish correlation between stock price movements and sentiments detected from stock twits on the same day and previous days. Finally, this sentiment signal is used as an additional signal to time series data for predicting stock price movements. We also implemented an end-to-end scalable and efficient software system for detailed analysis and experimentation with stocktwits data.
Back to top
Questions: Please email firstname.lastname@example.org