Sherie was born and raised on the island of Maui. She graduated from Maui High School, where she cultivated her love for information technology. Currently, she is pursuing a degree in Computer, Electronics, and Networking Technologies at Honolulu Community College. While attending, Sherie competed in the Collegiate Cyber Defense Competition (CCDC) and her team placed 4th for the West Region. Sherie hopes to continue her education in Information Security and Assurance at University of Hawaii–West Oahu. After graduation, she hopes to start a career in networking or cybersecurity. Apart from academics, she enjoys building computers, hiking, and playing basketball.
Home Island: Maui
Institution when accepted: Honolulu Community College
Akamai Project: Detecting Network Anomalies with Deep Learning: Implementing BiGAN
Project Site: Maui High Performance Computing Center – Kihei, Maui
Mentors: Dr. Robert Trevino, Dr. Wesley Emeneker
Collaborator: David Zane
Cyber defense is critical to the U.S. Department of Defense (DoD). DoD’s ability to defend the nation relies on secure network communications. Therefore, the DoD actively searches for new tactics to identify anomalous network behavior that could pose a security threat to the network. Detecting anomalies is a difficult task for cybersecurity professionals, as network threats come in different shapes, sizes, and speeds. The Maui High Performance Computing Center (MHPCC), a DoD Supercomputing Resource Center, is researching and implementing cutting-edge deep learning network anomaly detectors. The anomaly detection systems being developed at MHPCC inspects for potentially threatening events and network activities. In this project, I developed and trained a deep learning model called a Bidirectional Generative Adversarial Network (BiGAN) to perform network security threat analysis. Built with Python and TensorFlow, BiGAN utilizes historical network data to learn to differentiate between anomalous and normal network traffic. I trained the model to generate and detect various anomalies. A goal of the project is to create a model that can learn any network communication principle or network protocol, making it a malleable BiGAN. This will help DoD cybersecurity professionals respond to network security threats with quick anomaly detection and by reducing manual network analysis.