Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
Abstract for the AZ/NV Academy of Science
Joshua D. Bowen (Northern Arizona University, Prescott Valley, AZ)
Insider incidents are on the rise, just like the high-profile security breaches such as Snowden, thousands of insider-perpetrated
security breaches occur in United States businesses every day. While current commercial software can monitor, log, and prevent
access to designated files and directories, it remains difficult to predict and prevent unauthorized insider usage. Due to
the gaps in research in this area, the focus of this study is to more accurately predict insider threats within a terminal
Linux was chosen specifically because of its ubiquity on commercial servers around the globe. Amazon’s Machine Learning (AML)
service has been selected to analyze the data, reduce the necessary computing power, and to minimize human factors considerations
in the design of the machine learning architecture. AML uses multinomial logistic regression for multi-class classification and
uses the stochastic gradient descent optimization technique.
The method adapted is to utilize the AML software, it will be trained on a dataset comprised of normal user situations, crafted
mistakes, and malicious activity. After providing the training dataset, the software will be instructed to make predictions
against similar datasets to verify accuracy. It will then be tested against a human actor that will simulate multiple different
roles, and test predictions in a high-fidelity simulation.
In result, should it be demonstrated that ML software can accurately identify and predict insider threats, this research could
be a foundation for future cyber security software architectures. Other opportunities for research in this area would include
ML applications in intruder and malware detection.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.