• home
• current students
• graduate
• events

Presenter: Ahmed Farag
Supervisor:

Date: Tue, July 25, 2023
Time: 08:00:00 - 09:00:00
Place: via Zoom - please see link below

ABSTRACT

Zoom Meeting Link: https://uvic.zoom.us/j/89208328926?pwd=ZWhSaDlwRFppQjRac3NJUDcwZ2p5UT09

Meeting ID: 892 0832 8926

Password: 015919

You can also dial-in:

    +1 647 558 0588 Canada

        +1 778 907 2071 Canada

Abstract:

In today's data-driven world, the detection of unusual patterns or anomalies in data sets, especially in the security domain, is becoming increasingly vital, as it can pre-empt security threats. Our work serves as an extension to UNAVOIDS (Unsupervised and Nonparametric Approach for Visualizing Outliers and Invariant Detection Scoring), a distinctive model combining specialized techniques for detection algorithms and visualization methods within a unique realm known as the Neighborhood Cumulative Distribution Function (NCDF) space. In this space, each data point transforms into a unique 2D curve, allowing for visual identification and examination. Notably, UNAVOIDS is fully unsupervised, negating the need for prior training or specific data inputs, thereby eliminating the necessity for parameter selection or tuning, and it assigns a deviation score to each unusual data point, providing a clear abnormality indicator. The model was successfully implemented across the Python Package Index, a Restful API, a software called VAAD, which integrates UNAVOIDS with the Data Visualization Platform, and a custom Microsoft PowerBI visual. This implementation confronted and overcame two main challenges: handling large data sets within the RESTful API, for which we adopted compression over file streaming to enable efficient data transmission within the API constraints, and creating an interactive visual representation, which was challenging due to the unique nature of the data, where each observation is mapped to a 2D curve. We tackled this issue by mapping curve indices and implementing a reflection mechanism for interactivity between selected curves and other visuals. Our study demonstrates the practical implementation and effectiveness of UNAVOIDS, and all these implementations and their corresponding documentations are accessible from the official repository of the ISOT lab, catering to users across various sectors, including research and development, showcasing the versatility and effectiveness of UNAVOIDS in diverse environments.