Dos and memory exploitation sidebar attacks on graphics processing units on windows operating systems.

Thumbnail Image
Date
2021
Authors
Lungu, Nelson
Journal Title
Journal ISSN
Volume Title
Publisher
The University of Zambia
Abstract
This research investigated the vulnerabilities of Graphics Processing Units (GPUs) to Denial-of-Service (DoS) and Memory Exploitation sidebar attacks on Windows Operating Systems using recovery time. The simulation results showed that all three Windows operating systems and their respective GPUs are vulnerable to DoS attacks, as evidenced by a significant reduction in throughput when malicious polygons were introduced. The Intel GPU on Windows 7 experienced the most severe impact on performance, with a throughput of only 1,000,000 polygons per second when 8 million malicious polygons were introduced. The Nvidia GPU on Windows 10 and the iRiS GPU on Windows 11 experienced a throughput reduction of 500,000 and 250,000 polygons per second, respectively. Furthermore, the screen freeze times decreased as the GPU throughput decreased for each Windows operating system tested. The screen froze indefinitely for Windows 7 with an Intel GPU, with the highest throughput of 1,000,000 polygons/s, indicating a complete denial of service. This research provides quantitative evidence that GPU vulnerabilities can be exploited through malicious polygons and memory exploitation sidebar attacks on Windows Operating Systems, significantly impacting system performance and stability.
Description
Master of Engineering in ICT Security
Keywords
Graphics processing units. , Computer graphics. , Rendering (Computer graphics) , Real-time data processing. , Graphics processing units--Programming.
Citation
Collections