Date of Award:

5-2025

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Electrical and Computer Engineering

Committee Chair(s)

Koushik Chakraborty

Committee

Koushik Chakraborty

Committee

Sanghamitra Roy

Committee

Chris Winstead

Committee

Jonathan Phillips

Abstract

A Field Programmable Gate Array (FPGA) is a special type of computer chip that can be reconfigured to implement a nearly unlimited number of functions. Recent trends have led some companies to offer cloud-based FPGA solutions. Researchers are exploring how to properly secure multi-tenant environments where the designs from two or more customers are placed on the same FPGA with logical and spatial isolation, though multi-tenancy is not yet commercially available in cloud FPGAs. The digital circuits within an FPGA require a clock to synchronize timing within the design. The maximum speed that this clock can run at is determined by the amount of logic in a single clock cycle, quality of the silicon chip, supply voltage, and several other factors. Furthermore, if the voltage is reduced without reducing the clock speed, the device will encounter timing errors and produce erroneous outputs.

This work attempts to cause a voltage-induced timing error in a multi-tenant FPGA environment where one design is a victim AI application and the other is a malicious power plundering design intended to cause short disruptions in the supply voltage called voltage transients. This is done by dramatically increasing the power usage of the malicious design over a very short period of time. Although the power supply is capable of delivering this much power, it takes time for it to adjust to the increased demand. This brief period, where the power supply has not caught up with the demand, causes a voltage drop which can lead to timing errors if severe enough. The opposite, a voltage spike, occurs when power usage dramatically decreases.

Checksum

0e2a15edbd132b40525ae8e5325e021e

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS