The Electromagnetic Interference (EMI) Detection group aims to develop a software based methodology to detect the presence of EMI
and study the effects it has on software. Modern electronics are becoming more susceptible to EMI due to their rising clock speeds
and smaller overall size which makes it ever so important for critical systems to understand how EMI affects them. Having the ability
to understand when and how a device is being affected by EMI will allow critical systems to further harden against the effects of EMI
and provide more reliable services. Furthermore, a software based analysis method allows for a better understanding of hardware that
is infeasible to equip with laboratory sensors such as large-scale systems or consumer hardware. Our technique uses a lightweight
watchdog to log changes in key registers and uses these registers values to construct states. When the device is active these states
are recorded and used to give insight via categorical time series statistics and classification algorithms. Below you can find
information about the device we use and specific detection methods.
For our experiments we examine the operation of a USB 2.0 host controller on a Rock Pro 64. We chose the Rock Pro 64 as it is a well
documented and affordable system that has a built-in USB 2.0 controller. For further information on the Rock Pro 64 system please
visit the link below.
Time Series Statistics
Hidden Markov Model
Recurrent Neural Network Long/Short Term Memory Classifier
Artificial Neural Network
Support Vector Machine
Random Forest Classifier
Gradient Boosted Classifier
Joel Schott - Doctorate Student
Evan Hite - Undergraduate Scholar
Connor Jones - Accelerated Master's Student
Austin Potter - Undergraduate Scholar
This website contains additional information about the Rock Pro 64 system.