Date of Award

12-2026

Document Type

Project

Degree Name

Master of Science in Computer Science

Department

School of Computer Science and Engineering

First Reader/Committee Chair

Salloum, Ronald

Abstract

Cheating in online chess has become prevalent in recent years, especially as chess engines become more accessible and more powerful. In this work, we analyze the current landscape of cheat detection methods in online chess, and offer our own approach to detecting unfair games. We also analyze similar works to acknowledge the variety of cheating methods that are commonly used, and develop our methodology accordingly. Past works typically use convolutional neural networks (CNNs) to approach this problem, and we experimented with deep learning models such as densely connected neural networks (DNNs), recurrent neural networks (RNNs), CNNs, and ensemble methods. We also experimented with classical models such as clustering, k-nearest-neighbors (KNN), label spreading, and random forest. We formulated a preprocessing method that extracted valuable insights from each game, such as engine analysis, and evaluated its effect on deep learning models. We found that CNNs are the best choice for this problem in terms of accuracy, as our CNN model was able to perform at an 80% testing accuracy. Our ensemble model had a comparable accuracy as well, and overall we determined that our preprocessing method had a significantly positive impact on the testing accuracy.

Share

COinS