Software that covertly monitors user actions, also known as spyware, has become a first-level security threat due to its ubiquity and the difficulty of detecting and removing it. This is especially so for video conferencing, thin-client computing and Internet cafes.

CryptoGraphics: Exploiting Graphics Cards for Security explores the potential for implementing ciphers within GPUs, and describes the relevance of GPU-based encryption to the security of applications involving remote displays. As the processing power of GPUs increases, research involving the use of GPUs for general purpose computing has arisen. This work extends such research by considering the use of a GPU as a parallel processor for encrypting data. The authors evaluate the operations found in symmetric and asymmetric key ciphers to determine if encryption can be programmed in existing GPUs. A detailed description for a GPU based implementation of AES is provided. The feasibility of GPU-based encryption allows the authors to explore the use of a GPU as a trusted system component. Unencrypted display data can be confined to the GPU to avoid exposing it to any malware running on the operating system.

The focus of this book is the use of graphics processing units (GPUs) for cryptographies operations, hence the term CryptoGraphics. The computing power of GPUs has increased substantially over the past several years to the point that GPUs are more efficient than CPUs for certain tasks. As a result, even though GPUs are not intended to be general purpose processors, researchers have begun to study the use of GPUs for non-graphics applications. In most cases, the goal is to increase the rate at which computations can be performed by an application by using the GPU for specific types of calculations. Applications that are well suited to run in a GPU use data representations and types that are compatible with the GPU's abstraction of pixels. Compatible computations involve operations that take a single pixel's value, apply a simple function to it and output the result as a new pixel value. Parallel processing on multiple data sets can be performed by using multiple sets of pixels to represent the data sets and by applying the application simultaneously to each set and/or by treating each color component of a pixel as a separate set of data and applying the algorithm in parallel to each color component.