Towards Building Efficient Malware Detection Engines Using Hybrid CPU/GPU-Accelerated Approaches

Towards Building Efficient Malware Detection Engines Using Hybrid CPU/GPU-Accelerated Approaches

Ciprian Pungila, Viorel Negru
DOI: 10.4018/978-1-4666-4514-1.ch009
(Individual Chapters)
No Current Special Offers


This chapter presents an outline of the challenges involved in constructing efficient malware detection engines using hybrid CPU/GPU-accelerated architectures and discusses how one can overcome such challenges. Starting with a general problem description for malware detection and moving on to the algorithmic background involved for solving it, the authors present a review of the existing approaches for detecting malware and discuss how such approaches may be improved through GPU-accelerated processing. They describe and discuss several hybrid hardware architectures built for detecting malicious software and outline the particular characteristics of each, separately, followed by a debate on their performance and most suitable application in real-world environments. Finally, the authors tackle the problem of performing real-time malware detection and present the most important aspects that need to be taken into account in intrusion detection systems.
Chapter Preview


This section introduces the concept of malware, presenting the different forms under which the concept is found today, and presents a few common approaches for achieving malware detection. We also present the algorithmic background required for achieving malware detection, beginning with the most commonly used multiple pattern-matching algorithm and discussing a few common architectures: in particular, we present the RMAS (Run-time Malware Analysis System) and ClamAV approaches, also outlining known solutions to employing different heuristics for detecting malicious program behavior at run-time. The last part of this section covers NVIDIA's CUDA framework and architecture.

Complete Chapter List

Search this Book: