Project Details
Abstract
With the fast growing popularity of the Internet, there is an increasing demand for
network security. Traditional firewalls only provide basic network protection by examining
the information contained in the header of an incoming packet, which is not enough to
satisfy the increasing security requirements. Network Intrusion Detection Systems (NIDS)
have been proposed to complement firewalls. They monitor packets in the network and scan
packet payloads to detect malicious intrusions according to the predefined rules called
patterns or signatures. Studies revealed that pattern matching is the bottleneck in NIDS. For
example, it consumes about 70% of the total execution time in Snort, a popular open-source
NIDS. Thus, it is important to design a fast multi-pattern matching algorithm to cope with
the increasing network speed. The recent trend shows that General-Purpose Graphics
Processing Unit (GPGPU) computing density improves fasters than CPU. Moreover,
GPGPUs are cheap and readily available, which is suitable for computing intensive
applications. In this project, we aim to design an efficient multi-pattern matching algorithm
with GPGPU acceleration. Different from the existing approaches, this project focuses on
balancing the loading between the CPU and the GPGPU. The key idea behind the proposed
approach is to pre-filter packets with a fast algorithm executed by the CPU. Then GPGPU
performs an exact matching algorithm for the packets passing through the CPU. To
understand the design guideline of the pre-filtering algorithm, we analyze the relationship
between the filtering rate and the processing speed. The proposed pre-filtering algorithm is
also analyzed to show that it can fulfill its objectives. We will implement the proposed
multi-pattern matching algorithm in Linux to evaluate its performance.
Project IDs
Project ID:PB10207-1791
External Project ID:NSC102-2221-E182-034
External Project ID:NSC102-2221-E182-034
Status | Finished |
---|---|
Effective start/end date | 01/08/13 → 31/07/14 |
Keywords
- multi-pattern matching
- general-purpose graphics processing unit
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.