Parallel Firewalls on General-Purpose Graphics Processing Units

TL;DR

This paper explores implementing parallel firewall models on GPGPU to improve throughput and reduce latency, addressing bottlenecks and DoS vulnerabilities in high-speed network security.

Contribution

It introduces a new parallel firewall model on GPGPU that leverages massive parallelism for enhanced performance.

Findings

Improved firewall throughput on GPGPU

Reduced latency in high-speed networks

Enhanced resistance to DoS attacks

Abstract

Firewalls use a rule database to decide which packets will be allowed from one network onto another thereby implementing a security policy. In high-speed networks as the inter-arrival rate of packets decreases, the latency incurred by a firewall increases. In such a scenario, a single firewall become a bottleneck and reduces the overall throughput of the network.A firewall with heavy load, which is supposed to be a first line of defense against attacks, becomes susceptible to Denial of Service (DoS) attacks. Many works are being done to optimize firewalls.This paper presents our implementation of different parallel firewall models on General-Purpose Graphics Processing Unit (GPGPU). We implemented the parallel firewall architecture proposed in and introduced a new model that can effectively exploit the massively parallel computing capabilities of GPGPU.