FPGA Based Multi-Tier Artificial Neural Network Processor for Firewall Implementation

Abstract

Artificial intelligence (AI) tools, such as expert system, fuzzy logic, and neural network are likely to usher a new era in computer security and intrusion detection in the coming decades. Although these technologies have advanced significantly in recent years and have found wide applications, they have hardly penetrated in the areas of computer security mainly due to lack of interdisciplinary framework amongst the mathematicians, cyber experts and Very Large Scale Integration (VLSI) engineers specialized in the areas of reconfigurable soft computing platforms such as field programmable gate arrays (FPGA). The present paper showcases usefulness of the mathematized methodology for drawing inferences about the world from uncertain knowledge, such as the ever-changing hacking patterns posing grave security threats. The core theoretical principles for the implementation of firewall are derived from the well established mathematical concepts such as entropy and probability. FPGA based platform is used for realization of the multi-tier ANN processor for getting computational efficiency and less latency besides inculcation of the human expert intelligence. The novel features of the implemented semi custom Application Specific Integrated Circuit (ASIC) are parallelism, background processing and predicative output that decides whether to allow or deny the ongoing session between two hosts. Handel C based coding is adopted for C to Register Transfer Logic (RTL) conversion so as to get the advantages such as an early development cycle, ease of algorithmic expressions and fast prototyping. The paper also presents the main rationale behind application of Artificial Neural Network for firewall applications, formulation of the mathematical model and its subsequent coding to realize the device in FPGA paradigm.