---

ENHANCING AI MODEL SECURITY USING A HARDWARE-BASED APPROACH FOR PROTECTING FPGA IMPLEMENTATION

Abstract In the ever-evolving landscape of artificial intelligence (AI), the vulnerability of AI models to adversarial attacks has become a critical concern. The problem at hand lies in the lack of dedicated security mechanisms tailored for FPGA-based AI implementations, leaving them exposed to threats such as tampering, reverse engineering, and unauthorized access. This research addresses the pressing need for robust security measures by proposing a hardware-based approach to protect FPGA (Field-Programmable Gate Array) implementations of AI models. FPGAs offer a flexible and efficient platform for deploying AI models but are susceptible to attacks that compromise the integrity of the implemented algorithms. This involves the integration of specialized security modules within the FPGA architecture. These modules are designed to detect and thwart various forms of attacks, including side-channel attacks and unauthorized access attempts. Leveraging the inherent parallelism and reconfigurability of FPGAs, the security modules operate seamlessly alongside the AI model, imposing minimal overhead on performance. Results from experimental evaluations demonstrate the effectiveness of the hardware-based security approach in preventing unauthorized access and tampering with the FPGA-based AI model. The proposed solution showcases resilience against common attack vectors, ensuring the confidentiality and integrity of the deployed AI models. Authors Syed Arfath Ahmed1, R.K. Agrawal2, Neelam Labhade Kumar3, V.S. Narayana Tinnaluri4, Geogen George5 Maulana Azad National Urdu University, India1, SNJB K B Jain College of Engineering, India2, Shree Ramchandra College of Engineering, India3, Koneru Lakshmaiah Educational Foundation, India4, University of Technology and Applied Sciences, Sultanate of Oman5 Keywords Hardware Security, FPGA Implementation, AI Model Security, Field-Programmable Gate Array, Adversarial Attacks Yearly Full Views January February March April May June July August September October November December 14 2 3 7 3 0 1 0 0 1 0 0

Published By : ICTACT Published In : ICTACT Journal on Microelectronics ( Volume: 9 , Issue: 4 , Pages: 1663 - 1669 ) Date of Publication : January 2024 DOI : 10.21917/ijme.2024.0288 Page Views : 325 Full Text Views : 31