NuRedact: Non-Uniform eFPGA Architecture for Low-Overhead and Secure IP Redaction

TL;DR

NuRedact introduces a non-uniform eFPGA architecture that significantly reduces area overhead and enhances security for IP redaction, balancing efficiency and robustness against advanced attack models.

Contribution

It presents the first full-custom non-uniform eFPGA framework for IP redaction, improving efficiency and security over traditional uniform architectures.

Findings

Up to 9x area reduction compared to uniform fabrics

Achieves efficiency comparable to LUT-based and transistor-level techniques

Demonstrates enhanced resilience against SAT, cyclic, and sequential attacks

Abstract

While logic locking has been extensively studied as a countermeasure against integrated circuit (IC) supply chain threats, recent research has shifted toward reconfigurable-based redaction techniques, e.g., LUT- and eFPGA-based schemes. While these approaches raise the bar against attacks, they incur substantial overhead, much of which arises not from genuine functional reconfigurability need, but from artificial complexity intended solely to frustrate reverse engineering (RE). As a result, fabrics are often underutilized, and security is achieved at disproportionate cost. This paper introduces NuRedact, the first full-custom eFPGA redaction framework that embraces architectural non-uniformity to balance security and efficiency. Built as an extension of the widely adopted OpenFPGA infrastructure, NuRedact introduces a three-stage methodology: (i) custom fabric generation with pin-mapping irregularity, (ii) VPR-level modifications to enable non-uniform placement guided by an automated Python-based optimizer, and (iii) redaction-aware reconfiguration and mapping of target IP modules. Experimental results show up to 9x area reduction compared to conventional uniform fabrics, achieving competitive efficiency with LUT-based and even transistor-level redaction techniques while retaining strong resilience. From a security perspective, NuRedact fabrics are evaluated against state-of-the-art attack models, including SAT-based, cyclic, and sequential variants, and show enhanced resilience while maintaining practical design overheads.