Obfuscating the Interconnects: Low-Cost and Resilient Full-Chip Layout Camouflaging

TL;DR

This paper introduces a low-cost, generic full-chip layout camouflaging scheme based on interconnect obfuscation, which enhances security against reverse engineering and SAT attacks while maintaining acceptable overheads.

Contribution

The authors propose a novel, scalable layout camouflaging method that applies to the entire chip via BEOL interconnect obfuscation, addressing limitations of prior FEOL-centric techniques.

Findings

Significantly lower cost than previous methods

Achieves 12% power, 30% performance, 48% area overheads at 100% camouflaging

Effective against SAT attacks with large-scale camouflaging

Abstract

Layout camouflaging (LC) is a promising technique to protect chip design intellectual property (IP) from reverse engineers. Most prior art, however, cannot leverage the full potential of LC due to excessive overheads and/or their limited scope on an FEOL-centric and accordingly customized manufacturing process. If at all, most existing techniques can be reasonably applied only to selected parts of a chip---we argue that such "small-scale or custom camouflaging" will eventually be circumvented, irrespective of the underlying technique. In this work, we propose a novel LC scheme which is low-cost and generic---full-chip LC can finally be realized without any reservation. Our scheme is based on obfuscating the interconnects (BEOL); it can be readily applied to any design without modifications in the device layer (FEOL). Applied with split manufacturing in conjunction, our approach is the first in the literature to cope with both the FEOL fab and the end-user being untrustworthy. We implement and evaluate our primitives at the (DRC-clean) layout level; our scheme incurs significantly lower cost than most of the previous works. When comparing fully camouflaged to original layouts (i.e., for 100% LC), we observe on average power, performance, and area overheads of 12%, 30%, and 48%, respectively. Here we also show empirically that most existing LC techniques (as well as ours) can only provide proper resilience against powerful SAT attacks once at least 50% of the layout is camouflaged---only large-scale LC is practically secure. As indicated, our approach can deliver even 100% LC at acceptable cost. Finally, we also make our flow publicly available, enabling the community to protect their sensitive designs.