The “Backend Duplication” Method

Abstract

Several types of logic gates suitable for leakage-proof computations have been put forward [1,2,3,4]. This paper describes a method, called "backend duplication" to assemble secured gates into leakage-proof cryptoprocessors. To the authors' knowledge, this article is the first CADoriented publication to address all the aspects involved in the backend design of secured hardware. The "backend duplication" method achieves the place-and-route of differential netlists. It allows for 100 % placement density and for balanced routing of dual-rail signals. Wires of every other metal layer are free to make turns. In addition, the method does not require any modification to the design rules passed to the router. The "backend duplication" method has been implemented in 0.13 µm ASIC technology and successfully tested on various ciphers. The example of the design of a DES module resistant against side-channel attacks is described into details.

Security Evaluation of DPA Countermeasures Using Dual-Rail Pre-charge Logic Style

“…Therefore, incidental leakage can be likely to improve with the place-and-route in the manual operation or the semi-automatic operation using the special constraints. Actually, Tiri et al and Guilley et al proposed "Fat Wire" [7] and "Backend Duplication" [11], respectively as a countermeasure in the place-and-route to improve the DPA-resistance.…”

Divided Backend Duplication Methodology for Balanced Dual Rail Routing

“…Backend Duplication was proposed by Guilley et al [2] to address routing imbalances in DRP logic styles. The basic idea of backend duplication is based on placement and routing obstructions (constraints to the CAD tool).…”

“…The basic idea of backend duplication is based on placement and routing obstructions (constraints to the CAD tool). The first step of Backend Duplication is to constrain the CAD tool (1) to only use alternate rows for placing cells and routing horizontal routes (2) and to use the alternate routing pitches for routing vertical routes. Thus, when the placer has finished placing the single rail design, a dual rail design can be obtained from copying (and transforming) the single rail into the previously obstructed rows.…”

“…5 shows the overview of our proposed method for balanced dual rail routing. This method is similar to the Backend Duplication method, [2]. A single ended design is used for the initial place and route process and then duplicated to get the final dual rail design.…”

“…To address the routing problem, to date the following proposals have been put forward: DWDDL [3], FatWire [1], Backend Duplication [2], Three Phase Dual Rail [13], Path Switching [9], Double WDDL [14] and an iterative correction flow [15]. Of these, three proposals [1,2,3] impose some constraints on backend implementation flows. Three Phase Dual Rail [13] tries to avoid the routing problem by introducing a third phase, which is an additional overhead.…”