Global similarity tests of physical designs of circuits: A complex network approach

Abstract

Similarity testing for circuits is an important task in the identification of possible infringement of intellectual property rights. In this paper, we propose a novel procedure for global similarity measurement between circuit topologies (networks) and apply this procedure to the comparison of physical designs of circuits. We first construct networks to describe the way in which circuit elements interact. Then, we evaluate the properties of each node from the resulting networks by calculating the cumulative distribution of characteristic parameters such as degree, clustering coefficient, etc. Based on the maximum vertical distance of each pair of distributions, global similarity testing methods are proposed with consideration of the inhomogeneity of parameter distributions and the scale of the networks. Simulation results show the effectiveness of the strategy in terms of robustness and topological information mining. The methodology described here can be applied to the identification of physical designs of circuits that may contain suspected patent infringement, and it is suitable for a wide range of circuits and systems.

Introduction

Intellectual property disputes, such as legal proceedings related to alleged patent infringements, have attracted wide media attention in recent years as they often involve remarkable commercial and public interests. In most cases, allegations are often hard to establish due to the lack of convincing evidence that supports the alleged violations, and some lawsuits could be further complicated by the diversity of the detailed situations of usages and the interpretations of patent coverage. In this paper, we are interested in finding the distinct features in physical designs of electronic circuits in terms of quantitative measures rather than the visual appearance of circuit layouts.

Recent research in network science has aroused a lot of interest across a multitude of application areas [1], [2], [3], [4]. There are numerous physical and engineering systems whose structures naturally permit direct application of complex networks as the modeling platform [5], [6], [7]. For example, an electronic circuit can be modeled directly as a network which has a structure consisting of nodes connected by edges. The ubiquity of complex networks in science and technology has naturally led to a set of important research problems, such as similarity testing which attempts to tell whether two networks are alike, or/and being generated from the same underlying source or mechanism. Numerous methods have been proposed to identify equivalence classes in or between networks [8], [9], [10], [11], but the important practical problem of testing global similarity has been rarely considered.

In this paper, we propose to use a complex network approach to address the problem of similarity measurement, in quantitative terms, between physical designs of electronic circuits. Unlike previous studies in the characterization of complex networks where the average characteristic parameter over all nodes is considered, we evaluate the properties of each node in order to describe the internal differences between two physical designs of a circuit. Based on the properties of each node, we propose a novel approach for global similarity testing. Sections 2 Network construction, 3 Network analysis explain some basic concepts and provide the construction and analysis of the network model. Section 4 presents the global similarity testing strategy and experimental results. Section 5 gives the conclusion.