Abstract
Plagiarism of integrated-circuit (IC) layout is a problem encountered both in academia and in industry. A procedure was proposed that compares IC layouts based on the physical representation of particular electrical nets, i.e., on the shape of the features drawn on conducting layers (metals and polysilicon). At the heart of this method is the Needleman–Wunsch algorithm, used for decades in tools aligning sequences of amino acids or nucleotides. Here, it is used to quantify the visual similarity of nets within the pair of layouts being compared. The method was implemented in Python and successfully used to identify clusters of similar layouts within two pools of designs: one composed of logic gates and one containing operational transconductance amplifiers.
Highlights
Software tools are used to detect plagiarism in many fields
Much less work seems to have been dedicated to creating equivalent systems for comparing integrated circuit (IC) layouts
A methodology for IC layout comparison is proposed in [5], while software based on this methodology is described in [6,7,8]
Summary
Software tools are used to detect plagiarism in many fields. The most mature and widespread is software analyzing natural language in the search of plagiarism in student essays, theses, and other bodies of text [1]. The procedures presented there are mostly based on comparing statistical parameters of layouts such as the number of metal layers used, cumulative area of shapes on each layer, or number of devices and other shapes. As such, this methodology is mostly relevant to large VLSI systems. The comparison on the level of individual cells is limited to XORing the cells’ layouts to detect shaped and placed components This approach surely works for cases of downright plagiarism of VLSI circuits, where the copied design is synthesized either using standard cells from the same provider as the original or using (nearly) exact copies of those standard cells. Cases where the layout of one circuit is just “heavily inspired” by another may go undetected
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have