IC SynthLogo: A Synthetic Logo Image Dataset for Counterfeit and Recycled IC detection

• Main Authors: Mallaiyan Sathiaseelan, Mukhil Azhagan, Vutukuru, Manoj Yasaswi, Pandurangi, Suryaprakash Vasudev, Taheri, Shayan, Asadizanjani, Navid
• Format: Conference Proceeding
• Language: English
• Subjects: AutoBom; counterfeit detection; Deep Learning; Hardware Security; Image color analysis; Integrated circuits; Logo; Logo detection; Machine learning algorithms; Neural networks; PCB Assurance; Printed circuits; Quality assurance; Reliability; Synthetic images; Text recognition
• ISSN: 1946-1550
• DOI: 10.1109/IPFA53173.2021.9617352

We present and describe a novel and completely automatic method to create synthetic logos images of Integrated Circuits (ICs) for Printed Circuit Boards (PCB) assurance. Our method employs a hybrid of existing and newly developed techniques built specifically for the textures and colors in IC and PCB Images. This method can be utilized for reliability and quality assurance, evaluating counterfeit detection algorithms, and also to aid in PCB Marking detection as logos are prime false positives in Tesseract, EAST and CRAFT algorithms (popular text recognition algorithms in the vision community). In addition, we also make the script available to the community for creating the database of the images, along with their bounding box annotations to train Logo identification neural networks and machine learning algorithms. The data, along with the script for extending to further logos can be found at: https://github.com/mukhilazhagan/ic_synth_logo_code and https://www.trust-hub.org/#/data