{"id":2402,"date":"2022-07-27T12:39:23","date_gmt":"2022-07-27T12:39:23","guid":{"rendered":"http:\/\/cad4security.org\/?page_id=2402"},"modified":"2022-08-01T14:20:30","modified_gmt":"2022-08-01T14:20:30","slug":"supply-chain-security-track","status":"publish","type":"page","link":"http:\/\/cad4security.org\/index.php\/contests\/host-2022-microelectronics-security-challenge\/supply-chain-security-track\/","title":{"rendered":"Supply Chain Security Track"},"content":{"rendered":"\n<p class=\"is-style-subheading\"><strong>TRACK CO-CHAIR: NAVID ASADI<\/strong><\/p>\n\n\n\n<h2 class=\"is-style-subheading wp-block-heading\">Overview<\/h2>\n\n\n\n<p>The lack of traceability in the globalized electronics supply chain results in the infiltration of various counterfeit electronic parts, including recycled, remarked, overproduced, cloned, out-of-spec\/defective, forged documentation, and tampered types and pose a severe threat to the security of our critical infrastructures. Among them, recycled, remarked, and cloned parts constitute most counterfeit incidents. Over the years, a class of solutions has been proposed to mitigate the widespread infiltration of these fake parts. Physical Inspection methods have gained a lot of attention due to their one-size-fits-all nature as the same methods can be applied to all types of parts (analog, digital, memory, FPGAs. etc.). Among various modalities, including optical, X-ray, thermal, electron beam microscopy, etc., optical imaging is one of the fastest and most affordable modalities. <\/p>\n\n\n\n<p>&nbsp;This challenge requires the competitors to develop a highly-accurate automated method to identify counterfeit ICs from their labeled optical images. The minimum accuracy requirement is at least 60%. Detailed requirements are given in the following table.&nbsp;<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table><thead><tr><th>&nbsp;<strong>Accuracy Range&nbsp;<\/strong><\/th><th><strong>Competition Ranking Categories&nbsp;<\/strong><\/th><\/tr><\/thead><tbody><tr><td>90%-100%&nbsp;<\/td><td>Gold&nbsp;<\/td><\/tr><tr><td>More than or equal to 80%, but less than 90%&nbsp;<\/td><td>Silver&nbsp;<\/td><\/tr><tr><td>More than or equal to 70%, but less than 80%&nbsp;<\/td><td>Bronze&nbsp;<\/td><\/tr><tr><td>More than or equal to 60%, but less than 70%&nbsp;<\/td><td>Qualified, but no specified ranking&nbsp;<\/td><\/tr><tr><td>Less than 60%&nbsp;<\/td><td>Disqualified&nbsp;<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h2 class=\"is-style-default wp-block-heading\"><strong>Dataset Overview&nbsp;<\/strong><\/h2>\n\n\n\n<p>The image data presented here is collected using the FICS lab facilities with two different image acquisition modalities:&nbsp;<\/p>\n\n\n\n<ul><li>DSLR system with color normalization,&nbsp;<\/li><li>Zeiss Stemi 508 Stereo Microscope.&nbsp;<\/li><\/ul>\n\n\n\n<h3 class=\"is-style-subheading has-lg-font-size wp-block-heading\"><strong>Training Data<\/strong><\/h3>\n\n\n\n<p>The challenge package includes a folder called <strong>\u2018train\u2019, <\/strong>which contains two subfolders named <strong>\u2018DSLR\u2019 <\/strong>and <strong>\u2018STEMI_508\u2019. <\/strong>Inside the <strong>\u2018DSLR\u2019 <\/strong>folder, there are <strong>40 <\/strong>high-resolution images (both <strong>authentic <\/strong>and <strong>counterfeit<\/strong>) of the front and the back surface of <strong>3 <\/strong>different types of ICs. On the other hand, the <strong>\u2018STEMI_508\u2019 <\/strong>folder includes <strong>60 <\/strong>high-resolution images (both <strong>authentic <\/strong>and <strong>counterfeit<\/strong>) of the front and the back surface of the same <strong>3 <\/strong>types of ICs mentioned for DSLR. The <strong>\u2018train\u2019 <\/strong>folder also contains an annotation file called <strong>\u2018train.csv\u2019, <\/strong>which has <strong>8 <\/strong>columns as mentioned below:&nbsp;<\/p>\n\n\n\n<ul><li><strong>id <\/strong>&#8211; Unique identifier for each sample image. For example, the id <strong>A-M-16DIP-00F-D <\/strong>has 5 portions which contain 5 information of the sample:&nbsp;<ul><li><strong>A \u2013 <\/strong>Authentic&nbsp;<\/li><li><strong>M \u2013 <\/strong>Mouser Electronics (Vendor Acronym)<\/li><li><strong>16DIP \u2013 <\/strong>16 pins dual in-line package (Package type)<\/li><li><strong>00F \u2013 <\/strong>00 refers to the first image, i.e., serial number, and F refers to the Front side image<\/li><li><strong>D \u2013 <\/strong>DSLR (image acquisition modality)&nbsp;<\/li><\/ul><\/li><li><strong>sample_name \u2013 <\/strong>Manufacturer Product Number. For example, <strong>STM32F105R8T7&nbsp;<\/strong><\/li><li><strong>manufacturer &#8211; <\/strong>Manufacturing company\u2019s name. For example, <strong>STMicroelectronics&nbsp;<\/strong><\/li><li><strong>vendor \u2013 <\/strong>Name of the entity that supplies the product (IC). For example, <strong>Digi-Key Electronics&nbsp;<\/strong><\/li><li><strong>product_type \u2013 <\/strong>Type of the IC. For example, <strong>DAC <\/strong>(Digital-to-Analog converter) or <strong>Embedded \u2013 Microcontrollers&nbsp;<\/strong><\/li><li><strong>package_type \u2013 <\/strong>Type of the IC package. For example, <strong>16DIP <\/strong>(16 pins dual in-line package)<\/li><li><strong>modality \u2013 <\/strong>Type of image acquisition modality. For example, <strong>DSLR <\/strong>or <strong>Stereo Microscope&nbsp;<\/strong><\/li><li><strong>label \u2013 0 <\/strong>(authentic) or <strong>1 <\/strong>(counterfeit)&nbsp;<\/li><\/ul>\n\n\n\n<h3 class=\"is-style-subheading has-lg-font-size wp-block-heading\"><strong>Test Data<\/strong><\/h3>\n\n\n\n<p>The challenge package includes another folder called <strong>\u2018test\u2019, <\/strong>which contains two subfolders named <strong>\u2018DSLR\u2019 <\/strong>and <strong>\u2018STEMI_508\u2019<\/strong>. Inside the \u2018DSLR\u2019 folder, there are <strong>10 <\/strong>high-resolution images (both <strong>authentic <\/strong>and <strong>counterfeit<\/strong>) of the front and the back surface of a single type (different from training data) of ICs. On the other hand, the <strong>\u2018STEMI_508\u2019 <\/strong>folder includes <strong>10 <\/strong>high-resolution images (both <strong>authentic <\/strong>and <strong>counterfeit<\/strong>) of the front and the back surface of the same type of IC mentioned for DSLR. The <strong>\u2018test\u2019 <\/strong>folder also contains an annotation file called <strong>\u2018test.csv\u2019<\/strong>, with the same column information mentioned in the training data.&nbsp;<\/p>\n\n\n\n<h3 class=\"is-style-subheading has-lg-font-size wp-block-heading\"><strong>Submission Criteria<\/strong><\/h3>\n\n\n\n<ul><li>The participants are required to submit a zip file containing codes, sample submission, demo, and presentation. The file name should be &#8220;team_name_HOST_2022_SCS.zip&#8221; using <a rel=\"noreferrer noopener\" href=\"https:\/\/tigermailauburn-my.sharepoint.com\/personal\/uzg0005_auburn_edu\/_layouts\/15\/onedrive.aspx?p=26&amp;s=aHR0cHM6Ly90aWdlcm1haWxhdWJ1cm4tbXkuc2hhcmVwb2ludC5jb20vOmY6L2cvcGVyc29uYWwvdXpnMDAwNV9hdWJ1cm5fZWR1L0VrSXVxaWJHcW5oQW1nd1g1ZFVmcGhzQnp5MVVIOXJxYTFSLVo1RTY0ZU80QXc\" target=\"_blank\">OneDrive<\/a>.&nbsp;<\/li><li><strong>Code: <\/strong>Can be in any language, i.e., python (recommended) \/R\/Java\/C++ etc.<ul><li>A GitHub repository containing all necessary codes\/libraries\/helper functions to train, test, and visualize with a complete and comprehensive README file to implement on the hold-out test data.&nbsp;<\/li><li>If the README file and Demo video (see below) don\u2019t work\/cannot help in successful implementation as claimed in the presentation (see below) and demo (mentioned later), the competitor will receive a penalty.&nbsp;<\/li><\/ul><\/li><li><strong>Sample submission: <\/strong>a \u2018sample_submission.csv\u2019 file containing two columns \u2013<ul><li>id &#8211; Unique identifier for each sample test image as in \u2018test.csv\u2019&nbsp;<\/li><li>predicted_label \u2013 0 (if the prediction is \u2018authentic\u2019 for a certain sample test image), 1 (if the prediction is \u2018counterfeit\u2019 for a certain sample test image)&nbsp;<\/li><\/ul><\/li><li><strong>Demo: <\/strong>a demo video link (upload in <a href=\"https:\/\/tigermailauburn-my.sharepoint.com\/personal\/uzg0005_auburn_edu\/_layouts\/15\/onedrive.aspx?p=26&amp;s=aHR0cHM6Ly90aWdlcm1haWxhdWJ1cm4tbXkuc2hhcmVwb2ludC5jb20vOmY6L2cvcGVyc29uYWwvdXpnMDAwNV9hdWJ1cm5fZWR1L0VrSXVxaWJHcW5oQW1nd1g1ZFVmcGhzQnp5MVVIOXJxYTFSLVo1RTY0ZU80QXc\" target=\"_blank\" rel=\"noreferrer noopener\">OneDrive<\/a>) to demonstrate how to run the developed system (train, test, and visualize).<ul><li>Time limit: minimum &#8211; 5 mins, maximum &#8211; 30 mins, recommended \u2013 15 mins&nbsp;<\/li><\/ul><\/li><li><strong>Presentation: <\/strong>A PowerPoint presentation of a <strong>minimum 5 to maximum 15 slides <\/strong>in format to show the outcome (must include the method, results, and discussion).&nbsp;<\/li><\/ul>\n\n\n\n<h3 class=\"is-style-subheading wp-block-heading\"><strong>Evaluation Criteria***<\/strong><\/h3>\n\n\n\n<p><strong>Total: 100 pts&nbsp;<\/strong><\/p>\n\n\n\n<ul><li><strong>Training Strategy (e.g., Cross-validation, Data augmentation, Pre-processing, etc.) &#8211; 20 pts\u00a0<\/strong><\/li><li><strong>Model Complexity \u2013 10 pts\u00a0<\/strong><\/li><li><strong>Model Performance (Accuracy*, Confusion Matrix, Precision, Recall, F-Score, ROC AUC Score, The Matthews Correlation Coefficient (MCC)<\/strong> <strong>) \u2013 30 pts<\/strong><ul><li>Highly accurate on <strong>only DSLR <\/strong>images \u2013 <strong>10 pts\u00a0<\/strong><\/li><li>Highly accurate on <strong>only Stereo Microscope <\/strong>images \u2013 <strong>10 pts\u00a0<\/strong><\/li><li>Highly accurate on <strong>both DSLR and Stereo Microscope <\/strong>images \u2013 <strong>30 pts\u00a0<\/strong><\/li><\/ul><\/li><li><strong>Model\u2019s Generalizability\/Robustness &#8211; 10 pts<\/strong> (To clarify, it will be examined how consistent the model\u2019s performance is over different dataset distribution and\/or adversarial examples)\u00a0<\/li><li><strong>Inference time\/Computational Cost \u2013 20 pts\u00a0<\/strong><\/li><li><strong>Model\u2019s scalability \u2013 10 pts\u00a0<\/strong><\/li><li><strong>Model\u2019s explainability** &#8211; 10 pts (Bonus)\u00a0<\/strong><\/li><\/ul>\n\n\n\n<p class=\"has-sm-font-size\"><strong>* Mandatory&nbsp;<\/strong><br><strong>**Optional&nbsp;<\/strong><br><strong>***Tentative&nbsp;<\/strong><\/p>\n\n\n\n\n\n\n\n<h3 class=\"is-style-subheading wp-block-heading\"><strong>Some external resources for more clarification<\/strong><\/h3>\n\n\n\n<p><strong>More about counterfeit IC detection:&nbsp;<\/strong><\/p>\n\n\n\n<ul><li><a href=\"https:\/\/link.springer.com\/chapter\/10.1007\/978-3-030-62609-9_2\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/link.springer.com\/chapter\/10.1007\/978-3-030-62609-9_2&nbsp;<\/a><\/li><\/ul>\n\n\n\n<p><strong>Model\u2019s Performance Metrics: <\/strong><\/p>\n\n\n\n<ul><li><a href=\"https:\/\/neptune.ai\/blog\/evaluation-metrics-binary-classification\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/neptune.ai\/blog\/evaluation-metrics-binary-classification&nbsp;<\/a><\/li><\/ul>\n\n\n\n<p><strong>Model\u2019s Generalizability\/Robustness :&nbsp;<\/strong><\/p>\n\n\n\n<ul><li><a href=\"https:\/\/ieeexplore.ieee.org\/stamp\/stamp.jsp?tp=&amp;arnumber=7780854&amp;tag=1\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/ieeexplore.ieee.org\/stamp\/stamp.jsp?tp=&amp;arnumber=7780854&amp;tag=1&nbsp;<\/a><\/li><li><a href=\"https:\/\/proceedings.neurips.cc\/paper\/2020\/hash\/61d77652c97ef636343742fc3dcf3ba9-Abstract.html\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/proceedings.neurips.cc\/paper\/2020\/hash\/61d77652c97ef636343742fc3dcf3ba9-Abstract.html&nbsp;<\/a><\/li><\/ul>\n\n\n\n<p><strong>Model\u2019s scalability:&nbsp;<\/strong><\/p>\n\n\n\n<ul><li><a href=\"https:\/\/www.codementor.io\/blog\/scaling-ml-6ruo1wykxf\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/www.codementor.io\/blog\/scaling-ml-6ruo1wykxf&nbsp;<\/a><\/li><li><a href=\"https:\/\/www.codementor.io\/blog\/scalable-ml-models-6rvtbf8dsd\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/www.codementor.io\/blog\/scalable-ml-models-6rvtbf8dsd&nbsp;<\/a><\/li><li><a href=\"https:\/\/neptune.ai\/blog\/how-to-scale-ml-projects\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/neptune.ai\/blog\/how-to-scale-ml-projects&nbsp;<\/a><\/li><\/ul>\n\n\n\n<p><strong>Model\u2019s explainability:&nbsp;<\/strong><\/p>\n\n\n\n<ul><li><a href=\"https:\/\/neptune.ai\/blog\/explainability-auditability-ml-definitions-techniques-tools\" target=\"_blank\" rel=\"noreferrer noopener\">https:\/\/neptune.ai\/blog\/explainability-auditability-ml-definitions-techniques-tools&nbsp;<\/a><\/li><\/ul>\n","protected":false},"excerpt":{"rendered":"<p>TRACK CO-CHAIR: NAVID ASADI Overview The lack of traceability in the globalized electronics supply chain results in the infiltration of various counterfeit electronic parts, including recycled, remarked, overproduced, cloned, out-of-spec\/defective, forged documentation, and tampered types and pose a severe threat to the security of our critical infrastructures. Among them, recycled, remarked, and cloned parts constitute &hellip;<\/p>\n","protected":false},"author":8,"featured_media":2338,"parent":1405,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"_genesis_hide_title":false,"_genesis_hide_breadcrumbs":false,"_genesis_hide_singular_image":false,"_genesis_hide_footer_widgets":false,"_genesis_custom_body_class":"","_genesis_custom_post_class":"","_genesis_layout":"","footnotes":""},"acf":[],"_links":{"self":[{"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/pages\/2402"}],"collection":[{"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/comments?post=2402"}],"version-history":[{"count":3,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/pages\/2402\/revisions"}],"predecessor-version":[{"id":2425,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/pages\/2402\/revisions\/2425"}],"up":[{"embeddable":true,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/pages\/1405"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/media\/2338"}],"wp:attachment":[{"href":"http:\/\/cad4security.org\/index.php\/wp-json\/wp\/v2\/media?parent=2402"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}