LogoTrust: Leveraging BIMI to Build a Validated Dataset of Brands, Domain Names, and Logos

Most brand logo datasets are static, manually labeled, and solely rely on the trust in the dataset creator. We leverage Brand Indicators for Message Identification (BIMI) to build a dynamic, automatically generated logo dataset with trust anchored in Certificate Authorities (CAs). The BIMI DNS TXT records, in which the domain owners publish logos with Mark Certificates (MCs), link brand logos to domain names, streamlining the collection and the verification of the mapping while potentially enhancing logo-based phishing detection methods. Using global-scale DNS measurements, we collect BIMI records, retrieve and deduplicate logos, and validate Mark Certificates to obtain the mapping of 1,680 brands, 1,811 logos, and 2,821 domain names. Our dataset, available at https://github.com/josef0x/LogoTrust, demonstrates the BIMI potential as a reliable means for constructing logo datasets and potentially provides a valuable resource for applications such as phishing detection.