Logo Detection and Classification Toolkit

This information was sourced from "Logo Detection and Classification Toolkit" by AI.GOV.UK, https://ai.gov.uk/knowledge-hub/tools/logo-detection-and-classification-toolkit/?directTo=library. Licensed under Open Government Licence v3.0.

The challenge

Misuse of HMRC, HMG, and Valuation Office Agency (VOA) logos by third-party websites poses a serious risk to the public, as these logos falsely imply official endorsement or affiliation. This can mislead citizens into paying for services that are otherwise free, eroding public trust and exposing users to fraud. To tackle this, the team relied on a cumbersome review process that used third-party tools to scan the internet, followed by the manual verification of identified URLs. This approach was time-consuming and struggled to keep pace with the increasing sophistication of online fraudulent activities.

The solution

Developed in-house, the tool scans the internet for websites misusing HMRC or UK Government logos and automatically flags suspicious URLs. It uses YOLOv11, a powerful open-source object detection model, to identify logos on webpages. The findings are presented through an intuitive user interface, enabling fraud analysts to quickly review and take action. Following interest from the Government Security Centre for Cyber (Cyber GSeC) in rolling out the tool as a service across other departments, functionality has been expanded. The tool now allows users to create datasets with new logos to support model training and testing, thereby helping strengthen cyber resilience across government.

The results

• Cut total process time, which included scanning and flagging sites, from 3 days to just 3-4 hours.
• Has led to 10-20 fraudulent URLs being taken down monthly.
• Increased protection of government brand integrity and public trust.

Budget: Not disclosed