An innovative platform consisting of three toolkits ensuring trust, quality and explainability to AI-based decisions
University of Wolverhampton
Professional Cyber services, Secure and Trustworthy AI
This project has developed a secure and trustworthy AI platform suitable for AI developers and data scientists, which provides a scoring mechanism to measure the quality and trust levels of datasets and AI/ML algorithm during development and deployment phases. The TrustMe platform is running based on a local-host web application with enabled features for designing, developing, and implementing explainable and trustworthy AI applications. TrustMe platform also offers a data quality score using the Quality of Data (QoD) estimator.
Using QoD feature, end users can obtain the quality level of their data along with auto-generated reports highlighting all bad records. QoD also offers an automated solution to enrich the quality of data or perform data processing according to pre-defined and editable rules by data admins. To deal with possible bias within training/testing data samples, TrustMe offers Quality of Training/Testing (QoT) toolkits. This tool will help developers to automatically generate training/testing samples by considering over/below sampling for imbalanced data samples due to the data acquisition process.
We envisage that in the next three years TrustMe Score and generated reports would become like a universal score that all organisations would want to use to prove how secure and trustworthy their AI platforms are.
Tymlo Technology Demonstrator