challenge & innovation

The client needed an efficient way to test and evaluate the accuracy of a Controlled Unclassified Information (CUI) marking tool but faced strict restrictions on using actual sensitive documents for testing. This limitation created challenges in assessing the tool’s effectiveness without risking exposure of real data. To overcome this, Savan leveraged Generative AI and Natural Language Processing (NLP) to generate on-demand synthetic documents embedded with fake sensitive information elements. This innovative approach enabled realistic and controlled testing, ensuring the tool’s accuracy while maintaining compliance with security protocols.

impact

By generating synthetic documents, Savan enabled a rapid and accurate assessment of the CUI marking tool’s performance, allowing for thorough testing without compromising security. This approach improved the tool’s accuracy, ensuring more reliable identification and marking of sensitive information while streamlining the evaluation process.