LightleapAI Fraud Detection V1.2 Release
About a year ago, LightLeapAI was deployed at Foothill-De Anza Community College District around this time. Since then, we have been implementing the standard Admissions (m1.0) model that was originally released with remarkable success. In fact, this model is now in production at over 50 colleges across the nation and has processed about 5 million admissions applications and flagged almost 900,000 of them as potentially fraudulent. This success led us to the Statewide implementation of LightLeapAI, which was approved by the CCC Chancellor’s office.
While m1.0 has been extremely successful over the past few months, we have been constantly monitoring for potential weaknesses and areas for improvement by collaborating closely with our implementation partners. Because we believe that fraudsters will try to evolve their attack plans and invasion approaches as we fortify our fraud elimination efforts. As a result of the new CCC Collaboration and the expanded cohorts, we have identified a number of new patterns and inputs that would improve our fraud detection effectiveness.
I am pleased to report a significant enhancement to the Standard Admissions model (m1.2). This model is built by integrating some of the findings from the fraud clusters, false positives, and false negatives that were reported over the last few months. This model will be available to all LightLeapAI customers by 5 pm PT on Wednesday, July 23rd, 2025. While this model has shown significant improvement in reducing ‘false negatives’ (fraudsters who were incorrectly tagged as ‘not fraud’ by the model), the new model can have more ‘false positives’ as well. We will be addressing this by providing free ID verification tools and automatic clearance for any applicant that we mark as fraudulent.
Here are some enhancements of LightLeapAI’s patent-pending fraud detection framework:
Integration of Global and Internal Fraud Clusters into the Standard Model
Integration of AI-Powered ID verification tools and related AI Agents into the standard dashboards
API enhancements to support real-time and asynchronous fraud elimination
Fall back models to counteract evolving attack plans of the bots (ex: if a feature is disproportionately impacting the final score, the fallback model will suppress the weight of this feature to identify additional threat patterns).
In the coming days, we will share reports of new frauds we identified with the m1.2 model with your team. Additionally, we will provide concrete recommendations on how to clear these additional fraudsters without causing additional administrative overhead for your A&R, Admissions, IT, and Financial Staff. As part of this analysis, we would like to see any ‘false positives’ or ‘false negatives’ you have documented. Please do not hesitate to share this information with your customer success team.
Our goal is simple – to collaborate with each of you to eliminate fraud once and for all at California Community Colleges.
Please let me know if you have any questions or concerns regarding this.
--
Kiran Kodithala
N2N Services, Inc. | LightLeapAI
Founder and CEO
336.406.3631 (M) | 888.651.3309 (O)