The problem LitiBlocks solves

Insurance fraud is a significant challenge, costing companies billions annually and straining the overall insurance ecosystem. Detecting fraud often involves manual reviews, scattered data, and inconsistent patterns across organizations, leading to inefficiencies and missed fraudulent cases.

This system addresses these issues by:

• Automating Fraud Detection: Using AI to analyze legal documents for potential fraud, significantly reducing manual effort.
• Preserving Data Privacy: Ensuring companies' sensitive data remains secure through federated learning and ZK-SNARKs, where raw data never leaves the company’s system.
• Collaborative Intelligence: Aggregating insights from multiple companies without compromising privacy to detect broader fraud patterns that individual companies might miss.
• Rewarding Data Quality: Implementing tokenomics to incentivize companies to contribute high-quality data, improving the global model’s accuracy.
• Scalable and Efficient Processes: Providing tools for batch processing of documents, making the system scalable for companies of all sizes.
  By addressing these pain points, the solution not only improves fraud detection accuracy but also fosters trust and collaboration among insurers while ensuring robust data security.

Challenges we ran into

• Implementing ZK-SNARKs: Incorporating ZK-SNARKs into the system to ensure data privacy was a significant challenge. Understanding the cryptographic principles, designing proofs for secure parameter sharing, and integrating them with the federated learning process required meticulous planning and iterative testing. Balancing efficiency and security was particularly demanding.
• Integrating Smart Contracts: Developing smart contracts in Move for the Sui blockchain posed its own set of hurdles. The complexity of managing tokenomics, ensuring accurate incentive distribution, and maintaining seamless communication between AI systems and the blockchain required a deep understanding of both blockchain mechanics and smart contract design. Debugging and optimizing the contracts for real-world use cases added another layer of difficulty.
• Blending AI with Blockchain: Creating a system where AI models, federated learning, and blockchain technology work harmoniously was no easy task. Ensuring that the AI could securely interact with blockchain smart contracts without compromising performance or accuracy required innovative architectural solutions. Synchronizing the updates between the small and global models while integrating blockchain rewards added to the complexity.
  These challenges tested our technical expertise and forced us to think creatively, but overcoming them ultimately resulted in a robust, secure, and innovative solution.