Shield

This project empowers users with a solid layer of security against the growing number of Crypto scams. Our PWA offers users a suit where they can control and monitoring their exposure to potential scam or malicious actors. Shield will work hard for them in the background with its machine learning algorithms to securely analyze texts they are reading. Shield aggregates all the incoming visual input the user is seeing in their monitor and passes it into a pipeline specialized in scam analysis, to score the severity of potential harmful impact. Our users, when encountering a scam, get promptly informed as well as rewarded with SAT tokens for participating in proof of scam methodology. This way we encourage our users to participate in Shield's ecosystem and spread awareness of pitfalls in the industry.

This project is heavily using custom made super-secure AI models. We fine-tuned our main model using Hyperbolic GPU's rental and we used a dataset with the latest crypto scams. The first ML algorithm is a custom implementation of an OCR, aiming at parsing the content of what the user is viewing to make sense of it and optically read it. Then, our main machine learning algorithm works continuously in the background to evaluate incoming data being pushed into the pipeline, and isolate suspicious behavior. The first model is secure by definition, running locally on the user's machine. While the second (custom LLM backbone with a classification layer attached to its final neurons), with the help of Phala network and their TEEs, is hosted on a secure interface and provides a solid scam and fraud analysis layer, monitoring real time and promptly warning the user if something suspicious is reported. This one is secure by proof.

We also provided a DeFi rewards suite by customizing an erc20 SAT token which is responsible for rewards. Whenever an authenticated user catches a threat in our backend, by using Coinbase GCP SDK we found a way how to call our smart contract reward payout function. Our contract was deployed on Base Sepolia network.

Last, but not least, we followed the example of DevCon and developed a modern PWA to allow users track scam detections and ongoing progress of rewards.