ShadowML

ShadowML

ML model marketplace that provide predictions but maintains the privacy of ML model datasets

Created on 14th February 2025

ShadowML

ShadowML

ML model marketplace that provide predictions but maintains the privacy of ML model datasets

The problem ShadowML solves

In today's digital era, machine learning has become an essential tool for solving complex problems and making data-driven decisions. However, using ML models for sensitive applications can lead to privacy concerns and the potential for data leakage. This becomes especially challenging when the ML model provider needs to offer predictions without disclosing their proprietary model's critical attributes and thresholds. Traditional approaches for sharing ML models expose these sensitive details, making it difficult to strike a balance between functionality and privacy.
So your marktplace helps to keep the ML Model private and the dataset it has been trained . It helps the provider to earn money through per request fees he wants to set that keeps it Model private and other users can get predictions without devloping their own ML model by paying a fees to the provider and verifies the proof through zkVerify.
It can be employed in a wide range of applications and industries that require secure and privacy-preserving predictions. Some examples include:
Healthcare: Safeguard patient privacy while enabling healthcare providers to make data-driven decisions based on ML models. It ensures that personal health information (PHI) and model details are kept confidential, improving patient trust and compliance with data protection regulations.
Finance: Enhance security and compliance for financial institutions by enabling them to utilize ML models for tasks like credit scoring, fraud detection, and portfolio management, without disclosing sensitive customer data or proprietary model information.
Human Resources: Streamline the recruitment process by using It to analyze candidate profiles without exposing personal information or revealing details of the underlying ML model, thereby preserving privacy and reducing potential bias.
Marketing: Leverage customer data for targeted advertising and personalization without compromising user privacy or revealing proprietary ML

Challenges we ran into

We as a team had a very less knowledge about AI and ML models that's why there is only one example of the model . But it can easily be extended to various usecases that we had talked in the video.
We had also faced problem with the Arbitrum RPC URL because we were getting that the

from_block

is greater than

to_block

but we found a way to solve this issue.

Tracks Applied (3)

Arbitrum Web3 dApp with zkVerify

Deployed on Arbitrum, our ShadowML on-chain marketplace utilizes Arbitrum's scalable Layer 2 infrastructure to provide a...Read More
Arbitrum

Arbitrum

DeSci - Decentralized Science Web3 Application with zkVerify

Our project—ShadowML—fits seamlessly into the Open Campus EDU Chain’s DeSci vision and the zkVerify track by bringing de...Read More

Open Campus EDU Chain

zkVerify Main Prize Track

We had used zkVerify Contracts to verify the proof generated by the RISC0 ZKVM and Deployed all the contracts on the Arb...Read More

Discussion

Builders also viewed

See more projects on Devfolio