The modern world is moving towards an intelligent future where AI and ML will be playing an important role in improving the technologies across virtually all domains. The most basic requirement for training an ML model is data. Without the availability of data there is no use of AI or ML. But there are some hindrances to obtaining the data. Sometimes the data is freely and openly available, but there are some case where the privacy of the data is much more important that creating an ML model out of it.
Such scenarios can prove to be a hurdle in developing new technologies. So we suggest a way which can be used to train ML models on such private data while maintaining privacy of both the data as well as the ML model, using homomorphic encryption. We used a somewhat homomorphic encryption system proposed by Zhou and Wornell.
You can check out the report attached for more details.
Created By: Piyush and Vishal Sharma
Credits : i am trask