Artificial intelligence and machine learning have shown a huge increase in productivity in recent years. ML is all about having good quality data while maintaining all means of privacy and confidentiality. It is very important to bridge the gap between privacy and using the advantages of machine learning to solve problems. In today’s data-driven days, protecting one’s privacy has become very difficult. With machine learning becoming so prevalent today, the implications must be considered and customer information needs to be safeguarded. New advances such as Fully Homomorphic Encryption (FHE) have successfully protected user information while maintaining confidentiality.
Zama’s machine learning researchers have introduced an open source library called Concrete-ML that enables seamless conversion of ML models into their FHE counterparts. They recently presented Concrete ML during a Google Tech Talk. Whenever some of the data belonging to the user is sent to the cloud, homomorphic encryption schemes protect that data. Operations and all actions are performed on encrypted data with data security in mind. Fully homomorphic encryption can be explained with the help of an example. Suppose a doctor wants to evaluate the descriptive analysis of patients suffering from heart problems in a particular city. The internal team at hospitals in that city that securely stores patient data in their databases may not be able to disclose the data due to privacy concerns. That is where FHE encrypts the sensitive data so that the data is secure as well as being computerized.
Concrete ML is an open source toolkit that has been built on top of The Concrete Framework. It helps researchers and data scientists to automatically convert machine learning models into their identical homomorphic units. The key feature of Concrete ML is its ability to convert ML models into their FHE equivalent without necessarily having any prior knowledge of cryptography. With Concrete ML, users can have zero-trust conversations with different service providers without hindering the implementation of ML models. Data and user privacy are maintained, and ML models are put into production even on untrusted servers.
FHE, an encryption strategy that allows direct computing on encrypted data, can be used to develop applications with unique features. FHE does not require the need for decryption. Concrete ML uses some popular Application User Interfaces (APIs) from Scikit-learn and PyTorch. The Concrete ML model has been designed as follows:
- Model Training – The model is trained on some clear data using the Scikit-learn library. Concrete ML only uses integers during inference, since FHE only works with integers.
- Conversion and compilation: In this step, the model is converted to a Concrete-Numpy program, followed by the compilation of the quantized model into an FHE equivalent.
- Inference: The inference is made on the encrypted data. During model deployment on the server, the client encrypts the data, followed by secure processing by the server and decryption by the client.
Concrete ML is a breakthrough in using machine learning with complete privacy and trust. While the only limitation Concrete ML currently has is that it can only run within the supported precision of 16-bit integers, it still sounds promising for privacy preservation.
review the GitHub link. All credit for this research goes to the researchers of this project. Also, don’t forget to join our 14k+ ML SubReddit, discord channel, and electronic newsletterwhere we share the latest AI research news, exciting AI projects, and more.
Tanya Malhotra is a final year student at the University of Petroleum and Power Studies, Dehradun, studying BTech in Computer Engineering with a specialization in Artificial Intelligence and Machine Learning.
She is a data science enthusiast with good analytical and critical thinking, along with a keen interest in acquiring new skills, leading groups, and managing work in an organized manner.
