Protocols

The Nillion Network Blind Modules -- and the available SDKs that can be used to interoperate with them -- make it possible to leverage a number of secure data storage and computation protocols. These include what are broadly known as privacy-enhancing technologies (PETs).

Trust Assumptions

The Nillion Network consists of clusters of nodes, and builders have the option to rely on a subset of one or more of these nodes to securely store and compute over data. When working with a single node in conjunction with a PET such as homomorphic encryption, it is up to the builder to determine how secret and public keys are generated, maintained, and used. When working with a cluster of two or more nodes, builders can choose to rely on the non-collusion of network nodes (and may not even need to inform any individual node of the identities of other nodes in the cluster).

Privacy-Enhancing Technologies

The Blind Modules leverage a number of PETs to enable secure data storage and processing.

Secure Multi-Party Computation (MPC)

nilDB supports the use of additive secret sharing to store data and to compute over that data. Builders can choose two or more nodes across which data would be stored using this approach.

nilVM supports the use of a number of MPC protocols, including threshold secret sharing schemes. Most notably, it relies on an integrated implementation of CGGMP21 to enable threshold secure signing of messages.

Homomorphic Encryption (HE)

nilDB supports of the use of the Paillier cryptosystem to store data, compute over that data, and retrieve results while relying on a single-node cluster. This makes it possible to perform aggregation queries over encrypted data even when using a single-node cluster.

Trusted Execution Environments (TEEs)

Private LLM inference and the use of private RAG is supported via TEEs.