// Solution
The Cocos AI solution.
A distributed, microservice-based system for confidential and privacy-preserving AI/ML execution — built on Trusted Execution Environments and remote attestation.
// Secure Multi-Party Computation
Compute together, reveal nothing
Secure Multi-party Computation (SMPC) allows two or more parties to collectively perform computation and receive the resulting output without ever exposing any party's sensitive input.
With advances in machine learning, parties can be the owners of sensitive data on one hand and providers of machine learning models on the other. Cocos AI ensures that sensitive data is not exposed to the model providers and vice versa, while still delivering valuable results.
// Trusted Execution Environments
Hardware-rooted isolation
Cocos AI leverages Trusted Execution Environments (TEEs) like Intel TDX and AMD SEV to provide hardware-enforced isolation of in-use code and data.
Cryptographic keys used for code and data isolation are randomly generated and stored on the processors — never exposed to the hypervisor or operating system. This ensures that even the cloud provider cannot access the user's data or code.
// System Architecture
Multiple TEEs, one orchestrator
A robust architecture containing TEEs for code, data, and a central Security Policy Engine.
TEE_MANAGER
Microservice running on host hardware to dynamically deploy and configure TEEs for confidential computations.
IN_ENCLAVE_AGENT
Lightweight component within the enclave serving as a coordinator for execution and remote attestation.
HAL
Hardware Abstraction Layer providing a unified interface for AMD SEV, Intel TDX, and other architectures.
CLI
Command-line interface for users to configure, manage, and monitor confidential computations.
// Platform Features
Comprehensive capabilities for secure AI workloads
> SECURE_VM_PROVISIONING
Seamlessly deploy and maintain confidential virtual machines.
> HARDWARE_ABSTRACTION
Unified interface for AMD SEV and Intel TDX.
> END_TO_END_ENCRYPTION
Encrypted traffic, data, and execution state.
> REMOTE_ATTESTATION
Verify the integrity of remote systems and establish trust.
> MULTIPLE_RUNTIMES
Python, Docker, and WebAssembly workloads.
> SDK_AND_CLI
Comprehensive tools to streamline development.
// Get Started
Ready to deploy confidential AI?
Explore the documentation or check out the code on GitHub.