rectangle with AI network
AI Sparkles SECURE AI

Unlock the power of AI without compromising privacy.

Run AI/ML models on sensitive data with verifiable privacy guarantees, powered by TEEs.

Upside trending Arrow ADVANTAGES

Features

Data confidentiality

TEEs ensure that input data and model weights remain encrypted, even from the node operator.

Verifiable results

Generate attestations of correct model execution, ensuring trust in the output.

Model protection

Keep your proprietary AI models secure from theft or reverse engineering.

Performance isolation

Run AI workloads in dedicated secure enclaves with guaranteed resource allocation for consistent performance.

HOW TO DEPLOY

Your decentralized AI model

cog wheel
STEP 1

Encrypt

Send your sensitive data securely to the network using end-to-end encryption protocols

Dots connected with cube
STEP 2

Process

Oyster nodes execute your selected AI model within hardware-secured TEE environments

Cube inside circle with connected dots
STEP 3

Generate

AI model produces inference results while creating cryptographic proofs of execution

Cube inside circle with connected dots
STEP 4

Verify

Confirm the authenticity and integrity of your results through tamper-proof attestations

cog wheel
STEP 1

Encrypt

Send your sensitive data securely to the network using end-to-end encryption protocols

Dots connected with cube
STEP 2

Process

Oyster nodes execute your selected AI model within hardware-secured TEE environments

Cube inside circle with connected dots
STEP 3

Generate

AI model produces inference results while creating cryptographic proofs of execution

Cube inside circle with connected dots
STEP 4

Verify

Confirm the authenticity and integrity of your results through tamper-proof attestations

streamline arrows

LEVERAGE

Marlin TEEs in action

Learn how top tier AI teams are leveraging Marlin TEEs to make their AI apps confidential

Verida AI

Personalised AI assistant
Use the Marlin TEE Plugin to ensure user data remains secure

Autonolas

Verifiable off-chain computation
Integrate Marlin TEEs to run Autonolas’ AI framework with cryptographic guarantees.

Chromia

Trustless AI with on-chain memory
Chromia agents use Marlin TEEs to securely access short- and long-term memories.

Eliza

Deploy trust-minimized AI interactions
Use the Marlin TEE Plugin to ensure Eliza agents act as programmed.

Autonolas

Verifiable off-chain computation
Integrate Marlin TEEs to run Autonolas’ AI framework with cryptographic guarantees.

Chromia

Trustless AI with on-chain memory
Chromia agents use Marlin TEEs to securely access short- and long-term memories.

USE CASES USE CASES

Potential use cases

Brain with Arrows

Prediction Models

Dots connected to circle

Self Governing DAOs

Globe with cloud spark

Autonomous Worlds

Credit card swapping with Score

Credit Scoring

View our documentation

Can't find the answer you're looking for? We have a database of useful documents and resources dedicated to help you

Up circle arrow FAQ

Questions?

We have answers

A trusted execution environment (TEE) is a secure area of a main processor. It helps code and data loaded inside it to be isolated from other processes usually at a hardware level in order to provide confidentiality and integrity. Data integrity prevents unauthorized entities from outside the TEE from altering data, while code integrity prevents code in the TEE from being replaced or modified by unauthorized entities, which may also be the computer owner itself.

Oyster is a sub-network of Marlin that specializes in offering TEE-based coprocessors. It is fast, cheap and secure. It allows instances to be rented individually for any stretch of time or for tasks to be delegated amongst a pool of nodes in a serverless way without individually renting instances and paying only for the duration of the task's execution. Nodes can be rented and tasks delegated using both smart contract calls and web 2 APIs. A set of auditors routinely query Oyster nodes to ensure uptime guarantees.

Oyster is very flexible and versatile. Developers can port existing web 2 applications on Oyster, write code in the programming language of their choice and use popular libraries and frameworks. Additional benefits include:


1. Secure: Hardware isolation ensures that programs run correctly as intended without exposing their contents and prohibiting interference from the host machine or its owner.


2. Fast: Oyster nodes are essentially like usual servers. There’s no redundancy in computation, complex protocol logic or cryptography. Thus, apart from a little TEE overhead for encrypted memory access, Oyster nodes are very performant and provide an experience similar to running programs on a normal machine.


3. Cheap: TEEs can be considered commoditized. While a little more expensive than vanilla servers, they are very cheap compared to blockchains, MPC, FHE or ZK proofs. Consequently, TEEs are a very affordable mode to carry out computations verifiably.

Oyster is a permissionless network. As such, programmers can contribute to its codebase, developers can build trustless web 3 applications by deploying their backends on Oyster nodes and infrastructure providers can run TEE-enabled Oyster nodes.

Features Trusted Execution Environments (TEEs) Zero-Knowledge Proofs (zk)
Main Purpose Secure execution of code in an isolated environment Prove knowledge without revealing it
Scalability Generally good, limited by hardware Limited by computational overhead
Performance Faster as it leverages direct hardware support Generally slower due to the cryptographic operations required
Support for Complex Computations Better suited for complex computations, given adequate TEE capabilities Limited, best suited for simpler proofs
Development Complexity Medium, as it often involves utilizing existing hardware features High, requires expertise in cryptography
Integrity Verification Verified by hardware and software measures internal to the TEE Can be verified publicly without revealing underlying data
Trust Model Hardware-based trust; relies on the manufacturer and design of the TEE Cryptographic trust; no need for trusted hardware
Overhead Low to medium, depending on the implementation High, due to complex calculations
Portability Dependent on specific hardware and therefore less portable Can be implemented in software across different platforms
Data Privacy Yes, data is protected within the TEE Yes, no data is revealed
Regulatory Acceptance Subject to hardware security review and potential backdoor concerns Subject to cryptographic review
Attestation Capability Hardware and software attestation mechanisms provide trust assurance Mathematical proof provides attestation of correctness
Finality Instant execution within the TEE Instant verification once proof is computed
Maturity Well-established with industry standards Relatively new, rapidly evolving