Why Decentralized AI is Incomplete Without TEEs - Decrypt
06/28/2024 13:01Trusted Execution Environments will help realize the full potential of decentralized AI, a new whitepaper argues.
As the demand for more secure and efficient artificial intelligence (AI) computing grows, decentralized AI (DeAI) will become an increasingly vital component of AI infrastructure.
But in order to realize its full potential, a new whitepaper argues, DeAI will need to integrate the technology of Trusted Execution Environments (TEEs).
“As the demand for more secure and efficient AI computing grows, the role of TEEs will become increasingly vital,” the whitepaper’s co-author Nukri Basharuli, Founder and CEO at Super Protocol, told Decrypt. “Their ability to seamlessly integrate with existing blockchain architectures makes them indispensable in the development of future-proof DeAI systems.”
The current AI landscape
To deliver high-quality results, AI models require reliable and private user data. But in today’s AI community, sharing data and models creates risks—such as compromised confidentiality and ownership rights.
DeAI promises to create a collaborative environment that limits access to sensitive data for all participants. In its current state, though, the novel technology comes with scalability and reliability shortcomings that limit its potential, the whitepaper’s authors argue.
Blockchain's greatest strength lies in its ability to transparently verify transactions and smart contracts.
However, when it comes to the complex computations required for AI, traditional methods fall short, said David Attermann, Head of Web3 & AI Investments at M31 Capital, a global investment firm dedicated to crypto assets and blockchain technologies (M31 Capital does not hold a position in Super Protocol). “Verifying these computations using blockchain alone becomes prohibitively expensive and impractical,” he explained, a limitation that significantly hinders the scalability and efficiency of DeAI.
What are TEEs?
Embedded within CPUs and GPUs, TEEs provide a secure, isolated environment for executing code and processing data. This isolation ensures that even if the main system is compromised, the data and computations within the TEE remain protected.
Indeed, the device you’re reading this on right now almost certainly includes a TEE. They’re used in smartphones and computers to store sensitive information like biometric data, and in cryptocurrency hardware wallets to store cryptographic keys.
“TEEs are a cornerstone of many secure applications today,” Basharuli said, noting that their integration into blockchain systems “can transform them from merely 'trusted' to genuinely 'trustless,' by leveraging cryptographic primitives and consensus mechanisms.”
He added that platforms like Super Protocol demonstrate how TEEs can be effectively integrated with blockchain technology for DeAI applications, enabling trustless and secure AI computing by combining smart contract governance with clusterized TEE computing capacity.
“In a trustless environment, collaboration becomes trustless,” Basharuli said. “Everyone can trust the system to handle their assets securely and confidentially.”
How TEEs enhance DeAI
Trusted Execution Environments bring a number of critical benefits to decentralized AI applications, the whitepaper’s authors explained.
First off, said Attermann, “TEEs provide a level of security that is essential for sensitive AI computations,” ensuring that data remains encrypted and protected throughout the computation process.
By isolating sensitive data from the broader system, TEEs ensure that data “remains confidential and secure,” in the event that the broader system is compromised, Basharuli added.
They also enable the integrity of data and code to be maintained, preventing unauthorized modifications. “Integrity is crucial, and TEEs maintain this by isolating the execution environment,” said Attermann.
In addition, TEEs provide an isolated execution environment, ensuring secure processing of sensitive operations, and enable cryptographic attestation—verifying that the code running in the secure enclave is genuine and untampered.
TEEs and the secure computing landscape
TEEs sit alongside a number of other secure computing technologies, including zero-knowledge proofs (ZKPs) and fully homomorphic encryption (FHE).
But, Attermann said, these technologies have limitations that make them ill-suited for DeAI in their present state, including high cost and slow performance.
“The combination of TEEs with blockchain ensures security and confidentiality, while simultaneously enhancing scalability and performance. This makes TEEs the most practical and immediate solution for achieving verifiable, trustless AI computing,” Attermann told Decrypt.
“TEEs are extremely cheap and extremely fast,” he added, “but they don't have the sex appeal for some reason. We're trying to change that.”
Challenges and the path forward
Integrating TEEs into DeAI systems still presents some challenges. Ensuring that TEEs are used correctly and consistently across decentralized networks requires robust governance and oversight, the whitepaper argues, while fostering industry-wide acceptance and understanding of TEE technology is crucial.
Super Protocol and other pioneers in the field are actively addressing these challenges, Basharuli said. By demonstrating the practical benefits of TEEs in real-world applications and collaborating with industry leaders like Nvidia, they are paving the way for broader adoption. He added that “confidentiality is essential for secure collaboration in AI, and cryptocurrency incentivizes collaborating parties, thereby accelerating adoption.”
AI collaboration reimagined
TEEs provide the security, confidentiality, and verifiability necessary for DeAI to fully realize its potential, the whitepaper argues. “By integrating TEEs with blockchain, we can move beyond "trusted" systems to genuinely "trustless" environments, ensuring that the future of AI is decentralized, secure, and scalable,” Basharuli said.
For developers eager to start experimenting with DeAI, the Super Protocol AI Marketplace is live now. You can use it to collaborate on AI models and applications in a confidential, zero-disclosure computing environment, using everyone’s solutions, storage, data, or compute power.
⚡️Guess what? We've got some pretty cool news!
⏱By the end of this summer, in partnership with @nvidia , @intel , and @0xPolygon we will launch a public marketplace.
🛡But it's not just any marketplace. It's one for self-sovereign, verifiable confidential AI agents, all… pic.twitter.com/6Md9HHN0ed
— Super Protocol (@super__protocol) May 16, 2024
Soon, developers will also be able to monetize their AI agents and applications on the Super AI Marketplace through transparent blockchain instruments that split revenue among all contributors. They’re also using stablecoins for payments and payouts as part of a wider effort to make it easier for Web2 developers to onboard themselves into the ecosystem.
“Building a hub for confidential AI collaboration requires solving the chicken and egg problem of supply and demand—meaning that users should bring useful datasets and models before they start developing AI products”, Basharuli told Decrypt. Super Protocol’s solution is to offer a comprehensive incentivization programme to attract models, data and users to the AI marketplace in exchange for its token-based rewards.
To learn more about Super Protocol, visit the website, Twitter and LinkedIn. For more information about M31 Capital, visit the website, Twitter , LinkedIn and Substack.
Daily Debrief Newsletter
Start every day with the top news stories right now, plus original features, a podcast, videos and more.