Deep into Mind Network: When Fully Homomorphic Encryption Meets Restaking, the Consensus Security of Crypto AI Projects is Within Reach

Written by: Deep Tide TechFlow

AI and Restaking are widely recognized as the leading narratives accompanying this bull market cycle.

The former has spawned various AI star projects, while the latter has derived multiple LRT projects centered around EigenLayer, with various point-earning gameplay continuously emerging.

However, a very obvious feeling is that these two narratives seem to have entered a halftime stage. Although the number of projects in the field has increased, they are becoming increasingly homogenized, making it harder to find innovative stories from 0 to 1.

At the same time, when AI and Restaking become a "correct narrative," this "correctness" does not mean "perfection":

Are a large number of AI/Depin projects really decentralized? Recent data also shows that Eigenlayer's TVL is declining. Can Restaking only be used to ensure the security of Ethereum's ecological AVS?

Therefore, in the second half of the hot narratives, projects that solve key common problems are the treasures waiting to be discovered.

From this perspective, Mind Network has caught our attention in the current market -- it can solve the problem of insufficient decentralization in many AI/Depin projects and provide more uses and value for Restaking.

If EigenLayer is the restaking solution for the Ethereum ecosystem, then Mind is the restaking solution for the AI field.

By utilizing restaking more flexibly, combined with a fully homomorphic encryption consensus security solution, it ensures the security of the token economy and data of decentralized AI networks.

More importantly, the project completed a $2.5 million seed round financing in 2023 with participation from well-known institutions like Binance. It is currently also in deep cooperation with popular new AI/Depin projects like io.net and Myshell, and the anticipation for mainnet launch and incentive activities is also high.

However, for most readers seeing this project for the first time, on one side is the obscure fully homomorphic encryption, and on the other side is the profit-chasing Restaking. How can these two combine to solve the key problems of AI projects?

In this issue, let us delve into Mind Network and understand this potential project that integrates hot narratives like AI, Restaking, and fully homomorphic encryption.

AI projects rush to slay dragons but become the dragons themselves due to inability to achieve "zero trust."

To understand what Mind Network specifically does, we first need to grasp the problems currently faced by AI projects.

Perhaps, the slayers have gradually become the dragons, which serves as the best footnote to describe the current state of crypto AI projects.

From the perspective of slaying dragons, the core narrative of crypto AI (or DePIN) projects is decentralization, using more decentralized computing power, algorithms (models), and data to combat the monopoly of large companies over AI elements and break the trust in the authority of large companies.

This narrative, while correct and naturally resonating with the public, leaves a more significant problem that makes AI, once decentralized, more likely to become a dragon:

It cannot achieve "zero trust" for validators in a decentralized environment.

A bit hard to understand? Let’s look at specific examples.

For example, in common crypto AI projects, everyone needs to decentralize the verification/voting of AI models to select which model is better.

But in practice, the business model is often that the validators (nodes) in the project select the best-performing AI model. How can you ensure that the one they select is indeed the best-performing?

"Following their selection" under the POS mechanism does not equal "the best selection, the fairest selection."

Similarly, in AI agent services, when ranking well-performing services, how can you ensure that the services ranked at the top are genuinely the best?

As for the DePIN scenario, if a task is assigned to nodes in DePIN for computation, how can you ensure that the validators fairly assign this task to suitable nodes rather than cheating by giving it to familiar nodes?

These examples actually reflect a key common problem ---- In various decentralized AI networks, the decisions of validators become the center you must trust.

Thus, you must trust the decisions of validators or key participants in the network, hoping that they do not act maliciously or make incorrect decisions.

Projects that loudly proclaim decentralization are themselves constrained by trust limitations within the network. Zero trust has still not been achieved, and the current AI narrative is not perfect.

What else is needed to face the problem?

Clearly, we need to use some technical mechanisms and economic designs to maximally solve the trust dependency issues regarding key participants in verification/voting/decision-making within current AI project networks.

And this is also the area where Mind Network is deeply engaged.

The Holy Grail of Fully Homomorphic Encryption, placed in the most suitable position by Mind Network

Mind Network excels in what is known as the Holy Grail of cryptography: fully homomorphic encryption.

But what do the problems exposed in the above AI and DePIN projects have to do with fully homomorphic encryption?

If we look at the essence, these problems all point to the allocation, selection, and decision-making of resources --- it’s not about technology, but about "human governance."

Wherever there is human governance, the premise for wrongdoing is that network participants can fully and openly understand known information (I know a big holder invested, so I follow suit).

Smart as you are, you must have sensed the utility of FHE:

What if information is no longer known to everyone?

Fully homomorphic encryption (hereinafter referred to as FHE) has a perfect fit for solving the aforementioned human governance issues.

FHE, as the Holy Grail pursued by cryptography, has recently been emphasized by Vitalik Buterin for its role in the Web3 space. We won’t spend a lot of time explaining the principles of FHE; you just need to know its function --- it allows complex computations on encrypted data without decryption, thus providing a solution where data can remain secure and private throughout the analysis process.

But to hold the Holy Grail, one must bear its weight.

While FHE's encrypted computation is excellent, the resource overhead is significant. Using it for training AI models incurs extremely high costs, which is not a wise direction for crypto AI projects.

Mind Network's use of FHE has a sense of leveraging minimal resources for maximum output, placing the Holy Grail in the most suitable position.

That is, instead of using FHE for training AI models and changing parameters, it is used in areas filled with "human governance" such as cross-validation, selection, ranking, and voting after the AI models have been trained. The resource overhead is controllable, and the problems to be solved are very clear:

If participants in the AI network conduct business without knowing each other's selection/voting results, there will be no "following big holders, trusting authoritative nodes" behavior, eliminating decision biases caused by identity influence, allowing decentralized decision-making to return to its essence, thus identifying truly good AI models and AI services.

Thus, while the road for FHE to perform general computation is fraught with challenges, using FHE for decentralized specific processes --- validation, is self-consistent and feasible. It guarantees zero trust in the validation process, achieving consensus security and true decentralization for crypto AI projects.

And on the other side of security lies fairness.

We can also look at a specific case to see how Mind Network's fairness is reflected in the encrypted execution of validation:

• 1. AI projects integrate the fully homomorphic encryption validation service through the SDK provided by Mind;

• 2. At the same time, AI projects register on the Mind network to confirm the identity of the project. Mind will generate a smart contract on the target project network/chain to synchronize subsequent operation changes and execution results.

• 3. AI projects publish validation tasks that need to use fully homomorphic encryption (such as which AI model is good) on the Mind network. The FHE voting service begins to function, allowing the validation nodes of the AI project to execute the voting process while not seeing the plaintext results of the votes.

• 4. The voting results and related data changes are transmitted to Mind's own chain through the smart contract and are promptly synchronized and recorded.

• 5. In the above steps, when AI projects call Mind's services, they will be charged Mind project tokens as gas fees (tokens have not yet been released).

The reasoning is similar; if we focus on a DePIN project, using Mind Network will also yield a fairer resource allocation effect. Let’s take IO.net, which collaborates with Mind Network, as an example:

• 1. IO.net integrates the fully homomorphic encryption validation service through the SDK provided by Mind;

• 2. After integrating the service, nodes holding GPUs gain consensus capabilities under fully homomorphic encryption, making it possible to fairly allocate tasks to suitable nodes when AI computing tasks arise.

Wait, but what does this have to do with Restaking?

Everything discussed above seems to be at the technical level; what does it have to do with the asset-level Restaking?

Mind Network provides a solution based on FHE that technically facilitates the verification security of AI networks; however, to participate in verification and enjoy this security is closely related to the economic network structure of most AI/DePIN projects.

PoS, or Proof of Stake, is the underlying consensus logic for most crypto projects.

So, if any AI project accepts Mind Network's more equitable FHE-supported AI model/service selection, ranking, and verification, since most project nodes represent voting/verification rights through the POS mechanism, the size of the staked assets under that node is closely related to whether it has the right to participate in the fair verification guaranteed by FHE.

Mind Network's key initiative at the asset level is to expand the scope of Staking and Restaking in an open manner, using homomorphic encryption to ensure verification consensus within the AI network.

Different roles participating in the network can meet their respective interests.

For validation nodes of AI projects, increasing the amount of Restaking provides more opportunities and voting rights to conduct FHE verification tasks in Mind Network.

For general users, they can stake their LST/LRT assets to the aforementioned nodes through delegated agency to earn APR returns.

This seems to have similarities with the Restaking we are familiar with in EigenLayer, but essentially, they are different paths leading to the same destination:

EigenLayer uses restaking to ensure the security of different AVS in the Ethereum ecosystem; Mind Network uses restaking to ensure the consensus security of different AI networks in the entire crypto ecosystem.

It is worth noting that the reason it is called "entire ecosystem" is closely related to another key feature of Mind Network: Remote Restaking.

With Remote Staking, there is no need to cross-chain your LRT tokens on different chains; instead, you can indiscriminately stake your LRT from different chains to a validation node of a certain AI network through remote staking, greatly lowering the participation threshold for users and integrating fragmented liquidity in a multi-chain landscape.

Extensive ecosystem construction and solid technical strength

What other catalysts are worth paying attention to regarding Mind Network?

First, on the product side, the testnet has attracted 650,000 wallets and processed 3.2 million transactions, with the complete mainnet functionality expected to be launched soon;

Second, in terms of ecosystem construction, since the product is positioned as a platform to empower other AI projects, how many leading projects it can attract for cooperation is crucial.

Currently, Mind Network provides AI network consensus security services for io.net, Singularity, Nimble, Myshell, AIOZ, etc., offers FHE Bridge solutions for Chainlink CCIP, and provides AI data security storage services for IPFS, Arweave, Greenfield, etc. ---- leading AI, storage, and oracle projects are included, with the potential to become a "golden shovel."

Additionally, from a background perspective, in 2023, the project was also selected for the Binance Incubator and completed a $2.5 million seed round financing with participation from well-known institutions like Binance. It also received the Ethereum Foundation Fellowship Grant, was selected for the Chainlink Build Program, and became a signed Channel Partner of Chainlink.

In terms of technical strength, in addition to its team comprising top professors and PhDs from relevant fields of AI, security, and cryptography, a noteworthy point is its collaboration with top companies in the field of fully homomorphic encryption research.

In February of this year, Mind Network officially announced a partnership with ZAMA, a leading open-source encryption company in the field of fully homomorphic encryption, which has completed a $73 million Series A financing led by Multicoin and Protocol Labs;

Recently, the cooperation between the two parties has further expanded, jointly launching a new Hybrid FHE (Hybrid Fully Homomorphic Encryption) AI network, promoting the application of AI algorithms on encrypted data, undoubtedly adding another layer of technical benefits to the project itself.

According to close sources, Mind Network has chosen to use ZAMA's underlying technology library for its own technical development in its collaboration with ZAMA, which also demonstrates Mind's understanding of the industry:

Fully homomorphic encryption has significant resource overhead, and using the underlying library maximizes capability output without compromising performance.

Moreover, in addition to empowering itself with better technology, Mind Network is also outputting its capabilities to help the crypto ecosystem improve.

In May, the project also partnered with Chainlink to launch the first fully homomorphic encryption (FHE) interface based on the Cross-Chain Interoperability Protocol (CCIP). This move not only enhances the security of cross-chain communication and transactions but also realizes a more trustworthy and user-centered Web3 ecosystem.

As of the time of writing, Mind Network has already established partnerships with multiple leading projects across different ecosystems and sectors; considering its positioning to empower other projects, we might anticipate a subsequent "golden shovel" effect.

Conclusion

When fully homomorphic encryption meets Restaking, perhaps Mind Network will become a new driving force in the second half of this year's mainstream crypto narrative.

Fully homomorphic encryption serves as a medium, reaching a large number of crypto AI projects for business optimization, supporting the true "decentralization" and zero trust of decentralized AI projects; Restaking paves the way, further absorbing liquidity from different chains, and the rapid rise of the project's TVL can also be anticipated.

Undeniably, the Holy Grail of fully homomorphic encryption attracts the market's attention to new stories, while Restaking draws the market's liquidity. When the consensus security of AI projects becomes within reach, the concentration of attention and liquidity can undoubtedly be expected to drive the project's subsequent development.

Projects like Mind Network, which refine the correct narratives (AI, Restaking) through their own technology, represent a more gentle disruption in the second half of the mainstream narrative.