Mind Network is the first FHE-based re-staking solution designed for AI and PoS networks.

The Holy Grail of Cryptography - Fully Homomorphic Encryption

On May 5, Ethereum founder Vitalik Buterin once again shared his 2020 FHE (Fully Homomorphic Encryption) article on Twitter, which continued to ignite everyone's attention and discussion on the application of FHE technology. Vitalik's article deeply introduces the relevant mathematical principles, original English version.

FHE (Fully Homomorphic Encryption) in Chinese means fully homomorphic encryption computing. Like ZK, it is one of the frontiers of cryptography and is also known as the holy grail of cryptography.

Simply put, fully homomorphic encryption is to perform calculations directly on encrypted data without decryption.

When 1+2 is added, it is easy to get the result 3. However, after encryption, Encrypt(1)+Encrypt(2), we can still get Encrypt(3). This is FHE, and the ciphertext calculation = the encrypted plaintext calculation.

Unlike ZK, the application of FHE in Web3 focuses more on data privacy and security. It is not difficult to find from current applications that ZK is more reflected in the direction of capacity expansion.

Although Web3 is more well-known for ZK technology based on ZKRollup, FHE is gradually releasing its unique potential in many fields, especially AI.

Mind Network

Mind Network is the first FHE-based re-staking solution designed for AI and PoS networks.

Just as EigenLayer is a re-staking solution for the Ethereum ecosystem, Mind is a re-staking solution for the AI ​​field. Through re-staking and FHE consensus security solutions, the token economic security and data security of the decentralized AI network are guaranteed.

From the perspective of team background, the main members of Mind are professors and doctors in AI, security, and cryptography, from Cambridge, Google, Microsoft, IBM and other institutions. The core members were selected as one of the 12 Ethereum Foundation Fellows in the world, and conducted research in the fields of cryptography and security with the Ethereum Foundation research team. Mind's world-first FHE+Stealth Address solution - MindSAP (research paper link, original text is brain-burning and you can read it yourself), solved the problem in the Stealth Address Open Problem raised by Vitalik, and attracted considerable attention in the Ethereum community, and published papers and speeches many times.

Mind Network was selected for Binance Incubator in 2023 and completed a $2.5 million seed round of financing with the participation of well-known institutions such as Binance. It also received the Ethereum Foundation Fellowship Grant, was selected for the Chainlink Build Program, and became a Channel Partner signed by Chainlink.

In February 2024, Mind Network became a key partner of the well-known cryptography company ZAMA in the field of FHE.

Recently, Mind Network has further accelerated the expansion of its ecological territory, providing AI network consensus security services for io.net, Singularity, Nimble, Myshell, AIOZ, etc., providing FHE Bridge solutions for Chainlink CCIP, and providing AI data security storage services for IPFS, Arweave, Greenfield, etc.

FHE+AI, addressing the core pain points of AI

At the Hong Kong Web3 Conference in April this year, Vitalik expressed his expectations for the future of FHE in scenarios such as Encrypted Voting. As the forefront of cryptography, FHE is also the ultimate direction of cryptography pursued by Ethereum.

The founder of ZAMA recently published an article about his "Master Plan". It outlines the company's vision to create an end-to-end encrypted network HTTPZ ("Z" stands for "Zero Trust") and proposes to make FHE ubiquitous in the fields of blockchain and artificial intelligence.

Several key links in the AI ​​field, including training, tuning, use and evaluation, all face the same problem in the process of decentralization: how to remove the trust assumption. For example:

  • When the AI ​​model is being trained, cross-validation is required to select the best training result.

  • Before using AI services, existing services need to be ranked to determine the best service.

  • AI models also need to be constantly tuned and iterated, and independent evaluation is required

In centralized scenarios, these links are based on the assumption of compliance and trust in large companies, with large companies providing trust and guaranteeing that they will not do evil.

However, in the process of decentralization, there is no credit endorsement. How to verify whether the collaboration of all participants is fair and effective is a difficult point. This is exactly the starting point of FHE empowerment.

For example

  • When cross-validation is required during AI model training, the best training result is selected through secret voting, removing assumptions similar to those of OpenAI.

  • When AI services need to be ranked before use, the quality of each service is determined through anonymous ratings, removing trust assumptions similar to AI AppStore

  • AI models also need to be constantly tuned and iterated. When independent evaluation is required, a credible evaluation is completed through random sampling checks to remove the trust assumption of the evaluation agency.

The participation of FHE can also enable AI to achieve zero trust, making up for the trust assumption that ZK still requires off-chain aggregation.

There are many other AI examples that can be cited, including that such zero trust can enable AI Agents and Multi-Agents to better achieve intelligent interconnection and achieve benign governance.

At the same time, FHE's unique ciphertext computing characteristics can also achieve two other difficult problems: data privacy and data ownership:

  • Who can see our data? =Data privacy

  • Who owns the data given to us by AI? =Data ownership

FHE can ensure that data is always encrypted on the user side and exists only in ciphertext form outside the user, including storage + transmission + computing.

So far, except for FHE, data can only be encrypted during storage and transmission. However, once it involves calculation, the ciphertext needs to be decrypted into plaintext, which happens to make users lose ownership of the data. There are many such examples in real life. Once your plaintext data is copied by others, they can copy it many times. Users have no way of knowing whether others are using your data. They can only rely on the self-declaration of the data user and the supervision of three parties. FHE allows users to obtain their consent when decrypting and seeing the plaintext data even if the ciphertext data is copied. Then users can perceive the dynamics of the data at any time, making the data available and tradable but not visible, protecting data privacy while also truly protecting data ownership.

Such a feature is urgently needed by AI + Web3. It allows everyone to stake in an open manner and reach consensus in an encrypted manner, which can prevent evil and waste.

The next big thing in AI

From this point of view, the combination of AI and Web3 is inevitable. FHE is to AI as the [next big thing] is to Apple.

Recently, IO.NET and Mind Network announced a deep collaboration to jointly create solutions to enhance the security and efficiency of artificial intelligence. IO.NET will introduce Mind Network's fully homomorphic encryption solution into its distributed computing platform to help strengthen the security of its products.

For more information about the collaboration, please visit: Mind Network and io.net Partner up for Advanced AI Security and Efficiency

IO.NET has set a good example for combining AI and FHE with distributed computing.

Taking IO.NET as an example, users provide computing power and AI developers rent computing power.

When a developer comes to an AI project, he puts forward a requirement, which is split by the system and calculated by the computing power provided by the user.

At this time, several questions are involved: Whose computing power is being rented? Is the calculated result correct? Will the privacy of both parties be leaked when renting computing power?

1. Whose computing power should I rent?

Under normal circumstances, the node selection is based on test operations, that is, requirements are issued from time to time to test which nodes are online and ready to accept requirements.

During this process, targeted manipulation of related nodes may occur to obtain priority, similar to the MEV attack.

In response to this, Mind provides a fair distribution mechanism through FHE. Since both requests and data are encrypted, nodes cannot make favorable choices based on this.

2. Is the calculated result correct?

In distributed computing, ensuring that the calculation results are correct requires a certain consensus, namely voting.

When nodes know each other's selection results, follow-up investment may occur, resulting in unfair and incorrect results.

FHE encryption calculation, the voting results between nodes are encrypted with each other, but they can still participate in the final calculation, ensuring the fairness of the results.

3. Will the privacy of both parties be leaked when renting computing power?

The core of FHE is data security. The data itself is encrypted during calculation, and the problem to be calculated is also encrypted, so there is naturally no privacy leakage.

From the perspective of Restaking:

IO.NET itself can be seen as a PoS network. Nodes need to stake IO tokens in order to obtain IO rewards from their computing power contributions.

The problem that may arise is that if the price of the staked tokens fluctuates too much, the security of the validator and the network will be affected.

Mind’s solution to this is Dual Staking or even Triple Staking.

Staking supports BTC/ETH’s liquid staking tokens and blue-chip AI network tokens to diversify risks and increase the overall security of the network. It is essentially an advanced version of Restaking’s shared security.

At the same time, Mind also supports Remote Staking. For LST/LRT assets, there is no need for actual cross-chain to ensure the security of assets.

A few days ago, Mind just completed the Glaxe test network mission, with more than 650,000 active users participating and generating 3.2 million test network transaction data.

According to official news, Mind's official network protocol will also be launched in the near future, so you can pay attention to it.

Summarize

In general, we found that although Mind talks about FHE and AI, the key word is actually "security", using cryptography to solve various core security problems.

Restaking is token economic security; Remote Staking is asset security; FHE is data security; AI+FHE is consensus security.

The building of blockchain is based on cryptography, and perhaps the answers to the future will also be found in cryptography.

In addition to the AI ​​network, Mind Network is also expanding the scope of application of its solutions, collaborating in decentralized storage, EigenLayer AVS network, Bittensor Subnet, cross-chain bridges and other areas, demonstrating the huge potential of FHE.

In Web3 of 2024, if ZK kicked off the cryptography field, then FHE will be the main theme of the second half of the year. At the same time, the popularity of AI remains high. With the triple narrative of AI+FHE+Restaking and the halo of the Ethereum Foundation and Binance's investment, whether Mind can take the lead in FHE will be revealed as the mainnet goes online.