Aoba Protocol Litepaper

Under-Collateralized Lending & Borrowing Protocol Based on On-Chain Credit

TradFi (traditional finance) is riddled with issues. One of the biggest problems is that the financial system is individually created in each country while conforming to strict regulations in each country, and is fragmented from country to country. For example, have you ever had the experience of doing business with a Tokyo-based company while in New York, only to be charged a large transfer fee, have to wait days for the money to arrive, and then have to have your accounting department go through the hassle of repeatedly contacting the customer for confirmation? The Internet has made it possible to transfer information globally. But what about value transfer? The best way to facilitate the global transfer of value is to use crypto assets or DeFi. However, DeFi also has many issues. In particular, Lending Protocol has two major issues: (1) Excessive collateral is required, and (2) Funds are circulated only in the DeFi world and are not used much in the real world.

Look at this diagram. This diagram summarizes the evolution of the Lending Protocol.

※Based on an article and drawn by Aoba.

Typical protocols for the first generation of Lending include Aave and Compound. If you were to use these protocols, if you wanted to borrow $6,000, you would have to pledge at least $9,000 worth of tokens as collateral. This captures the need to borrow other tokens while keeping a highly volatile asset such as ETH as collateral without selling it.However, the problem is that it cannot meet the demand for funds to spend more than the value of the asset you have using your own credit.

TrueFi and Maple Finance are representative of the second generation of Lending protocols. These protocols provide Under-Collateralized loans by limiting borrowers to institutional investors and other users with high creditworthiness. A separate contract is executed in the borrower’s real name, allowing for legal action in the event of default.One of the advantages of using DeFi instead of TradFi is that it is open finance for everyone, but in DeFi Lending 2.0, anonymity is disregarded. This is problematic in that many users who cannot create bank accounts will not be able to use it. Also, in these Protocols, credit decisions are determined by human voting. In other words, DeFi Lending 2.0 Protocols are similar to the top-down financial systems created by the state or central bank, and not the bottom-up financial systems unique to DeFi.

The problems of DeFi Lending 1.0~2.0 can be solved by utilizing data science technology. In a nutshell, all we need is an Under-Collateralized Lending Protocol where AI/ML models make credit decisions. The first generation issue of over-collateralization was caused by the lack of credit evaluation for each borrower. This was due not only to the anonymity of the Wallet, but also to the lack of data on that Wallet at the time of the Protocol’s launch, making it difficult to make individual credit decisions. The second generation brought credit in existing finance into DeFi to achieve individual credit evaluation. However, as of 2022, not only has on-chain data about individual Wallets accumulated to the point where analysis can begin, but there is also a trend toward establishing an on-chain identity with the concept of SBT (Soul Bound Token) being proposed. Therefore, 2022 is the best year to use data science technology to evaluate Wallet credit and launch the Under-Collateralized Lending Protocol.

Aoba is an Under-Collateralized Lending Protocol being developed by a team of finance background members, data scientists, and smart contract engineers. It features the use of data science technology to evaluate the creditworthiness of each wallet and issue a Non-Fungible Credit Score, and the ability to borrow tokens with less than 100% collateral based on that NFCS.


Aoba’s vision is to create a decentralized credit economy in which anyone in the world can participate. There are three steps to realize this vision.

Phase I: Establishing the Concept of Credit Creation in DeFi

Our first task is to establish the concept of credit creation in DeFi. Currently, there is no concept of credit creation in the crypto world. However, there is a wave of DIDs (decentralized identities) that are becoming more prevalent. This is a movement to integrate one on-chain identity and tie credit to it. In the future, Aoba’s NFCS will be used as the basis for developing tools for recruiting people for on-chain projects, and NFCS will be tied to Social Graph.

Phase II: Expansion of Aoba use cases in TradFi

In the second phase, we will expand the use cases of Aoba outside of DeFi. Demand for token borrowing has already emerged in games, etc. As more and more real estate purchases and business investments are made with tokens in the future, the demand for token borrowing will explode. At this time, even other Protocols will refer to Aoba’s NFCS to determine how creditworthy a borrower is.

Phase III: DAO

In the third phase, the management team responsible for credit scoring will be disbanded, leaving Protocol operations to the community of Aoba token holders. Incentives will be designed to ensure the continued development of the Protocol, such as operating the machine learning model and updating the Lending functionality. This will create a bottom-up decentralized credit economy instead of the current top-down credit decision financial system.


There are two most important points in the operation of Under-Collateralized Lending.

(1) Accuracy and efficiency of data analysis

The accuracy of machine learning models varies greatly depending on the data input. However, when only on-chain data is input, it is difficult to differentiate by the data input because anyone can access the data source. On the other hand, there is very little on-chain repayment performance data for Under-Collateralized loans as of August 2022. This means that it is difficult to create a machine learning model with default as the objective variable. In this situation, it is important to find a way to put data that can substitute for default as the objective variable. In order to create a unique model, the model will be improved multiple times, and it is also important to increase the number of trials by streamlining the preprocessing of the data from the data lake to clean data that can be fed into the analysis before shaping.

(2) Ability to attract borrowers

For Aoba to become “the foundation of a decentralized credit economy in which anyone in the world can participate,” the most important KPI is the amount of borrowing. This is due to the way interest rates are determined. The borrower’s interest rate is determined individually by the creditworthiness of each borrower. On the other hand, the lender’s interest rate is highly dependent on the Pool’s Utilization Rate. For example, if USDC is lent out and no one borrows USDC from that Pool, the lender will obviously earn no more interest than a stable investment of USDC. Conversely, if most of the USDC in the lending Pool is borrowed, the lender will clearly earn a higher interest rate than in other Lending Protocols. In other words, the more borrowers you attract, the higher the lender’s interest rate will be, which will attract more lenders.

There are many points to increase the number of borrowers as follows

・Collaboration with Social Graph

The higher the value of NFCS issued by Aoba, the higher the cost for borrowers to default and dump on-chain credit. This would lower the volatility charge (see below) and lower the borrower’s cost of funding. In other words, the more popular Aoba’s NFCS becomes, the more borrowers will want to borrow tokens with Aoba.

・Guarantor System

The Aoba team believes that data science will solve financial issues. However, there are times when it is more efficient for humans to make decisions directly.

Example) When Alice borrowed USDC from Aoba, she had to pay an annual interest rate of 20% because of her poor on-chain credit. However, Bob, a close friend of Alice’s with good on-chain credit, argues that Alice deserves more credit than Aoba’s ML model gave her. At this time, Bob becomes Alice’s guarantor, which allows Alice to borrow USDC at a lower interest rate; by guaranteeing Alice’s credit, Bob is rewarded with a percentage of the initially reduced interest rate when Alice repays the loan.

This guarantor system complements the weaknesses of the ML model and allows borrowers to obtain financing on good terms.

In addition to the above, other key points include the type of collateral, chain development, design of credit and debt transfers, and token incentives for lenders and borrowers.

Interest Rate

The borrower’s interest rate is calculated by a formula derived from the CAPM (Capital Asset Pricing Model).

Risk free rate is the interest rate that would be paid if the asset were invested risk-free.

You can get this data from protocols that offer fixed-rate loans.

Risk premium is the cost calculated from the borrower’s default rate.

The interest rate will vary depending on the utilization of the POOL, as in the formula used by Aave. When there is money in the POOL, the borrowing rate will be lowered to attract borrowers.

Volatility charge is the cost of hedging the risk that a borrowers choose to default due to changes in collateral value.

It can be derived as a function of the borrower’s threshold, the borrower’s collateral rate, and implied volatility.

The borrower’s threshold is a value that indicates how far the price of the collateral asset drops to the point at which the borrower would choose to abandon on-chain credit and collateral and default.

The higher this threshold is, the lower the volatility charge. The higher the borrower’s collateral rate, the lower the 𝑉𝑐. Implied volatility is derived from option trading information.

FAQ: Under-Collateralized Lending

Q. There is a perception that “wallets are meant to be thrown away,” but will there ever be a situation where everyone defaults?

A. It is unlikely that everyone defaults. Under-collateralized loans are already being provided by Maple Finance and TrueFi, and users have their wallets and reputations tied to them.

There are other organizations and individuals who do not want to default because of the damage to their reputation.

Since DeFi has been in widespread use for more than three years and enough data has been accumulated to provide credit scores to wallets, it is technically feasible to select and finance users with high reputations such as those mentioned above. We plan to operate with strict standards for lending to users with high credit scores at the beginning of the launch, and then gradually lower those standards.

In addition, as evidenced by Ethereum founder Vitalik Buterin’s publication of a document about SBT, there will be many Crypto Projects about on-chain identity formation in the future. This trend is irreversible, and the more popular this concept becomes, the lower the risk of default for loans using Aoba. Until now, wallets were easily disposable, but the time has come to use the main wallet with care.

Q. Does Protocol suffer a loss if defaults increase?

A. Aoba does not take direct default risk, but the user who provided the funds to Pool takes that risk. This risk is reflected in the interest rate earned by the lender.

Q. Is there a standard for measuring the creditworthiness of a wallet?

A. We will create standards for measuring credit in the near future, but we will not open them until we have stabilized the system in order to prevent fraudulent activity. Of course, we are open to the idea that intuitively positive things such as using various DeFi, moving large amounts of money, or having a high number of transactions are worthy of evaluation.

Q. Is there a risk of malicious users appearing?

A. Of course there is that risk, but the risk can be kept low.

For example, for a scoring rule, there are users who meet the standards from the start, and there are users who cheat to meet the standards. The latter can be eliminated using an approach similar to anomaly detection. Similar techniques are widely used in the world outside of DeFi, including detection of cheaters in games, credit card fraud, and machine malfunctions.

Q. Who are the target users?

A. At launch, institutional investors, DAOs, and reputable individuals are the target users, but as the operation progresses, users with a higher risk of default will also be available.

In particular, Aoba is a beneficial protocol for users who are unable to raise funds on the terms they seek under TradFi’s standards. For example, there are many self-employed people who cannot get a mortgage despite having a large amount of ETH in their wallet, and there are also many blockchain-related projects that cannot do business with banks or list on stock exchanges.

Q. What percentage of wallets or individuals would you consider eligible for these unsecured loans?

A. We don’t have specific numbers, but we can say that the top 10% are really good and we focus on these initially, the 80% are average, and the remaining 10% are pretty bad users.

The top 10% of users do not need to worry too much, but we will gradually give them preferential loan amounts, interest rates, and collateral rates.

We plan to expand this to average users based on the data obtained above.

Q. Why are so few people interested in creating a Under-collateralized Lending protocol?

A. To begin with, since DeFi has only been around for about 3 years, the main reason is that until now there has not been enough data to analyze. We believe that 2022 is the best time to start this project.

In addition, there are high hurdles to overcome in order to move forward with this project, as described below.

Whether or not to conceive the idea

If you don’t understand what data science can and can’t do, you won’t come up with the idea in the first place.

For example, can you determine that DeFi has accumulated enough data to be able to analyze it?

Can we trust the idea?

Are you convinced that the trend toward DID penetration and growing on-chain identities is here to stay?

Veteran crypto users, if they haven’t updated their views on value, say, “Wallets are thrown away. Scams should be rampant,” they tend to think. We need to switch to shipping that credit creation will occur in the future at DeFi because it is “more convenient” to have trust, even if it is an anonymous wallet. credit creation has occurred at TradFi because it is “convenient”.

Q. Does the under-collateralized loan have a higher level of liquidation than in the past?

A. Rather than liquidation, they must be repaid. The risk of users not repaying the loan is factored into the borrower/lender’s interest rate.

Q. What happens to the user if the loan is not repaid?

A. The user will not be able to take out another loan. Since the default information will be public, the wallet will receive a bad reputation in other credit granting protocols as well.

Especially since the concept of credit is beginning to be developed in the financial and human resources sectors, once a user defaults, they will be “unable to borrow money or work”.

If we work with other DeFi and DID type projects, users who default on Aoba will be excluded from the on-chain world.

FAQ: Similar Projects

Q. What are similar projects?

A. The closest is RociFi; Teller, TrueFi, and Maple Finance also have Under-collateralized Lending in common.

Q. What is the difference between TrueFi and Maple Finance and Aoba?

A. They share with TrueFi and Maple Finance the fact that they lend and borrow tokens with under-collateral. On the other hand, their customers are limited to institutional investors or DAOs. They can also take legal action in the TradFi world in the event of default; Aoba’s customers are likely to be in the same segment as their customers at launch, but soon anonymous private investors will also be able to use the protocol. There is also no specific legal action in the event of default.

Q. How is this different from RociFi?

A. RociFi is launching on Polygon, while Aoba will be launching on Ethereum first.We plan to implement features that allow people to make decisions when it is better for them to make decisions than a machine, such as adding a guarantor system. For example, if Aoba’s algorithm determines that Alice can borrow 1000 USDC at 15% interest per annum, but Bob trusts Alice and wants to lend her 12%, Bob can become Alice’s guarantor to lower Alice’s interest rate. If Alice repays the loan, Bob will receive a percentage of the reduced interest rate due to Bob’s guarantee.

We need to strengthen our partnerships with social graph, gaming and institutional investors. We plan to focus on hiring personnel. Since the most important KPI for this project is considered to be the number of borrowers developed, it is necessary to increase the number of contacts with institutional investors and DAOs, who are the main users, in addition to expanding awareness of Aoba’s credit score by collaborating with various projects.

We plan to partner with an organization with a proven track record in data science. We will continue to refine the accuracy of our models by ensuring that we always have access to a high level of technology and resources.

FAQ: Aoba Token

Q. What are the advantages of having Aoba Token?

・The ability to get a share of Protocol’s earnings.

A percentage of the interest rate is distributed to Aoba Token holders as a Protocol Fee. The more you lend, the more Aoba Token you own, the more revenue you receive.

・Participation in key decisions of Aoba.

Aoba Token holders have voting rights in important Aoba decisions, such as the distribution of Protocol Fees.

Q. Will any measures be taken to reduce selling pressure?

A. We are considering a design, like the Lock Drop implemented by Bastion, where it would be more beneficial if LPs had their tokens locked.

However, the most focus is on maximizing the value of the Protocol itself, which is the main point of attracting borrowers, alliances with other protocols, etc.

Q. Is there a possibility of dilution through capital increase?

A. Basically, we do not plan to dilute Token. On the other hand, in the event of unforeseen circumstances, such as Protocol being sued for a large amount of damages, there is a possibility that Aoba Token holders will vote for a dilutive capital increase.

FAQ: Others

Q. What positions are you currently recruiting for?

A. We are currently recruiting data scientists, engineers, and business development personnel.

Q. Are there any restrictions on applying?

A. There are no restrictions on nationality, age, or gender. We communicate in English or Japanese.

Our work schedule is fully remote and fully flexible.

Full-time and part-time positions are available.

Q. Tell us about your chain expansion.

A. We will launch on Ethereum first. Eventually, we will deploy on multiple chains, but since the data to be analyzed is in EVM, it will take time to deploy on chains other than EVM.

Q. What is your long-term vision?

A. The creation of a decentralized credit economy. In the real world, there are huge credit economies governed by governments, central banks, and other organizations. However, the concept of credit is almost non-existent in the on-chain. Aoba falls under the root layer protocol of the on-chain credit economy. After the rise of Aoba, other protocols will be developed using Aoba’s credit scores.


<For future members>

We are looking for engineers, data scientists, and business managers. If you are interested, please contact us .

<For Investors>

Aoba currently has a team of 7 people: 1 business member, 3 data scientists, 1 engineer, and 1 data analyst and web designer. We are raising funds to develop the protocol at a faster pace. If you are an investor who believes in the future of Aoba, please contact us through the link below.

Contact us at the link below:

Aoba official Twitter:

Founder Twitter:

Founder LinkedIn:



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Aoba Protocol

Under-Collateralized Lending & Borrowing Protocol Based on On-Chain Credit