Encrypted Credit Card Approval Prediction Using Fully Homomorphic Encryption

When you apply for a credit card, banks assess your financial details like income, job status, and debts with help from credit bureaus and scoring agencies, who calculate a credit score based on your credit history. This process involves sharing sensitive data with multiple parties, increasing the risk of breaches or errors. Fully Homomorphic Encryption (FHE) addresses this by allowing secure analysis of encrypted data, ensuring privacy while improving accuracy. With FHE, your information stays protected throughout the process, reducing risks and helping banks make more informed credit decisions.

Step 1: Enter Information.

Applicant, Bank and Credit bureau setup

Set up the applicant, bank, and credit bureau profiles by selecting the relevant details to evaluate credit card eligibility. This model uses three data sources:

  • Personal Information: For assessing the applicant’s creditworthiness.
  • Bank Account History: Including key details like how long the account has been active.
  • Credit Bureau Information: Additional data, such as employment history, to provide further insights.

After making updates, remember to encrypt and send the information using the buttons on the right before running the FHE process.

Step 1.1 - Applicant information 🧑‍💻

Which of the following do you actively hold or own?

Number of children

How many children do you have ?

0

Household size

How many members does your household have ?

0

Income

What's you total yearly income (in euros) ?

0

Age

How old are you ?

0

What is your main type of income ?

What is your education background ?

What is your family status ?

What is your main occupation ?

In what type of housing do you live ?

Step 1.2 - Bank information 🏦

Account age (months)

How long have this person had this bank account (in months) ?

6

Step 1.3 - Credit bureau information 🏢

Is the person employed ?

How long have this person been employed (in years) ?

Step 2: Run the FHE evaluation.

Server Side

Once the encrypted data is sent, the server uses a Decision Tree model to make predictions—without ever needing to decrypt the information.

Step 3: Receive and Decrypt the Output.

Applicant, Bank and Credit bureau decryption

After the server finishes processing, the encrypted result is sent back to the applicant. Only the applicant, bank, and credit bureau can decrypt the result, as each holds a part of the decryption key. Once all three parties decrypt their shares, the applicant can use their private key to view the final result.