The system detects fraudulent transactions that have never been seen before and explains the reason for each detection in plain language.
JAPAN, November 18, 2026: Urata Soft Co., Ltd. has been granted Japanese Patent No. 7944650, “Anomalous Transaction Detection and Notification System Using Quantum Circuits and Anomalous Transaction Detection and Notification Method Using Quantum Circuits” (registered on October 26, 2026).
The company is seeking buyers and licensees for its patent portfolio, which includes this patent (three granted patents and two pending applications).
Background:
Losses from bank fraud and scams worldwide reached US$579.4 billion in 2025 (Payments Dive).
Card fraud losses worldwide were US$33.41 billion in 2024 (Nilson Report).
Conventional fraud detection systems used for credit card payments, bank transfers and securities trading can detect fraud schemes that have appeared before, but they miss schemes that have never been seen before.
This system is a state-of-the-art challenger to the conventional IDS (Intrusion Detection System). An IDS checks transactions one by one against known signatures, while the quantum oracle of this patent applies the detection condition to all combinations at once.
Features of the System:
(1) Viewpoints: For example: the system examines each transaction from five viewpoints. Whether the amount is high, whether the time is late at night, whether the destination is high-risk, whether the card is used frequently, and whether the payment is split into amounts just below the limit.
(2) Quantum detection: Among the 32 combinations of the five viewpoints, Grover search finds the combinations that match fraud patterns. The system outputs the probability that the transaction matches a fraud pattern as a “quantum score.” The Quantum Fourier Transform converts irregularity in the cycle of transaction times into a number.
(3) Combination with AI: The system combines the quantum score with the result of an AI model that finds statistical outliers (Isolation Forest), and it decides whether the transaction is anomalous.
(4) Rescue of missed fraud: The system also compares transactions judged as normal with confirmed past fraudulent transactions in terms of amount, time, destination risk and frequency of use. If a transaction closely resembles past fraud, the system changes its judgment to anomalous.
(5) Notification of reasons: The system uses SHAP to identify the items that decided the judgment, and it notifies the person in charge of the fraud type, the recommended action and an explanation.
(6) Operating environment: The system runs on standard CPUs.
Test Results (Proof of Concept):
Data: The public Kaggle dataset “Credit Card Transactions Fraud Detection Dataset” (simulated U.S. card transactions, about 1.3 million records)
Preprocessing: The company extracted 1,500 fraudulent transactions and 1,500 normal transactions (3,000 in total) and prepared four items: amount, time, destination risk and frequency of use.
Results: detection by (2) and (3) only ? with the rescue in (4) added
- Fraudulent transactions detected out of 1,500 (recall): 839 (55.9%) ? 1,244 (82.9%)
- Share of transactions judged anomalous that were actually fraudulent (precision): 69.5% ? 77.1%
- AUC (an index of how well the system separates fraudulent transactions from normal ones; closer to 1 is better): 0.7622 ? 0.7674
Product and Source Code:
The system is sold on AWS Marketplace as the product “Quantum Anomaly Transaction Detection System.”
The source code is uploaded to GitHub (the repository is currently private and is not displayed).
Patent Portfolio:
The company welcomes inquiries about the sale or licensing of all of the following.
Granted
- Japanese Patent No. 7944650: Anomalous Transaction Detection and Notification System Using Quantum Circuits and Anomalous Transaction Detection and Notification Method Using Quantum Circuits
Espacenet: https://worldwide.espacenet.com/patent/search?q=JP7944650 - Japanese Patent No. 7791512: Credit Scoring System and Model Generation System
Google Patents: https://patents.google.com/patent/JP7791512B1 - Japanese Patent No. 7859641: Information Processing System and Information Processing Method for Vectorizing the Meaning of Words and Searching for Reasons and the Like Based on Vector Similarity
Google Patents: https://patents.google.com/patent/JP7859641B1
Pending
- Japanese Patent Application No. 2026-181432: Quantum-Cryptography Hybrid Secure Computation System, Quantum-Cryptography Hybrid Secure Computation Method, and Program Therefor
- Japanese Patent Application No. 2026-228366: Cryptographic Replacement System, Cryptographic Replacement Method and Program for Replacing Cryptography with Post-Quantum Cryptography Without Decryption and Without Downtime
About Urata Soft Co., Ltd.:
Urata Soft contributes to the development of industry through the development of patented technologies based on AI and quantum principles.
Website: https://www.urata-soft.org/
Media Contact: Tetsuo Urata, CEO

Join the Discussion
No comments yet.