This site is fictional demo content. It is not real news or affiliated with any real organization. Do not treat it as fact or professional advice.

NextPaper Logo
nextpaper.ccCYBER CHRONICLE
RSS
EN中文

Full article

FULL TEXT

View this issue
Deep diveAI

Quantum Machine Learning for Financial Risk: How Quantum Feature Mapping Detects Fraud in Millionths of a Second

Quantum ML algorithm QFraudNet surpasses classical AI in financial risk control for the first time, compressing fraud detection latency to microseconds, though quantum hardware costs remain a barrier to scale deployment.

Quantum Machine Learning for Financial Risk: How Quantum Feature Mapping Detects Fraud in Millionths of a Second

JPMorgan Chase and IBM's jointly developed quantum machine learning risk control system QFraudNet completed a six-month large-scale live test in December. Results show QFraudNet has surpassed classical deep learning models in fraud detection accuracy for the first time, while compressing inference latency to 1.2 microseconds.

QFraudNet's core innovation lies in quantum feature mapping. Traditional risk models must compress transaction data into fixed-length vectors, inevitably losing information. QFraudNet leverages quantum superposition to encode transaction data as qubit sequences, searching for fraud pattern clustering structures in high-dimensional Hilbert space.

IBM quantum researcher Maria Santos explained: Classical AI is like finding anomalies on a 2D plane. QFraudNet searches in an exponentially high-dimensional space. Certain fraud patterns overlap completely with normal transactions in low dimensions but have clear boundaries in high dimensions.

The live test covered 130 million JPMorgan users' credit card data. Over six months, QFraudNet processed 4.7 billion transactions with a fraud detection accuracy of 99.7%, a 0.4 percentage point improvement over classical models, intercepting approximately 19,000 additional fraudulent transactions and recovering roughly $320 million.

Running on IBM's latest 1,121-qubit Heron-3 processor, QFraudNet achieves single-transaction inference in 1.2 microseconds, about 98% faster than classical GPU inference at 80 microseconds.

Commercial deployment faces two challenges: hardware costs (annual maintenance of approximately $12 million per quantum computer) and quantum decoherence issues requiring cross-validation with classical models. JPMorgan's quantum technology lead Robert Smith said the team is developing a hybrid quantum-classical inference architecture expected to cover 50% of real-time risk control scenarios by late 2029.