# Falcon Fraud Detection
**Entity class:** Payment-card fraud detection system
## Definition and history
HNC's Falcon fraud-detection system dates to 1992 in the company's account, while later agreements describe institutional data supplied for model updates. [[articles/Intelligence Inside|Intelligence Inside]] uses it to show how separate payment streams could be pooled into a wider risk picture. A transaction's score influences review or intervention, and the resulting disposition can become a later label. The 2003 consortium filing describes monthly data provision, which is a concrete retraining cadence rather than an assumed instantaneous learning loop.
## Mechanism and limits
The system joins transaction features, model inference, operational thresholds, and institutional feedback. Performance depends on the distribution of legitimate and fraudulent behavior and on which reviewed cases receive confirmed labels. The architecture is a counterpart to DIF in finance, though the operators, objectives, and legal conditions differ.
## Relationships
**Company:** HNC Software. **Function:** [[wiki/Classifier|Classifier]] and [[wiki/Decision Threshold|Decision Threshold]]. **Comparison:** [[wiki/IRS Discriminant Function|IRS Discriminant Function]]. **Problem:** [[wiki/Selective Labels|Selective Labels]]. **Article route:** [[articles/Intelligence Inside|Intelligence Inside]].
## Sources
[HNC Software SEC filing and Falcon history](https://www.sec.gov/Archives/edgar/data/814547/000095014902000962/f81563xe425.htm); [[articles/Intelligence Inside|Intelligence Inside]].