Authors:
Vikrant Maheshwari, Rhonda Foshee, Adrian Rodriguez, Barney Barnett, Vipul Chotaliya, Karoline Gayla, Paul Scrutton
Abstract:
Disclosed is a system and method to correlate and analyze transactions across multiple retailers using anonymized shopper information to detect fraudulent behavior or scam activity. The system identifies if a shopper is unknowingly a scam victim by analyzing transaction characteristics while maintaining user privacy.
Background:
Online and retail scams are increasing, often targeting unaware customers. Fraudulent actors typically instruct victims to purchase high-value gift cards (e.g., Amazon, Visa, Amex) across multiple retailers to avoid triggering individual store alerts. Existing systems do not detect fraud distributed across stores, nor do they effectively notify customers unless tied to loyalty programs.
This invention addresses the need for a privacy-preserving, cross-retailer system that flags unusual or scam-related transactions by recognizing behavioral trends using biometric data such as anonymous facial features or optional identification like fingerprints or driver’s licenses.
Description:
The system works as follows:
1. Data Capture:
- When a shopper performs a transaction, facial attributes (not actual photos) are captured.
- Item details (e.g., quantity of gift cards, medications, etc.) are recorded.
2. Anonymized Linking Across Retailers:
- Facial attributes or optional personal identifiers are used to anonymously link transactions across stores.
3. Fraud Trend Analysis:
- Aggregated transactions are compared against scam behavior databases.
- Suspicious patterns (e.g., multiple high-value gift card purchases + international shipping) increase a fraud confidence score.
4. Customer Interaction:
- As fraud likelihood increases, the shopper receives soft warnings (e.g., “Are you sure you need these gift cards?”).
- If scam is likely, the system prompts the shopper to optionally de-anonymize and provide more information.
5. Escalation:
- Prompts may include requests for:
- A photo or license
- Reason for purchase
- Emotional state via image detection (e.g., if someone is coercing them)
- Staff may be alerted for in-store intervention.
Alternate Embodiments:
- Using OCR and metadata from product images to match items across stores
- Embedding rich metadata in shopping lists for better cross-store substitution and fraud context
Usages:
- Fraud detection in cross-retail transactions
- Scam prevention at POS
- Behavioral pattern detection using biometric data
- Anonymous shopper alert systems
Enabling Technology:
- Biometric (non-identifying) facial recognition
- Controlled item POS monitoring
- Metadata and OCR integration
- Cross-retailer data-sharing services
Relevant Existing Art:
- Pharmacy systems tracking controlled medications
- Fingerprint authentication for large banking transactions
- Anonymous facial recognition in surveillance
- Lottery machine behavior detection systems
TGCS Reference 2167
