MediumRe-identification◐Demonstrated
Re-identification from outputs
Privacy & Data ProtectionDescription
Aggregated or summarised outputs allow re-identification of individuals thought to be anonymised.
Example scenario
A portfolio summary inadvertently makes a high-net-worth individual identifiable.
Real-world evidence◐Demonstrated
Facebook's shadow-profile construction for non-users (combining off-platform tracking, contact uploads, and inferred connections) is a confirmed large-scale incident. However, AI-specific shadow profiling that goes beyond conventional data aggregation — where an ML model itself synthesises disparate data points into profiles beyond user consent — is primarily demonstrated in academic work and browser fingerprinting research rather than confirmed in a BFSI production system.
Primary mitigations
- k-anonymity/aggregation thresholds
- output suppression of rare combos
- re-identification testing.
Detection signals
Re-identification risk scoring; small-cell suppression checks.
Mitigating controls
3 Non-agentic controls