> For the complete documentation index, see [llms.txt](https://opora.gitbook.io/opora-health-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://opora.gitbook.io/opora-health-documentation/extensions/continuous-re-identification-risk-scoring.md).

# Continuous re-identification risk scoring

This feature implements ongoing privacy audits by probing trained generative and predictive models with [membership inference attacks (MIAs)](https://medium.com/data-reply-it-datatech/membership-inference-attacks-mias-and-data-leakage-in-generative-models-737e6ed88e21) and shadow models. The tool estimates the likelihood that individual patient records can be re-identified from synthetic or model outputs, producing calibrated risk scores over time. These scores are logged into the governance knowledge graph and visualised in dashboards, enabling regulators, clinicians, and data stewards to monitor privacy leakage, set acceptable thresholds, and trigger retraining or stricter privacy budgets when risk exceeds tolerance.

**Why is it important?**

{% hint style="warning" %}
Membership Inference Attacks (MIAs) aim to infer whether a data record was used to train a target model or not. MIAs on ML models can directly lead to a privacy breach. For example, via identifying the fact that a clinical record that has been used to train a model associated with a certain disease, an attacker can infer that the owner of the clinical record has the disease with a high chance. [Source](https://dl.acm.org/doi/10.1145/3523273)
{% endhint %}