Re-Identification Risk Evaluation Using k-NN Distance Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anonymization methods lack a comprehensive and reliable method to evaluate the risk of re-identification of anonymized data, particularly in the context of distance-based correspondence-seeking attacks, which can lead to imperfect protection against re-identification.
Innovation Solution
A computer-implemented method that transforms original and anonymized data into a Euclidean space using factor analysis or auto-encoder neural networks, calculates Euclidean distances, and applies the k-NN method to identify nearest neighbors, determining a protection rate based on the density of non-valid anonymous individuals around each original individual.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If anonymization algorithms modify data by deleting, generalizing, or replacing personal information, then privacy protection is improved, but information content and data utility deteriorate
Solution Approach 1:
The patent replaces traditional mechanical anonymization operations (deletion, generalization, replacement) with a neural network-based system that learns optimal transformation patterns. The auto-encoder neural network transforms data in a way that preserves information structure while achieving anonymization, substituting rule-based mechanical operations with intelligent learned transformations.
Solution Approach 2:
The patent changes the parameters of anonymization by introducing a learnable transformation process through neural networks. Instead of fixed anonymization rules, the system learns optimal parameter transformations that balance privacy protection with information preservation, allowing dynamic adjustment of anonymization intensity based on data characteristics.
2Ease of operation
If existing anonymization methods are used without comprehensive risk evaluation, then data sharing is simplified, but re-identification risk increases
Solution Approach 1:
The patent implements a feedback mechanism by evaluating the anonymized data through the same neural network model to assess re-identification risk. This feedback loop allows the system to verify whether the anonymization achieved the desired protection level and to adjust parameters if necessary, ensuring reliability while maintaining ease of operation.
Solution Approach 2:
The patent performs preliminary risk evaluation as part of the anonymization process itself. By integrating the evaluation step within the anonymization workflow, the system proactively identifies and addresses potential re-identification risks before data sharing occurs, rather than treating evaluation as a separate post-processing step.
3Device complexity
If distance-based correspondence-seeking attacks are not considered in anonymization, then anonymization process is simpler, but protection against re-identification is insufficient
Solution Approach 1:
The patent applies preliminary anti-action by training the neural network to recognize and counteract distance-based correspondence-seeking attacks during the anonymization process. The model learns to transform data in ways that specifically prevent attackers from using distance metrics to re-identify individuals, building defense against these attacks into the anonymization itself.
Data Source
AI summary
(EN) The method of the invention provides a protection rate (txP2) representative of the risk of re-identification of data. In the case of a distance-based correspondence-seeking attack, the method comprises the steps of: a) linking an original dataset (EDO) comprising a plurality of original individuals (IO) with an anonymised dataset (EDA) comprising a plurality of anonymised individuals (IA); b) transforming (PCA, MCA, FAMD) the original individuals and the anonymous individuals in a Euclidean space; c) identifying for each original individual, one or more nearest anonymous individuals based on a distance, by a method referred to as the “k-NN” method; and d) calculating the protection rate, being a mean number (Nm) of anonymous individuals, nearest to a considered original individual (IOi), who are not a valid anonymous individual corresponding to the original individual considered, the nearest anonymous individuals being those identified in step c) and having a distance (dy) relative to the considered original individual less than the distance between the considered original individual and the valid anonymous individual.