Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Table 1.

PDPL privacy concepts and operational interpretation.

More »

Table 1 Expand

Fig 1.

Linkage-based re-identification risk in integrated data systems.

The figure illustrates how quasi-identifiers from multiple organizational systems can be combined by an adversary to re-identify individuals. It highlights the increased privacy exposure caused by cross-system data integration and motivates the need for risk assessment in integrated environments.

More »

Fig 1 Expand

Table 2.

Employee Dataset (EMP-DS) attributes and categories.

More »

Table 2 Expand

Table 3.

Employee Dataset (EMP-DS) data model.

More »

Table 3 Expand

Table 4.

Class instance and the ReID risk in Employee Dataset (EMP-DS) sample.

More »

Table 4 Expand

Fig 2.

Record-level re-identification risk in the employee dataset (EMP-DS).

Each bar represents the calculated prosecutor-style re-identification risk for an individual record based on indirect identifiers (Job Title and Gender). Higher values indicate greater vulnerability to linkage attacks.Calculate The Overall Risk for The Population Dataset.

More »

Fig 2 Expand

Table 5.

Maximum and average reidentification probability and uniqueness.

More »

Table 5 Expand

Table 6.

Comparison of uniqueness thresholds.

More »

Table 6 Expand

Fig 3.

Exact-match linkage success under Local Differential Privacy.

The figure shows exact-match linkage success rates for varying privacy budgets (ε). Lower ε values provide stronger privacy protection and substantially reduce the attacker’s ability to link records.

More »

Fig 3 Expand

Fig 4.

Distributional distortion under Local Differential Privacy.

Average Total Variation (TV) distance between raw and LDP-perturbed attribute distributions across different ε settings. Lower distances indicate higher statistical utility.

More »

Fig 4 Expand

Fig 5.

Classification accuracy under LDP-perturbed features.

Classification accuracy for predicting Job Title using LDP-perturbed features. Accuracy improves as ε increases, reflecting the privacy–utility trade-off.

More »

Fig 5 Expand

Fig 6.

Weighted F1-score under LDP-perturbed features.

Weighted F1-score for Job Title prediction across different LDP privacy budgets. Larger ε values improve predictive performance while maintaining privacy guarantees.

More »

Fig 6 Expand

Fig 7.

Effect of repeated data releases on linkage success.

The figure shows the attacker’s exact-match linkage success under repeated data releases with fixed ε. Linkage success increases as the number of releases grows, demonstrating cumulative privacy loss.

More »

Fig 7 Expand

Fig 8.

Reduction of uniqueness after applying Local Differential Privacy.

Uniqueness levels in the sample dataset before and after applying LDP. The results show that LDP significantly reduces linkability and supports PDPL compliance.

More »

Fig 8 Expand

Table 7.

Key differences and complementarity between Arcolezi et al. (PVLDB 2023) and this paper.

More »

Table 7 Expand