Bavarian Microdata Dataset for ARX Data Anonymization Tool — Teaching & Demo Dataset

Description This dataset is a synthetic, fictitious microdata record designed for educational use with the ARX Data Anonymization Tool(https://arx.deidentifier.org/) (version ≥ 3.9). It simulates a small patient registry from the Ansbach district , Bavaria, Germany, and is intended to demonstrate the full ARX anonymization workflow — including attribute classification, generalization hierarchy configuration, and evaluation of privacy models such as k-Anonymity, l-Diversity, t-Closeness, and Differential Privacy. **All records are entirely synthetic. No real individuals are represented. The dataset was created solely for teaching and demonstration purposes.** Dataset Overview | Property | Value ||---|---|| File name | `microdata_bavaria.csv` || Format | CSV (comma-separated, UTF-8) || Records | 100 (+ 1 header row) || Columns | 5 || Geographic scope | Ansbach district, Bavaria, Germany (ZIP codes 915xx – 916xx) || Disease categories | Bone fractures, Diabetes & metabolic conditions || Intended tool | ARX Data Anonymization Tool | Column Description | Column | Type in ARX | Description ||---|---|---|| `Name` | **Explicit Identifier (EI)** | Fictional first name. Must be removed or suppressed before any anonymized release. Not used for generalization. || `Age` | **Quasi-Identifier (QI)** | Integer age in years (range: 18–85). Can be linked to external sources for re-identification. Accompanied by a 5-level generalization hierarchy (`hierarchy_age_bavaria.csv`): exact age → 5-year bands → 10-year bands → 20-year bands → 18–85 → `*`. || `Sex` | **Quasi-Identifier (QI)** | Biological sex, values: `M` / `F`. Accompanied by a 1-level suppression hierarchy (`hierarchy_sex_bavaria.csv`). || `Zipcode` | **Quasi-Identifier (QI)** | 5-digit German postal code from the Ansbach region (e.g., `91522`, `91604`). Accompanied by a 5-level prefix-masking hierarchy (`hierarchy_zipcode_ansbach.csv`): exact → 4-digit prefix → 3-digit prefix → 2-digit prefix → 1-digit prefix → `*****`. A separate suppression-only hierarchy (`hierarchy_zipcode_suppression.csv`) is also included. || `Disease` | **Sensitive Attribute (SA)** | Diagnosed condition, drawn from two clinically coherent groups: **bone fractures** (e.g., Hip Fracture, Vertebral Fracture, Tibia Fracture) and **metabolic/diabetes conditions** (e.g., Type 2 Diabetes, Hypoglycemia, Diabetic Neuropathy). A 3-level semantic hierarchy (`hierarchy_disease_bavaria.csv`) maps individual diagnoses → subcategory (e.g., *Lower Limb Fracture*, *Diabetic Complication*) → ICD chapter group (*Bone Fracture (ICD S/M)*, *Metabolic Disease (ICD E)*) → `*`. | Accompanying Generalization Hierarchies The following CSV hierarchy files are included and ready for import into ARX: | File | Attribute | Levels ||---|---|---|| `hierarchy_age_bavaria.csv` | Age | 5 (exact → 5yr → 10yr → 20yr → 18–85 → `*`) || `hierarchy_sex_bavaria.csv` | Sex | 1 (M/F → `*`) || `hierarchy_zipcode_bavaria.csv` | Zipcode | 5 (prefix masking) || `hierarchy_disease_bavaria.csv` | Disease | 3 (diagnosis → subcategory → ICD group → `*`) | Intended Use This dataset is suitable for: Teaching: k-Anonymity, l-Diversity, t-Closeness, and Differential Privacy in university courses on data privacy, information security, or data science Demonstrating: the complete ARX workflow: data import → attribute configuration → hierarchy setup → privacy model selection → lattice exploration → anonymized export Hands-on lab exercises: where students can experiment with different suppression limits, privacy parameters, and information loss metrics Illustrating re-identification risks: through quasi-identifier combinations (Age + Sex + Zipcode) in a realistic but non-sensitive context Related Resources ARX Data Anonymization Tool: https://arx.deidentifier.org/ ARX GitHub Repository: https://github.com/arx-deidentifier/arx