adulau/imsi-dataset

Format-Preserved Pseudonymised IMSI-Catcher Observations

Dataset Description

This dataset contains pseudonymised mobile-network observation records derived from IMSI-catcher-style telemetry. It is intended for research on mobile-network measurements, cellular security, anomaly detection, roaming patterns, and privacy-preserving analysis of telecom metadata.

The dataset has been transformed using deterministic, keyed, format-preserving pseudonymisation (Format-preserving pseudonymisation for IMSI-catcher observation datasets tool). The output keeps the same field names and general value formats as the original observations, while replacing sensitive identifiers and serving-location values with consistent synthetic equivalents.

The timestamp field is preserved unchanged to support temporal analysis.

Important Privacy Notice

This dataset is pseudonymised, not guaranteed anonymous.

Direct identifiers such as raw IMSI, TMSI, MCC, MNC, LAC, and cell identifiers have been transformed. However, mobility traces, timestamps, rare observation patterns, repeated sightings, and combinations of fields may still create re-identification or linkage risks.

Researchers must not attempt to re-identify individuals, subscribers, operators, physical locations, devices, or original radio infrastructure.

Intended Use

This dataset may be useful for:

• studying cellular observation patterns;
• evaluating IMSI-catcher detection methods;
• testing telecom-security analytics pipelines;
• researching privacy-preserving publication of mobile-network datasets;
• developing visualisation, clustering, and anomaly-detection methods;
• teaching mobile-network metadata analysis with reduced exposure to raw identifiers.

Out-of-Scope Use

This dataset must not be used for:

• identifying or tracking individuals;
• inferring the real-world location of captured observations;
• deanonymising subscribers, devices, operators, or cells;
• operational surveillance;
• targeting specific people, groups, networks, or infrastructure;
• commercial profiling or advertising;
• law-enforcement or intelligence use without appropriate legal basis and independent review.

Dataset Format

Each record is a JSON object with the same field layout as the source observations.

Example record:

{
  "_table": "observations",
  "stamp": "2020-05-10 15:53:19.961716",
  "tmsi1": null,
  "tmsi2": null,
  "imsi": "734 42 8392018472",
  "imsicountry": "Country-734",
  "imsibrand": "Brand-128403",
  "imsioperator": "Operator-991204",
  "mcc": 734,
  "mnc": 18,
  "lac": 51902,
  "cell": 12044
}

Field Description

Field	Type	Description
_table	string	Original table/source category, preserved when present.
stamp	string	Original observation timestamp. This field is not modified.
tmsi1	string or null	Format-preserved pseudonym of the first TMSI-like value, or null.
tmsi2	string or null	Format-preserved pseudonym of the second TMSI-like value, or null.
imsi	string or null	Format-preserved pseudonym of the IMSI. Grouping and digit lengths are preserved where possible.
imsicountry	string or null	Pseudonymous home-country label derived from the original IMSI metadata.
imsibrand	string or null	Pseudonymous brand label derived from the original IMSI metadata.
imsioperator	string or null	Pseudonymous operator label derived from the original IMSI metadata.
mcc	integer or null	Synthetic serving MCC-like value. This does not represent the real observed MCC.
mnc	integer or null	Synthetic serving MNC-like value. This does not represent the real observed MNC.
lac	integer or null	Synthetic LAC-like value. This does not represent the real observed LAC.
cell	integer or null	Synthetic cell-like value. This does not represent the real observed cell.

Pseudonymisation Method

The dataset was transformed using a deterministic keyed pseudonymisation process.

The transformation preserves:

• field names;
• JSON structure;
• null values;
• timestamp values;
• IMSI digit grouping, where present;
• integer-like format for mcc, mnc, lac, and cell;
• consistency across records.

The transformation does not preserve:

• real subscriber identifiers;
• real serving MCC/MNC values;
• real LAC or cell identifiers;
• real operator names when label pseudonymisation is enabled;
• geographic meaning of serving-location fields.

The secret pseudonymisation key is not included in this repository.

Consistency Guarantees

Within this release:

• the same original IMSI maps to the same pseudonymous IMSI;
• the same original serving network maps to the same synthetic serving network;
• the same original LAC within a serving network maps to the same synthetic LAC;
• the same original cell within a LAC maps to the same synthetic cell;
• repeated observations remain linkable within the released dataset.

Across different releases, consistency depends on the pseudonymisation key and release identifier used by the data publisher.

Known Limitations

Although direct identifiers are transformed, the dataset may still contain sensitive structure. In particular:

• timestamps are unchanged;
• rare sequences of observations may remain unique;
• mobility or co-occurrence patterns may remain linkable;
• the dataset may not be representative of all mobile networks, operators, regions, or devices;
• pseudonymous MCC/MNC/LAC/cell values must not be interpreted as real geography;
• pseudonymous country, brand, and operator labels are analytical labels, not ground truth about the transformed identifiers.

Ethical and Responsible Use

Users of this dataset are expected to:

• use the data only for legitimate research, education, or defensive security purposes;
• avoid re-identification attempts;
• avoid combining this dataset with external data to infer real persons, locations, cells, or operators;
• report privacy concerns or suspected leakage to the dataset maintainers;
• comply with applicable privacy, data-protection, telecommunications, and research-ethics requirements.

Contact

For questions, responsible disclosure, or requests related to this dataset, contact:

Alexandre Dulaunoy -> foo.be

Changelog

v1.0

• Initial public release.
• Format-preserving pseudonymisation applied to IMSI, TMSI-like values, serving network, LAC, and cell fields.
• Timestamps preserved for temporal research.