Research Data De-Identifier
Detect direct and quasi-identifiers, apply masking, generalization, and pseudonymization, check k-anonymity, and export a de-identified dataset with a methods log. Your raw data never leaves your browser.
Drop your CSV file here, or click to browse
CSV and Excel files supported. Your data never leaves your browser.
Who uses Research Data De-Identifier?
Real teams solving real problems — see if your use case is here.
Academic Researcher
Social SciencePreparing a survey dataset for upload to an open-access repository. IRB approval requires the data to be de-identified before sharing, and the journal requires a methods log.
Suppresses name and email, generalizes DOB to decade age band and ZIP to 3-digit prefix, hits k=7, exports the de-identified file + methods log for the IRB appendix.
Clinical Research Coordinator
HealthcareSharing a patient cohort extract with a collaborating institution. Cannot upload the file to any third-party SaaS because it contains PHI.
Runs the entire de-identification in the browser with no upload. Pseudonymizes patient IDs, masks diagnosis codes, checks k-anonymity on age + ZIP. Exports for transfer.
Graduate Student
Public HealthThesis dataset contains participant names, exact DOBs, and home ZIP codes. Committee requires the shared version meet k≥5 before committee members can review the raw results.
Age bands + 3-digit ZIP raise k from 1 to 6. Exports the de-identified version for committee review and the methods log for the appendix.
Data Librarian
University LibraryAssisting a faculty member in preparing a dataset for the institutional repository. Needs to verify de-identification before accession.
Runs the tool, reviews the column classification, confirms k≥3 with the faculty member, and deposits the de-identified version with the auto-generated methods log.
How to de-identify a research dataset
Upload your CSV or Excel file
Drag and drop your dataset. The tool auto-classifies each column as a direct identifier, quasi-identifier, sensitive attribute, or safe — based on column names and patterns.
Review and configure transforms
Override any auto-classification. Choose a per-column action: suppress the column entirely, mask values, pseudonymize (stable fake IDs), generalize dates to age bands, generalize ZIP codes to 3-digit prefixes, or top/bottom-code outliers.
Check k-anonymity (optional)
Select your quasi-identifier columns and a target k. k=5 means no individual can be singled out from the combination of quasi-identifier values alone. The tool shows the current k and specific generalization suggestions if you fall short.
Export the de-identified file and methods log
Download the transformed dataset as CSV, plus a plain-text methods log describing every transform applied — suitable for an IRB application or published dataset appendix.
Why in-browser de-identification matters for human-subjects data
Most de-identification tools upload your file to a server. For any dataset containing human-subjects data, health information, or data covered by GDPR or HIPAA, that upload is itself a data transfer — and often one that requires a data processing agreement with the vendor. Researchers frequently cannot share data with a SaaS tool at all before de-identification.
This tool runs entirely in your browser. Your raw data never leaves your device. There is no server, no account, no third-party data processor. The de-identification happens locally, and only the already-de-identified output is saved anywhere.
The k-anonymity check implements the model described by Samarati and Sweeney (1998): a dataset is k-anonymous when every combination of quasi-identifier values appears in at least k rows. k=5 is a common threshold for research data repositories; HIPAA Safe Harbor de-identification requires specific field removal rather than a k threshold, but k-anonymity is a useful complement for verifying quasi-identifier combinations.