The EOSC-SIESTA project is developing practical tools to support the secure analysis and sharing of sensitive data across different scientific and technological domains. One of these tools focuses on the anonymization of sensitive textual data, addressing a key challenge for organizations that work with confidential information and need to reuse data safely for research, analysis, or machine learning purposes.
This prototype has been developed within Use Case 4 of the EOSC-SIESTA project by the Text Anonymization team from Universidad de León. The tool is designed to help incident response teams prepare cyber incident reports that can be safely shared for purposes such as research, collaborative threat analysis, and collective cyber defence.
Cyber incident reports often contain sensitive information that may be linked to individuals, organizations, or specific events. This can include names of victims, locations, contact details, institutions, dates, or other identifying references. For this reason, these reports cannot usually be shared in their original form. To address this problem, the tool combines Named Entity Recognition AI models with four anonymization techniques to automatically detect and replace sensitive information, allowing the content of the report to be reused while protecting the identities of the people and entities involved.
The tool allows users to anonymize cyber incident reports directly from the interface. Users can paste a single report or upload several reports in Excel or CSV format, making it possible to work both with individual texts and larger datasets.
The prototype currently offers four anonymization techniques: data tokenization, data masking, data substitution, and data hashing. Tokenization replaces sensitive entities with labels indicating their type; masking hides them with fixed tokens; substitution replaces them with plausible but false alternatives; and hashing generates deterministic codes that make it possible to track repeated entities across documents without revealing their real identity.
These options provide flexibility depending on the intended use of the anonymized data, from classification tasks and machine learning training to secure information sharing and cross-document analysis.
The tool is expected to be further integrated into the EOSC-SIESTA ecosystem. Public repositories will soon be available, providing access to the tool’s code, documentation, and technical resources. In addition, the tool will also be available through the SIESTA Dashboard, making it easier for users to access and test it within the project’s secure environment.
Overall, the text anonymization tool developed within EOSC-SIESTA contributes to one of the project’s central objectives: enabling the secure use and sharing of sensitive data while preserving its value for analysis. By offering different anonymization techniques, the tool provides flexibility depending on the final purpose of the data, whether that is classification, machine learning training, cross-document analysis, or secure information sharing.