Abstract:The financial industry deals with a vast amount of sensitive data in its business operations.However, the conventional approach of binding financial data for desensitization and using desensitization algorithms is becoming inefficient due to the fast-paced growth of financial businesses and the proliferation of data types.Additionally, manual verification and assessment of desensitized data by security experts are time-consuming and may carry potential privacy risks due to the improper selection of desensitization algorithms.While prior research has emphasized desensitization methods and privacy-preserving technologies, limited work has been conducted on desensitization algorithms from the perspective of automation.To address this issue, an adaptive recommendation framework was propose for selecting desensitization strategies that consider various factors, such as existing privacy protection technologies, data quality requirements of business scenarios, security risk requirements of financial institutions, and data attributes.Specifically, a dual-objective evaluation function was established for privacy risk and data quality to optimize the selection of desensitization algorithm parameters for different algorithms.Furthermore, the desensitization algorithm and parameters were adaptively selected by considering the data attributes through a multi-decision factor system and desensitization effect evaluation system.Compared to traditional approaches, the proposed framework effectively tackle issues of reduced data usability and inadequate personal data privacy protection that derive from manual intervention.Testing on a dataset with multiple financial institution types, the experiments show that the proposed method achieves a recommendation accuracy exceeding 95%, while the desensitized privacy risk level differed by less than 10% from the expected level.Additionally, the recommendation efficiency is 100 times faster than expert manual processing.

Data Sanitization Based on Weaken_Sensitive_Pattern Tree

Hiding Sensitive Sequential Patterns by Computing Impact Weight

An Effective Approach for Hiding Sensitive Knowledge in Data Publishing

On the Vulnerability of Text Sanitization

Preserving Private Knowledge in Frequent Pattern Mining

Mining Sequential Patterns Based on Privacy Preserving

Sensitive-rule Sanitization Algorithm Based on Computing Dynamic Conflict Degree

An Improved EDP Algorithm to Privacy Protection in Data Mining.

An Optimized Sanitization Approach for Minable Data Publication

ID-based sanitizable signature data integrity auditing scheme with privacy-preserving

Big Data Set Privacy Preserving Through Sensitive Attribute-Based Grouping.

A novel method for preventing the disclosure of restrictive patterns

A Chunked and Disordered Data Privacy Protection Algorithm: Application to Resource Platform Systems

A Scheme of Privacy Protection Based on Genetic Algorithm for Behavior Pattern of Social Media Users

Research on Desensitization Algorithm of Power Consumption Data Based on Anonymity and Differential Privacy Technology

Decouple-and-Sample: Protecting sensitive information in task agnostic data release

Clustering-assisted privacy perseveration model for data mining

Adaptive Selection Method of Desensitization Algorithm Based on Privacy Risk Assessment

Privacy Preserving Data Mining on Decision Tree

Stronger Data Poisoning Attacks Break Data Sanitization Defenses

Sensitive Label Privacy Preservation with Anatomization for Data Publishing