Data Anonymization via Event Sequence Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data anonymity methods using k-anonymity algorithms lose significant information when applied to event logs, as they fail to effectively anonymize datasets containing quasi-identifiers and event sequences without substantial data loss.
Innovation Solution
A data anonymity method and system that replaces direct-identifiers with pseudonyms, classifies quasi-identifiers using a group-by algorithm, links activities based on timestamps to form event sequences, constructs a similarity hierarchy tree, and groups event sequences using k-anonymity to minimize information loss, allowing for generalized event sequences without losing critical data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional k-anonymity algorithm is applied to event logs, then data anonymity is achieved, but significant information is lost
Solution Approach 1:
The patent segments the data processing into distinct phases: first processing direct-identifiers and quasi-identifiers separately, then handling event sequences independently. This segmentation allows event sequences to be anonymized without being forced through the traditional quasi-identifier generalization process that causes information loss.
Solution Approach 2:
The patent introduces a new dimension for event sequence anonymization by creating equivalence classes based on event sequence similarity rather than traditional quasi-identifier values. This dimensional shift allows preserving the temporal and sequential characteristics of events while achieving anonymity.
2Reliability
If event sequences are anonymized using traditional generalization, then k-anonymity is achieved, but variability of event sequences is lost
Solution Approach 1:
The patent changes the parameters used for equivalence class formation from traditional quasi-identifier values to event sequence characteristics (activities and timestamps). This parameter change allows maintaining event sequence variability while achieving k-anonymity through similarity-based grouping.
3Object-affected harmful factors
If direct-identifiers are replaced by pseudonyms, then identification privacy is protected, but data linkage capability is reduced
Solution Approach 1:
The patent uses pseudonyms as intermediaries that preserve linkage capability. By replacing direct-identifiers with pseudonyms while maintaining consistent mapping, the system protects identification privacy but retains the ability to link events to the same entity across different records through the pseudonym key.
Data Source
AI summary
A data anonymity method and a data anonymity system are provided. The data anonymity method includes the following steps. A data set comprising a plurality of direct-identifiers, a plurality of quasi-identifiers and a plurality of event logs each of which includes an activity and a timestamp is obtained. A content of each of the direct-identifiers is replaced by a pseudonym. The quasi-identifiers are classified, via a group-by algorithm with k-anonymity, as a plurality of equivalence classes. The activities corresponding to each of the direct-identifiers are linked according to the timestamps to obtain a plurality of event sequences. A similarity hierarchy tree is obtained according to a plurality of edit distances among the event sequences. The event sequences are grouped according to the similarity hierarchy tree with k-anonymity to obtain at least one group. The event sequences which are in the group are generalized.