Partitioning based incremental marginalization algorithm for anonymizing missing data streams

The IoT and its applications are the inseparable part of modern world. IoT is expanding into every corner of the world where internet is available. IoT data streams are utilized by many organizations for research and business. To benefit from these data streams, the data handling party must secure the individuals’ privacy. The most common privacy preservation approach is data anonymization. However, IoT data provides missing data streams due to the varying device pool and preferences of individuals and unpredicted devices’ malfunctions of IoT. Minimization of missingness and information loss is very important for anonymizing of missing data streams. To achieve this, we introduce IncrementalPBM (Incremental Partitioning Based Marginalization) for anonymizing missing data streams. IncrementalPBM utilizes time based sliding window for missing data stream anonymization, and it aims to control the number of QIDs for anonymization while increasing the number of tuples for anonymization. Our experiment on real dataset showed IncrementalPBM is effective and efficient for anonymizing missing data streams compared to existing missing data stream anonymization algorithm. IncrementalPBM showed significant improvement; 5% to 9% less information loss, 4500 to 6000 more number of re-use anonymization while showing comparable clustering, suppression and runtime.