Utility-based Privacy Preserving Data Mining
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916957677158400 |
|---|---|
| author | Zhou, Qingfeng Gan, Wensheng Qi, Zhenlian Yu, Philip S. |
| author_facet | Zhou, Qingfeng Gan, Wensheng Qi, Zhenlian Yu, Philip S. |
| contents | With the advent of big data, periodic pattern mining has demonstrated significant value in real-world applications, including smart home systems, healthcare systems, and the medical field. However, advances in network technology have enabled malicious actors to extract sensitive information from publicly available datasets, posing significant threats to data providers and, in severe cases, hindering societal development. To mitigate such risks, privacy-preserving utility mining (PPUM) has been proposed. However, PPUM is unsuitable for addressing privacy concerns in periodic information mining. To address this issue, we innovatively extend the existing PPUM framework and propose two algorithms, Maximum sensitive Utility-MAximum maxPer item (MU-MAP) and Maximum sensitive Utility-MInimum maxPer item (MU-MIP). These algorithms aim to hide sensitive periodic high-utility itemsets while generating sanitized datasets. To enhance the efficiency of the algorithms, we designed two novel data structures: the Sensitive Itemset List (SISL) and the Sensitive Item List (SIL), which store essential information about sensitive itemsets and their constituent items. Moreover, several performance metrics were employed to evaluate the performance of our algorithms compared to the state-of-the-art PPUM algorithms. The experimental results show that our proposed algorithms achieve an Artificial Cost (AC) value of 0 on all datasets when hiding sensitive itemsets. In contrast, the traditional PPUM algorithm yields non-zero AC. This indicates that our algorithms can successfully hide sensitive periodic itemsets without introducing misleading patterns, whereas the PPUM algorithm generates additional itemsets that may interfere with user decision-making. Moreover, the results also reveal that our algorithms maintain Database Utility Similarity (DUS) of over 90\% after the sensitive itemsets are hidden. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_15755 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Utility-based Privacy Preserving Data Mining Zhou, Qingfeng Gan, Wensheng Qi, Zhenlian Yu, Philip S. Databases With the advent of big data, periodic pattern mining has demonstrated significant value in real-world applications, including smart home systems, healthcare systems, and the medical field. However, advances in network technology have enabled malicious actors to extract sensitive information from publicly available datasets, posing significant threats to data providers and, in severe cases, hindering societal development. To mitigate such risks, privacy-preserving utility mining (PPUM) has been proposed. However, PPUM is unsuitable for addressing privacy concerns in periodic information mining. To address this issue, we innovatively extend the existing PPUM framework and propose two algorithms, Maximum sensitive Utility-MAximum maxPer item (MU-MAP) and Maximum sensitive Utility-MInimum maxPer item (MU-MIP). These algorithms aim to hide sensitive periodic high-utility itemsets while generating sanitized datasets. To enhance the efficiency of the algorithms, we designed two novel data structures: the Sensitive Itemset List (SISL) and the Sensitive Item List (SIL), which store essential information about sensitive itemsets and their constituent items. Moreover, several performance metrics were employed to evaluate the performance of our algorithms compared to the state-of-the-art PPUM algorithms. The experimental results show that our proposed algorithms achieve an Artificial Cost (AC) value of 0 on all datasets when hiding sensitive itemsets. In contrast, the traditional PPUM algorithm yields non-zero AC. This indicates that our algorithms can successfully hide sensitive periodic itemsets without introducing misleading patterns, whereas the PPUM algorithm generates additional itemsets that may interfere with user decision-making. Moreover, the results also reveal that our algorithms maintain Database Utility Similarity (DUS) of over 90\% after the sensitive itemsets are hidden. |
| title | Utility-based Privacy Preserving Data Mining |
| topic | Databases |
| url | https://arxiv.org/abs/2509.15755 |