Location Correlated Differential Privacy Protection Based on Mobile Feature Analysis

Abstract

Recently, with the popularity of smartphones and other GPS embedded devices, location-based service applications are being rapidly developed. In addition, individual privacy protection is also receiving increasing attention. Currently, most studies assume that individual location records are independent. However, the records are mostly interrelated in the real world. If the information is protected without considering the location-correlated information between users, an attacker can use a background knowledge attack to obtain the user's private information. Therefore, this paper proposes a method to protect multiuser location-correlated information under a strict privacy budget. First, a method for group movement analysis based on adaptive time segmentation is proposed in this paper. In addition, based on the time dimension, timecontinuous hotspot areas are constructed by adaptively segmenting and merging the stay areas, which are established for subsequent location-correlated privacy protection. Second, a data publishing mechanism is proposed to resist inferred attacks and to adaptively protect user-correlated location information. In addition, this paper also proposes the individual user correlation sensitivity concept and extends differential privacy by building an individual sensitivity matrix to correct noise. The experiments on real datasets show that under the same conditions, compared with the existing methods, the heat value of the hotspot areas formed by the method is increased by 10.11% under the same time slice length. In addition, the method reduces the similarity of 26.98% of group users.INDEX TERMS Correlated differential privacy, location protection, mobile feature analysis, time division. I. INTRODUCTIONRecently, with the popularity of GPS mobile devices, location-based service applications have been widely used in social and commercial fields. However, users generate a large amount of spatiotemporal data when using these services, which leads to leaking sensitive personal information. Therefore, it is vital to protect users' spatiotemporal data [1].The associate editor coordinating the review of this manuscript and approving it for publication was Kaigui Bian.Taking a wide view of current studies on the privacy protection of user location information, most of the methods assume that the dataset is independent and the internal data records are also independent. However, in real life, people have their social circles, and they often engage in certain social activities [2]. For example, some colleagues in a company may arrive at a designated place within a specified time in the morning and have breakfast together. Therefore, combining location data and time can reflect some of the user's social attributes [3], such as their home address, eating

An Efficient Differential Privacy-Based Method for Location Privacy Protection in Location-Based Services

“…[10] k-anonymity multi-center clustering; based on geographic semantics [11] double k-anonymity hides location and request information of users [12] STA-LPPM multi-objective particle swarm optimization [13] blockchain anonymous proof for P2P energy-trading location [14] DAK k-anonymity dynamically adjustable by adding and deleting points [16] tree structure attribute-aware privacy-preserving scheme [17] Markov model expands the size of the area [18] Markov decision achieved over an infinite time range [19] cluster, k-anonymity groups users by learning their trajectory data [20] graph trajectory data measurement method based on edges and nodes [21] differential privacy adds random noise and gathers the centroids of clusters [22] personalized security considers users' historical GPS trajectory and attributes [23] P-STM mines social connections of users' trajectories [24] DP, cluster constructs continuous time regions [25] PrivSem combines k-anonymity, l-diversity, and DP…”

“…Wang et al [23] proposed a privacy-protected social tie mining (P-STM) method, which can identify social connections from users' daily trajectories, and offered an indicative dense region to calibrate personal daily trajectories. In addition, a clustering analysis method for spatiotemporal sequence data was proposed in [24]. This method provides a basis for privacy protection by constructing continuous time regions and includes a data publishing mechanism that can prevent inferential attacks.…”

An Efficient Location Privacy Protection Method for Location-Based Services based on Differential Privacy

“…Thus, Wang et al [23] proposed a privacy-protected social tie mining (P-STM) method, which could find their social connections from users' daily trajectories, and offered an indicative dense region (IDR) to calibrate personal daily trajectories. At the same time, a clustering analysis method for spatiotemporal sequence data [24] was proposed, which provided the basis for privacy protection by constructing continuous time regions, and the data publishing mechanism was also offered to resist inferential attacks, but it mainly distributed the offline group location data, and could not update other relevant information. In [25] a new framework PrivSem was represented, which combined k-anonymity, l-semantic diversity and different privacy.…”

An Analysis of Differential Privacy Research in Location and Trajectory Data

“…Another privacy attack is possible by considering the location-correlated information between users and background knowledge attack to obtain the user's private trajectory as shown in [72]. A solution to neutralize such attacks is presented [73] RQ3.…”