First Phase CSR() Require:

// AddAllSegmentToCSR(junction n) : add all segments connected to the junctionn to CSR

// UserNowLocation : current user junction

// Mset junction : all destinations of a specific junction // way1 : select the junction which has the highest degree // way2 : randomly select junction

m← 0;

AddAllSegmentToCSR(UserNowLocation);

Add destinations of UserNowLocation to Mset UserNowLocation;

m← m + 1;

Randomly select way1 or way2;

if way1 then

while m != user determined M do

next path← current destination with highest degree;

AddAllSegmentToCSR(next path);

Add destination of next path to Mset next path;

m← m + 1;

check criteria();

Add the new path to Mset;

end while else if way2 then

while m != user determined M do Randomly select next path;

AddAllSegmentToCSR(next path);

Add the destination of the next path to Mset m;

end while check criteria();

Add the new path to Mset;

end if

from LBS data servers as main performance metrics to com-pare Junction Sharing and Random Selection methods in this paper.

A. Simulation Setup

Simulator is built with event-driven manner, and we eval-uate our protocols under two kinds of road network models:

virtual grid network and physical road network. In the virtual environment, a grid network is used with 416 road segments.

In contrast, a map of Jhongli City from Taiwan is applied in the physical environment. There is an AZ in the network and mobile user issues continuous query to AZ asking for LBS result. Table I summarizes the parameters in our simulation.

Furthermore, since the purpose of this research is mainly focus on path privacy protection, the computation overhead of LBS server is only virtually implemented in our simulation.

LBS server’s computation overhead is proportional to CSR’s size, including total length of RSi, denoted as length |RS|

(unit length), and number of BNs, denoted as|BN s|, within a CSR. This consequently affects the number of data that server needs to process. Therefore, the computation overhead of service provider is transferred to number of data, which

Algorithm 2 Random Selection CSR() // there will be m Mset(1 . . . m)

AddAllSegmentToCSR(UserNowLocation);

Clean Mset UserNowLocation;

Add destinations of UserNowLocation to Mset UserNowLocation;

next path[] ← randomly select a junction from the rest of Mset;

for i = 0 ; i < next path.size ; i++ do AddAllSegmentToCSR(next path[i]);

Clean Mset i;

Add the destination of next path to Mset i;

end for

check criteria();

Add all destination to right Mset;

TABLE I PARAMETERS

Parameter Virtual road Physical road

Road network area 3000 x 3000

-No. of road segments 416 160

Road map grid Jhongli City, Taiwan

M (for M-cut) 3∼5 3∼5

User path length 6∼14 6∼14

Moving speed 16 16

Continuous query time 900 900

Mobility model network-constrained network-constrained

computed as (length |RS|)/w1 + |BN s| × w2 among the simulation. Since BN s usually produce more data than road segments, w1 is set to350 (unit length) and w2 is set to five in our simulation.

B. Simulation Results

1) Virtual Road Network: In virtual network, 416 grid segments are used and the length of each road segment is the same. Except the outer bound of the grid network, all junction nodes in this environment have degree equal to four. The total computation overhead versus length of path is shown in Fig. 6. The Junction Sharing, denoted as Sharing, produces less data than the Random Selection, denoted as Random. This is mainly because Junction Sharing selects physically closer road junctions and segments to decrease the computation overhead.

With M-cut requirement increasing, the computation overhead grows as well for both methods.

The average number of data computation per CSR versus path length is illustrated in Fig. 7. As it shows, the path length does not affect the average overhead for each CSR. Each CSR’s overhead is only related to M-cut requirement. We ob-serve that Junction Sharing method increases the performance by 35% to 45% compared with Random Selection one. Thus, the Junction Sharing method can reduce the overhead of LBS servers significantly.

Path length (number of road segments)

Total number of data computation

Sharing M=3

Fig. 6. Total data computation in virtual road network versus path length.

6 7 8 9 10 11 12 13 14

Path length (number of road segments)

Average number of data computation per CSR

Sharing M=3

Fig. 7. Average data computation per CSR over virtual road network.

2) Physical Road Network: We utilize the real map of Jhongli City and extract the corresponding road model, degree of connectivity, and the length of each road to perform this simulation. Fig. 8 shows the total computation overhead versus length of path over this real road network. The Junction Sharing method is still better than the Random Selection method in physical network. Also, the total overhead increases while the path length increases. However, the increment of data computation is not strictly proportional to the value of M-cut.

Fig. 9 plots the average number of data computation per CSR, and Fig. 10 illustrates the number of CSR for each method andM with respect to different path length, denoted as Path. These two figures explain the unusual result of Fig. 8.

Average data computation per CSR is reasonable as it only related toM . However, Fig. 10 shows that AZ produces fewer number of CSRs when M increases. It is attributable to that largerM tends to enlarge the area of CSR, and user is more likely to stay in the same CSR without stepping out, hence reducing the number of total CSRs. With more number of CSRs, smallerM may involves more data processing, resulting in shown result.

We observe that Junction Sharing method only increases the performance by 9% to 24% compared with the Random Selection method over physical road network. It is because that physical road network is much more complicated than virtual one, thereby the effectiveness of Junction Sharing may decrease. Nevertheless, both Junction Sharing and Random Selection can support M-cut path privacy, and Junction Sharing provide better performance in average.

6 7 8 9 10 11 12 13 14

Path length (number of road segments)

Total number of data computation

Sharing M=3

Fig. 8. Total data computation in real road network versus path length.

6 7 8 9 10 11 12 13 14

Path length (number of road segments)

Average number of data computation per CSR

Sharing M=3

Fig. 9. Average data computation per CSR over real road network.

VI. CONCLUSIONS

This paper has addressed the issue of location privacy on continuous spatial query over road networks. We have proposed a concept of path privacy to protect user’s route from being inferred by an adversary. Furthermore, to achieve the objective of path privacy, we proposed a novel M-cut requirements. Besides, two methods of constructing CSRs are provided, namely Random Selection and Junction Sharing.

Experiment result shows that proposed construction methods are feasible to support user’s requirement and Junction Sharing can further reduce system overhead. Future research directions are to analyze the proposed M-cut, and to derive the complex-ity of blurred cloaked region for further evaluation.

ACKNOWLEDGEMENT

This work was supported in part by the National Science Council, Taiwan, under Grants NSC 101-2221-E-011-100.

REFERENCES

[1] K. Virrantaus, J. Markkula, A. Garmash, V. Terziyan, J. Veijalainen, A. Katanosov, and H. Tirri, “Developing gis-supported location-based

Fig. 10. Number of CSR in physical road network.

services,” in Proc. International Conference on Web Information Systems Engineering, vol. 2, Dec. 2001, pp. 66–75.

[2] B. Rao and L. Minakakis, “Evolution of mobile location-based services,”

Communications of the ACM, vol. 46, no. 12, pp. 61–65, dec 2003.

[3] J. Voelcker, “Stalked by satellite - an alarming rise in gps-enabled harassment,” IEEE Spectrum, vol. 43, no. 7, pp. 15 – 16, july 2006.

[4] M. F. Mokbel, C.-Y. Chow, and W. G. Aref, “The new casper: query processing for location services without compromising privacy,” in Proceedings of the 32nd international conference on Very large data bases, ser. VLDB ’06, 2006, pp. 763–774.

[5] P. Kalnis, G. Ghinita, K. Mouratidis, and D. Papadias, “Preventing location-based identity inference in anonymous spatial queries,” IEEE Transactions on Knowledge and Data Engineering, vol. 19, no. 12, pp.

1719 –1733, dec. 2007.

[6] J.-H. Um, M.-Y. Jang, K.-J. Jo, and J.-W. Chan, “A new cloaking method supporting both k-anonymity and l-diversity for privacy protection in location-based service,” International Symposium on Parallel and Distributed Processing with Applications, vol. 0, pp. 79–85, 2009.

[7] T. Wang and L. Liu, “Privacy-aware mobile services over road net-works,” Proceedings of the VLDB Endowment, vol. 2, no. 1, pp. 1042–

1053, aug 2009.

[8] C.-Y. Chow, M. F. Mokbel, J. Bao, and X. Liu, “Query-aware location anonymization for road networks,” Geoinformatica, vol. 15, no. 3, pp.

571–607, jul 2011.

[9] T. Xu and Y. Cai, “Location anonymity in continuous location-based ser-vices,” in Proceedings of the 15th annual ACM international symposium on Advances in geographic information systems, 2007, pp. 39:1–39:8.

[10] X. Pan, X. Meng, and J. Xu, “Distortion-based anonymity for continuous queries in location-based mobile services,” in Proceedings of the 17th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 2009, pp. 256–265.

[11] R. Dewri, I. Ray, I. Ray, and D. Whitley, “Query m-invariance: Pre-venting query disclosures in continuous location-based services,” in Proceedings of the 2010 Eleventh International Conference on Mobile Data Management, 2010, pp. 95–104.

[12] H. Shin, J. Vaidya, V. Atluri, and S. Choi, “Ensuring privacy and security for lbs through trajectory partitioning,” in Proc. International Conference on Mobile Data Management, 2010, pp. 224–226.

[13] B. Palanisamy and L. Liu, “Mobimix: Protecting location privacy with mix-zones over road networks,” in Proceedings of the 2011 IEEE 27th International Conference on Data Engineering, 2011, pp. 494–505.

[14] X. Liu, H. Zhao, M. Pan, H. Yue, X. Li, and Y. Fang, “Traffic-aware multiple mix zone placement for protecting location privacy,” in Proceedings IEEE INFOCOM 2012, march 2012, pp. 972 –980.

[15] T.-H. You, W.-C. Peng, and W.-C. Lee, “Protecting moving trajectories with dummies,” in Proceedings of the 2007 International Conference on Mobile Data Management, 2007, pp. 278–282.

[16] O. Abul, F. Bonchi, and M. Nanni, “Never walk alone: Uncertainty for anonymity in moving objects databases,” in Proc. IEEE 24th Interna-tional Conference on Data Engineering, april 2008, pp. 376 –385.

[17] M. E. Nergiz, M. Atzori, and Y. Saygin, “Towards trajectory anonymiza-tion: a generalization-based approach,” in Proc. International Workshop on Security and Privacy in GIS and LBS, 2008, pp. 52–61.

[18] B. Hoh, M. Gruteser, H. Xiong, and A. Alrabady, “Achieving guaranteed anonymity in gps traces via uncertainty-aware path cloaking,” IEEE Transactions on Mobile Computing, vol. 9, no. 8, pp. 1089 –1107, aug.

2010.

[19] C.-Y. Chow and M. F. Mokbel, “Trajectory privacy in location-based services and data publication,” ACM SIGKDD Explorations Newsletter, vol. 13, no. 1, pp. 19–29, aug 2011.

[20] K. Mouratidis and M. L. Yiu, “Anonymous query processing in road networks,” IEEE Transactions on Knowledge and Data Engineering, vol. 22, no. 1, pp. 2 –15, jan. 2010.

[21] J. Xue, X. Liu, X. Yang, and B. Wang, “Protecting location privacy using cloaking subgraphs on road network,” in Web Information Systems and Applications Conference, aug. 2010, pp. 65 –68.

[22] A.-A. Hossain, A. Hossain, H.-K. Yoo, and J.-W. Chang, “H-star:

Hilbert-order based star network expansion cloaking algorithm in road networks,” in 2011 IEEE 14th International Conference on Computa-tional Science and Engineering (CSE), aug. 2011, pp. 81 –88.

[23] S. Mascetti, C. Bettini, X. S. Wang, D. Freni, and S. Jajodia, “Provi-denthider: An algorithm to preserve historical k-anonymity in lbs,” in Proc. International Conference on Mobile Data Management: Systems, Services and Middleware, 2009, pp. 172–181.