1 / 28

PRIVÉ: Anonymous Location-Based Queries in Distributed Mobile Systems

PRIVÉ is a decentralized architecture that preserves user anonymity in spatial queries to Location-Based Services (LBS). It offers fault tolerance, load balancing, and superior construction mechanism for anonymizing queries. This article presents the system architecture, drawbacks of existing approaches, and the HILBASR algorithm used in PRIVÉ. Experimental evaluation confirms the efficiency and scalability of PRIVÉ.

bhiggins
Download Presentation

PRIVÉ: Anonymous Location-Based Queries in Distributed Mobile Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. PRIVÉ: Anonymous Location-Based Queries in Distributed Mobile Systems Gabriel Ghinita, Panos Kalnis, Spiros Skiadopoulos WWW 07

  2. Outline • Introduction • System Architecture • Drawbacks of Existing Approaches • The HILBASR Algorithm • Fault Tolerance and Load Balancing • Experimental Evaluation • Conclusions

  3. Introduction (1/4) • PRIVÉ, a decentralized architecture for preserving the anonymity of users issuing spatial queries to LBS. • Good fault tolerance and load balancing properties. • PRIVÉ avoids the bottleneck caused by centralized techniques both in terms of anonymization and location updates. • The system state is distributed in numerous users, rendering PRIVÉ resilient to attacks

  4. Introduction (2/4) • Instead of reporting the exact coordinates to the LBS, an Anonymizing Spatial Region (K-ASR) is constructed, which encloses the locations of K-1 additional users. • Each user sends his query to the anonymizer, which constructs the appropriate K-ASR and contacts the LBS. • The LBS computes the answer based on the K-ASR, instead of the exact user location. • Finally, the anonymizer filters the result from the LBS and returns the exact answer to the user.

  5. Introduction (3/4) • We develop a superior K-ASR construction mechanism based on the Hilbert space-filing curve, that guarantees query anonymity even if the attacker knows the location of all users. • We introduce a distributed protocol used by mobile entities to self-organize into a fault-tolerant overlay network. • The structure of the network resembles a distributed B+-tree.

  6. Introduction (4/4) • We also conduct an extensive experimental evaluation. • The results confirm that PRIVÉ achieves efficient anonymization and load balancing with low maintenance overhead. • It is scalable to large numbers of mobile users.

  7. System Architecture (1/5)

  8. System Architecture (2/5) • A trusted central Certification Server (CS), where users are registered. • Prior to entering the system, a user u must authenticate against the CS and obtain a certificate. • Users having a certificate are trusted by all other users. • Typically, a certificate is valid for a few hours; it can be renewed by recontacting the CS.

  9. System Architecture (3/5) • The CS returns to u the IP addresses of some users who are currently in the system. • Note that the CS does not know the locations of the users and does not participate in the anonymization process. • Therefore the workload of the CS is low; moreover it does not store any sensitive information.

  10. System Architecture (4/5) • Each user corresponds to a peer. • Peers are grouped into clusters, according to their location. • Within each cluster, peers elect a cluster head, and the set of heads is grouped recursively to form a tree. • To achieve load balancing, cluster heads are rotated in a round-robin manner.

  11. System Architecture (5/5)

  12. Drawbacks of Existing Approaches (1/2) • QUADASR

  13. Drawbacks of Existing Approaches (2/2) • CLOAKP2P • The query source initiates a K-ASR request by contacting all peers within a given physical radius r, which is a fixed system parameter. • If the set of peers S0 found in the initial iteration is larger than K, the nearest K of them are chosen to form the K-ASR. • Otherwise, the process continues, and all peers in S0 issue a request to all peers within radius r. • The process stops when K or more users have been found. • Unfortunately, this simple heuristic fails to achieve anonymity in many cases.

  14. The HILBASR Algorithm (1/5) • The Hilbert space-filling curve is a continuous fractal which maps each region of a multi-dimensional space to an integer. In our case, the 2D coordinates of user locations are mapped to a 1D value. • With high probability, if two points are close in the 2D space, they will also be close in the Hilbert transformation.

  15. The HILBASR Algorithm (2/5)

  16. The HILBASR Algorithm (3/5) • HILBASR computes and sorts the Hilbert values of all users. Then, the algorithm conceptually groups the sorted Hilbert values into K-buckets that contain K users.

  17. The HILBASR Algorithm (4/5) • HILBASR performs a search for H(u) in the index and computes ranku, which corresponds to the position of H(u). • From ranku, we calculate the start and end positions defining the K-bucket which includes H(u), as: • start = ranku – (ranku mod Ku) • end = start + Ku -1 • To compute ranku efficiently, we use an annotated B+-tree.

  18. The HILBASR Algorithm (5/5)

  19. Index Operations (1/3) • The users of each cluster C elect a leader called head(C). The head (marked with an asterisk) handles all index operations.

  20. Index Operations (2/3) • Join: Newly joining users authenticate at the certification server and receive the address of a user already inside the system. • Departure, Relocation

  21. Index Operations (3/3) • K-request

  22. Fault Tolerance and Load Balancing (1/4) • Fault Tolerance • Each cluster leader sends periodically a membership_update message to all cluster members. The message contains the membership list of the current cluster C and that of parent(C). • Cluster members respond to these messages; if a cluster member does not respond to 2 consecutive messages, it its considered disconnected and removed from the cluster. The change is broadcast by the cluster head to the remaining cluster members.

  23. Fault Tolerance and Load Balancing (2/4) • If a non-head cluster member u does not receive a membership_update from its head for a 2δt period, it initiates a leader election process. • Alternatively, when u attempts to initiate a operation, such as query or relocation, but cannot contact the cluster head for two consecutive attempts, it triggers the leader election protocol without waiting for the timer to expire.

  24. Fault Tolerance and Load Balancing (3/4) • Load Balancing • We propose a cluster head rotation mechanism. • Rotation is triggered when a node reaches a certain load threshold, denoted by load unit. • Starting with its highest layer • The member with the least load is appointed as new head.

  25. Fault Tolerance and Load Balancing (4/4) • The granularity of load unit choice is important in practice, in order to achieve a good tradeoff between load balancing and communication cost.

  26. Experimental Evaluation (1/2)

  27. Experimental Evaluation (2/2)

  28. Conclusions (1/1) • PRIVÉ, a distributed system for query anonymization in LBS. • PRIVÉ supports our HILBASR anonymization technique, which guarantees anonymity under any user distribution. • We show experimentally that our system is efficient, scalable, fault tolerant and achieves load balancing.

More Related