User defined privacy grid system

USER-DEFINED PRIVACY GRID SYSTEM FOR CONTINUOUS LOCATION-
BASED SERVICES
Abstract—Location-based services (LBS) require users to continuously report
their location to a potentially untrusted server to obtain services based on their
location, which can expose them to privacy risks. Unfortunately, existing privacy-
preserving techniques for LBS have several limitations, such as requiring a fully-
trusted third party, offering limited privacy guarantees and incurring high
communication overhead. In this paper, we propose a user-defined privacy grid
system called dynamic grid system (DGS); the first holistic system that fulfills four
essential requirements for privacy-preserving snapshot and continuous LBS. (1)
The system only requires a semi-trusted third party, responsible for carrying out
simple matching operations correctly. This semi-trusted third party does not have
any information about a user’s location. (2) Secure snapshot and continuous
location privacy is guaranteed under our defined adversary models. (3) The
communication cost for the user does not depend on the user’s desired privacy
level, it only depends on the number of relevant points of interest in the vicinity of
the user. (4) Although we only focus on range and k-nearest-neighbor queries in
this work, our system can be easily extended to support other spatial queries
without changing the algorithms run by the semi-trusted third party and the

database server, provided the required search area of a spatial query can be
abstracted into spatial regions. Experimental results show that our DGS is more
efficient than the state-of-the-art privacy-preserving technique for continuous LBS.
EXISTING SYSTEM:
Spatial cloaking techniques have been widely used to preserve user location
privacy in LBS. Most of the existing spatial cloaking techniques rely on a fully-
trusted third party (TTP), usually termed location anonymizer, that is required
between the user and the service provider. When a user subscribes to LBS, the
location anonymizer will blur the user’s exact location into a cloaked area such that
the cloaked area includes at least k – 1 other users to satisfy k-anonymity. The TTP
model has four major drawbacks. (a) It is difficult to find a third party that can be
fully trusted. (b) All users need to continuously update their locations with the
location anonymizer, even when they are not subscribed to any LBS, so that the
location anonymizer has enough information to compute cloaked areas. (c)
Because the location anonymizer stores the exact location information of all users,
compromising the location anonymizer exposes their locations. (d) k-anonymity
typically reveals the approximate location of a user and the location privacy
depends on the user distribution. In a system with such regional location privacy it

is difficult for the user to specify personalized privacy requirements. The
feelingbased approach [29] alleviates this issue by finding a cloaked area based on
the number of its visitors that is at least as popular as the user’s specified public
region. Although some spatial clocking techniques can be applied to peer-to-peer
environments, these techniques still rely on the k-anonymity privacy requirement
and can only achieve regional location privacy. Furthermore, these techniques
require users to trust each other, as they have to reveal their locations to other peers
and rely on other peers’ locations to blur their locations., another distributed
method was proposed that does not require users to trust each other, but it still uses
multiple TTPs. Another family of algorithms uses incremental nearest neighbor
queries, where a query starts at an “anchor” location which is different from the
real location of a user and iteratively retrieves more points of interest until the
query is satisfied. While it does not require a trusted third party, the approximate
location of a user can still be learned; hence only regional location privacy is
achieved.
PROPOSED SYSTEM:
In this paper, we propose a user-defined privacy grid system called dynamic grid
system (DGS) to provide privacy-preserving snapshot and continuous LBS. The
main idea is to place a semitrusted third party, termed query server (QS), between
the user and the service provider (SP). QS only needs to be semi-trusted because it

will not collect/store or even have access to any user location information. Semi-
trusted in this context means that while QS will try to determine the location of a
user, it still correctly carries out the simple matching operations required in the
protocol, i.e., it does not modify or drop messages or create new messages. An
untrusted QS would arbitrarily modify and drop messages as well as inject fake
messages, which is why our system depends on a semi-trusted QS.
The main idea of our DGS. In DGS, a querying user first determines a query
area, where the user is comfortable to reveal the fact that she is somewhere within
this query area. The query area is divided into equal-sized grid cells based on the
dynamic grid structure specified by the user. Then, the user encrypts a query that
includes the information of the query area and the dynamic grid structure, and
encrypts the identity of each grid cell intersecting the required search area of the
spatial query to produce a set of encrypted identifiers. Next, the user sends a
request including (1) the encrypted query and (2) the encrypted identifiers to QS,
which is a semi-trusted party located between the user and SP. QS stores the
encrypted identifiers and forwards he encrypted query to SP specified by the user.
SP decrypts the query and selects the POIs within the query area from its database.
For each selected POI, SP encrypts its information, using the dynamic grid
structure specified by the user to find a grid cell covering the POI, and encrypts the
cell identity to produce the encrypted identifier for that POI. The encrypted POIs

with their corresponding encrypted identifiers are returned to QS. QS stores the set
of encrypted POIs and only returns to the user a subset of encrypted POIs whose
corresponding identifiers match any one of the encrypted identifiers initially sent
by the user. After the user receives the encrypted POIs, she decrypts them to get
their exact locations and computes a query answer.
Module 1
Dynamic grid system (dgs)
In this section, we will describe how our DGS supports privacypreserving
continuous range and k-NN queries. This section is organized as follows: Section
3.1 describes the details of our DGS for processing continuous range queries and
incrementally maintaining their answers, and Section 3.2 extends DGS to support
k-NN queries.
3.1 Range Queries Our DGS has two main phases for privacy-preserving
continuous range query processing. The first phase finds an initial answer for a
range query and the second phase incrementally maintains the query answer based
on the user’s location update. Range Query Processing a continuous range query
is defined as keeping track of the POIs within a user-specified distance Range of
the user’s current location (xu, yu) for a certain time period. In general, the
privacy-preserving range query processing protocolhas six main steps.

Step 1. Dynamic grid structure (by the user). The idea of this step is to construct
a dynamic grid structure specified by the user. A querying user first specifies a
query area, where the user is comfortable to reveal the fact that she is located
somewhere within that query area. The query area is assumed to be a rectangular
area, represented by the coordinates of its bottom-left vertex (xb, yb) and top-right
vertex (xt, yt).
Step 2. Request generation (by the user). In this step, the querying user generates
a request that includes (1) a query for a SP specified by the querying user and (2) a
set of encrypted identifiers, Se, for a QS. The user first selects a random key K and
derives three distinct keys: (HK,EK,MK) ← KDF(K) (1) where KDF(・) is a key
derivation function ( [24]). Then, the user sets query and Se.
Step 3. Request processing (by QS). When QS receives the request from the user,
it simply stores the set of encrypted identifiers Se and forwards the encrypted
query to SP specified by the user.
Step 4. Query processing (by SP). SP decrypts the request to retrieve the POI-
type, the random key K selected by the user in the request generation step (Step 2),
and the query area defined by m, (xb, yb), and (xt, yt). SP then selects a set of np
POIs that match the required POI-type within the user specified query area from its
database. For each selected POI j with a location (xj , yj)) (1 ≤ j ≤ np), SP

computes the identity of the grid cell in the user specified dynamic grid structure
covering j by (cj , rj) = _j xj−xb (xt−xb)/mk , j yj−yb (yt−yb)/mk_.
Step 5. Encrypted identifier matching (by QS). Upon receiving np triples, QS
determines the set of matching POIs by comparing the encrypted identifiers Cj (1 ≤
j ≤ np) of the received POI with the set of encrypted identifiers Se previously
received from the user. A match between a Cj and some Ci in the set Se indicates
that the POI j is in one of the grid cells required by the user. Thus, QS forwards
every matching POI hlj , σji to the user. If the query is a snapshot query, QS then
deletes the received POIs and their encrypted identifiers. However, if the query is a
continuous one, QS keeps the received POIs along with their encrypted identifiers
until the user unregisters the query.
Step 6. Answer computation (by the user). Suppose that there are μ matched
POIs received by the user. For each of these matched POIs, say hlj , σji, the user
decrypts lj using EK and gets access to the exact location (xj , yj) of the POI. From
(xj , yj) and lj, the user verifies σj by re-calculating the MAC value and compares it
against σj . If they match, the user finds the answer that includes the POI whose
location is within a distance of Range of the user’s current position (xu, yu).
Module 2
K-Nearest-NeighborQuery Processing

A continuous k-NN query is defined as keeping track of the knearest POIs to a
user’s current location (xu, yu) for a certain time period, as presented in Section 2.
In general, the privacypreserving k-NN query processing has six major steps to
find an initial query answer.
Step 1. Dynamic grid structure (by the user). This step is the same as the
dynamic grid structure step (Step 1) in the range query processing phase. It takes a
user-specified query area with a left-bottom vertex (xb, yb) and a right-top vertex
(xt, yt) and divides the query area into m × m equal-sized cells.
Step 2. Request generation (by the user). The required search area of the k-NN
query is initially unknown to the user. The user first finds at least k POIs to
compute the required search area as a circular area centered at the user’s location
with a radius of a distance from the user to the k-th nearest known POI. The user
therefore first attempts to get the nearby POIs from a specific SP. In this step, the
user requests the POIs in the cell containing the user and its neighboring cells from
SP. Given the user’s current location (xu, yu) and a query area specified by the
user in Step 1, she wants to get the POIs within a set of grid cells Sc that includes
the cell containing herself, i.e., (cu, ru) = _j xu−xb (xt−xb)/mk , j yu−yb
(yt−yb)/mk_, and its at most eight neighboring cells (cu −1, ru−1), (cu, ru−1),
(cu+1, ru−1), (cu − 1, ru), (cu + 1, ru), (cu − 1, ru + 1), (cu, ru + 1), and (cu + 1, ru
+ 1). For each cell i in Sc, the user generates an encrypted identifier Ci using

Equations 3 and 4, as in the request generation step (Step 2) in the range query
processing phase. The user also creates a query to be sent to SP. Finally, the user
sends a request, which includes the identity of SP, the query, and the set of
encrypted identifiers (in random order), to QS.
Step 3. Request processing (by QS). This step is identical to Step 3 for range
queries in the query processing phase.
Step 4. Query processing (by SP). This step is identical to Step 4 for range
queries in the query processing phase Thanks to this query abstraction feature, our
DGS can be easily extended to support other continuous spatial query types, e.g.,
reverse NN queries and density queries.
Step 5. Required search area (by the user and QS). This step is similar to the
encrypted identifier matching step (Step 5) for range queries in the query
processing phase, with the difference that this step may involve several rounds of
interaction between the user and QS. QS matches the encrypted identifiers of the
encrypted POIs returned by SP with the encrypted identifiers in Se sent by the user
in Step 2, and sends the matching encrypted POIs to the user.
Module 3
Privacy Against Service Provider (SP)

We require that SP cannot learn the user’s location any better than making a
random guess. Formally, we consider the following game played between a
challenger C and a (malicious) SP, denoted by A. The challenger prepares the
system parameters, and gives them to A. A specifies a POI-type, the grid structure,
a query area and two locations (x0, y0) and (x1, y1) in this area, and gives them to
C. C chooses at random b ∈ {0, 1}, uses (xb, yb), the specified grid structure and
POI-type to generate Msgb 2 with respect to the identity of A, i.e., the message that
the malicious SP expects to receive. C then gives Msgb 2 to A. A outputs a bit b′
and wins the game if b′ = b.
Module 4
Privacy Against Query Server (QS)
This requires that QS cannot tell from the user’s request or SP’s transcript about
where the user is, provided that it does not collude with the intended SP. Formally,
we consider the following game played between an adversary A (which is the
dishonest QS) and a challenger C which acts the roles of the user and service
providers. Given the system parameters, A begins to issue Private Key Query for
polynomially many times: it submits the identity of a SP to C, and receives the

corresponding private key. This models the case that QS colludes with a (non-
intended) SP. A then specifies the POI-type, he identity of the intended SP (the
private key of which has not been queried), the grid structure, a query area, and
two user locations (x0, y0) and (x1, y1) in the query area, and gives them to C. C
tosses a coin b ∈ {0, 1}, and uses (xb, yb) and the other information specified by A
to generate Msgb 1 as the user’s message to QS, and the corresponding SP
message Msgb 3. It sends both Msgb 1 and Msgb 3 to A. A continues to issue
queries as above except that it cannot ask for the private key of the intended SP.
Finally, A outputs a bit b′ as its guess of b, and wins the game if b′ = b.
CONCLUSION
In this paper, we proposed a dynamic grid system (DGS) for providing privacy-
preserving continuous LBS. Our DGS includes the query server (QS) and the
service provider (SP), and cryptographic functions to divide the whole query
processing task into two parts that are performed separately by QS and SP. DGS
does not require any fully-trusted third party (TTP); instead, we require only the
much weaker assumption of no collusion between QS and SP. This separation also
moves the data transfer load away from the user to the inexpensive and high-
bandwidth link between QS and SP. We also designed efficient protocols for our
DGS to support both continuous k-nearest-neighbor (NN) and range queries. To

evaluate the performance of DGS, we compare it to the state-of-the-art technique
requiring a TTP. DGS provides better privacy guarantees than the TTP scheme,
and the experimental results show that DGS is an order of magnitude more
efficient than the TTP scheme, in terms of communication cost. In terms of
computation cost, DGS also always outperforms the TTP scheme for NN queries;
it is comparable or slightly more expensive than the TTP scheme for range queries.
REFERENCES
[1] B. Bamba, L. Liu, P. Pesti, and T.Wang, “Supporting anonymous location
queries in mobile environments with PrivacyGrid,” in WWW, 2008.
[2] C.-Y. Chow and M. F. Mokbel, “Enabling private continuous queries for
revealed user locations,” in SSTD, 2007.
[3] B. Gedik and L. Liu, “Protecting location privacy with personalized
kanonymity: Architecture and algorithms,” IEEE TMC, vol. 7, no. 1, pp. 1–18,
2008.
[4] M. Gruteser and D. Grunwald, “Anonymous Usage of Location-Based Services
Through Spatial and Temporal Cloaking,” in ACM MobiSys, 2003.
[5] P. Kalnis, G. Ghinita, K. Mouratidis, and D. Papadias, “Preventing location-
based identity inference in anonymous spatial queries,” IEEE TKDE, vol. 19, no.
12, pp. 1719–1733, 2007.

[6] M. F. Mokbel, C.-Y. Chow, and W. G. Aref, “The new casper: Query
processing for location services without compromising privacy,” in VLDB, 2006.
[7] T. Xu and Y. Cai, “Location anonymity in continuous location-based services,”
in ACM GIS, 2007.
[8] “Exploring historical location data for anonymity preservation in location-
based services,” in IEEE INFOCOM, 2008.
[9] G. Ghinita, P. Kalnis, A. Khoshgozaran, C. Shahabi, and K.-L. Tan, “Private
queries in location based services: Anonymizers are not necessary,” in ACM
SIGMOD, 2008.
[10] M. Kohlweiss, S. Faust, L. Fritsch, B. Gedrojc, and B. Preneel, “Efficient
oblivious augmented maps: Location-based services with a payment broker,” in
PET, 2007.
[11] R. Vishwanathan and Y. Huang, “A two-level protocol to answer private
location-based queries,” in ISI, 2009.
[12] J.M. Kang,M. F.Mokbel, S. Shekhar, T. Xia, and D. Zhang, “Continuous
evaluation of monochromatic and bichromatic reverse nearest neighbors,” in IEEE
ICDE, 2007.

User defined privacy grid system

More Related Content

What's hot (19)

Similar to User defined privacy grid system (20)

More from nexgentech15 (20)

Recently uploaded (20)

User defined privacy grid system