Supporting Privacy Protection in Personalized Web Search

ABSTRACT
Personalized web search (PWS) has demonstrated its effectiveness in improving the
quality of various search services on the Internet. owever! evidences show that users"
reluctance to disclose their private information during search has become a ma#or barrier for the
wide proliferation of PWS. We study privacy protection in PWS applications that model user
preferences as hierarchical user profiles. We propose a PWS framewor$ called %PS that can
adaptively generalize profiles by queries while respecting userspecified privacy requirements.
&ur runtime generalization aims at stri$ing a balance between two predictive metrics that
evaluate the utility of personalization and the privacy ris$ of e'posing the generalized profile.
We present two greedy algorithms! namely (reedy)P and (reedyI*! for runtime generalization.
We also provide an online prediction mechanism for deciding whether personalizing a query is
beneficial. +'tensive e'periments demonstrate the effectiveness of our framewor$. ,he
e'perimental results also reveal that (reedyI* significantly outperforms (reedy)P in terms of
efficiency.
GLOBALSOFT TECHNOLOGIES
IEEE PROJECTS & SOFTWARE DEVELOPMENTS
IEEE FINAL YEAR PROJECTS|IEEE ENGINEERING PROJECTS|IEEE STUDENTS PROJECTS|IEEE
BULK PROJECTS|BE/BTECH/ME/MTECH/MS/MCA PROJECTS|CSE/IT/ECE/EEE PROJECTS
CELL: +91 9!9" #9$"% +91 99&&' #"(% +91 9!9" "(9$% +91 9($1! !$!$1
V)*)+: ,,,-.)/012304546738+*-649 M0)1 +6:)333.)/01*3:546738+*;9:0)1-86:
GLOBALSOFT TECHNOLOGIES
IEEE PROJECTS & SOFTWARE DEVELOPMENTS
IEEE FINAL YEAR PROJECTS|IEEE ENGINEERING PROJECTS|IEEE STUDENTS PROJECTS|IEEE
BULK PROJECTS|BE/BTECH/ME/MTECH/MS/MCA PROJECTS|CSE/IT/ECE/EEE PROJECTS
CELL: +91 9!9" #9$"% +91 99&&' #"(% +91 9!9" "(9$% +91 9($1! !$!$1
V)*)+: ,,,-.)/012304546738+*-649 M0)1 +6:)333.)/01*3:546738+*;9:0)1-86:
Existing System
,he e'isting profile-based Personalized Web Search do not support runtime profiling. .
user profile is typically generalized for only once offline! and used to personalize all queries
from a same user indiscriminatingly. Such /one profile fits all0 strategy certainly has drawbac$s
given the variety of queries. &ne evidence reported in is that profile-based personalization may
not even help to improve the search quality for some ad hoc queries! though e'posing user
profile to a server has put the user"s privacy at ris$.
,he e'isting methods do not ta$e into account the customization of privacy requirements.
,his probably ma$es some user privacy to be overprotected while others insufficiently protected.
1or e'ample! in! all the sensitive topics are detected using an absolute metric called surprisal
based on the information theory! assuming that the interests with less user document support are
more sensitive. owever! this assumption can be doubted with a simple countere'ample2 If a
user has a large number of documents about /se'!0 the surprisal of this topic may lead to a
conclusion that /se'0 is very general and not sensitive! despite the truth which is opposite.
%nfortunately! few prior wor$ can effectively address individual privacy needs during the
generalization.
3any personalization techniques require iterative user interactions when creating personalized
search results. ,hey usually refine the search results with some metrics which require multiple
user interactions! such as ran$ scoring! average ran$! and so on. ,his paradigm is! however!
infeasible for runtime profiling! as it will not only pose too much ris$ of privacy breach! but also
demand prohibitive processing time for profiling. ,hus! we need predictive metrics to measure
the search quality and breach ris$ after personalization! without incurring iterative user
interaction.
isadvantage!
.ll the sensitive topics are detected using an absolute metric called surprisal based on the
information theory.
Proposed System!
We propose a privacy-preserving personalized web search framewor$ %PS! which can
generalize profiles for each query according to user-specified privacy requirements. 4elying on
the definition of two conflicting metrics! namely personalization utility and privacy ris$! for
hierarchical user profile! we formulate the problem of privacy-preserving personalized search as
4is$ Profile (eneralization! with its5P-hardness proved.
We develop two simple but effective generalization algorithms! (reedy)P and (reedyI*!
to support runtime profiling. While the former tries to ma'imize the discriminating power ()P)!
the latter attempts to minimize the information loss (I*). 6y e'ploiting a number of heuristics!
(reedyI* outperforms (reedy)P significantly.
We provide an ine'pensive mechanism for the client to decide whether to personalize a
query in %PS. ,his decision can be made before each runtime profiling to enhance the stability
of the search results while avoid the unnecessary e'posure of the profile.
Advantages!
7. It enhances the stability of the search quality.
8. It avoids the unnecessary e'posure of the user profile.
Architecture!
Enhanced
"#$%ES0
&' Profile-6ased Personalization.
(' Privacy Protection in PWS System'
)' (eneralizing %ser Profile.
*' &nline )ecision.
"odules escription
&' Profile-6ased Personalization
,his paper introduces an approach to personalize digital multimedia content based on
user profile information. 1or this! two main mechanisms were developed2 a profile
generator that automatically creates user profiles representing the user preferences! and a
content-based recommendation algorithm that estimates the user9s interest in un$nown
content by matching her profile to metadata descriptions of the content. 6oth features are
integrated into a personalization system.
(' Privacy Protection in PWS System
We propose a PWS framewor$ called %PS that can generalize profiles in for each
query according to user-specified privacy requirements. ,wo predictive metrics are
proposed to evaluate the privacy breach ris$ and the query utility for hierarchical user
profile. We develop two simple but effective generalization algorithms for user profiles
allowing for query-level customization using our proposed metrics. We also provide an
online prediction mechanism based on query utility for deciding whether to personalize a
query in %PS. +'tensive e'periments demonstrate the efficiency and effectiveness of our
framewor$.
)' (eneralizing %ser Profile
,he generalization process has to meet specific prerequisites to handle the user
profile. ,his is achieved by preprocessing the user profile. .t first! the process initializes
the user profile by ta$ing the indicated parent user profile into account. ,he process adds
the inherited properties to the properties of the local user profile. ,hereafter the process
loads the data for the foreground and the bac$ground of the map according to the
described selection in the user profile.
.dditionally! using references enables caching and is helpful when considering an
implementation in a production environment. ,he reference to the user profile can be
used as an identifier for already processed user profiles. It allows performing the
customization process once! but reusing the result multiple times. owever! it has to be
made sure! that an update of the user profile is also propagated to the generalization
process. ,his requires specific update strategies! which chec$ after a specific timeout or a
specific event! if the user profile has not changed yet. .dditionally! as the generalization
process involves remote data services! which might be updated frequently! the cached
generalization results might become outdated. ,hus selecting a specific caching strategy
requires careful analysis.
*' &nline )ecision
,he profile-based personalization contributes little or even reduces the search
quality! while e'posing the profile to a server would for sure ris$ the user"s privacy. ,o
address this problem! we develop an online mechanism to decide whether to personalize a
query. ,he basic idea is straightforward. if a distinct query is identified during
generalization! the entire runtime profiling will be aborted and the query will be sent to
the server without a user profile.
System Con+iguration!,
:W System ;onfiguration2-
Processor , Pentium -...
Speed - 7.7 (hz
4.3 - 8<= 36 (min)
ard )is$ - 8> (6
1loppy )rive - 7.?? 36
@ey 6oard - Standard Windows @eyboard
3ouse - ,wo or ,hree 6utton 3ouse
3onitor - SA(.
S:W System ;onfiguration2-
&perating System 2WindowsB<:BC:8>>>:DP
.pplication Server 2 ,omcat<.>:=.D
1ront +nd 2 ,3*! Eava! Esp
Scripts 2 EavaScript.
Server side Script 2 Eava Server Pages.
)atabase 2 3ysql
)atabase ;onnectivity 2 E)6;.
C#/C%$S.#/
,his paper presented a client-side privacy protection framewor$ called %PS for
personalized web search. %PS could potentially be adopted by any PWS that captures user
profiles in a hierarchical ta'onomy. ,he framewor$ allowed users to specify customized privacy
requirements via the hierarchical profiles. In addition! %PS also performed online generalization
on user profiles to protect the personal privacy without compromising the search quality. We
proposed two greedy algorithms! namely (reedy)P and (reedyI*! for the online generalization.
&ur e'perimental results revealed that %PS could achieve quality search results while preserving
user"s customized privacy requirements. ,he results also confirmed the effectiveness and
efficiency of our solution.
RE0ERE/CES
F7G H. )ou! 4. Song! and E.-4. Wen! /. *arge-Scale +valuation and .nalysis of Personalized
Search Strategies!0 Proc. Int"l ;onf. World Wide Web (WWW)! pp. <C7-<B>! 8>>I.
F8G E. ,eevan! S.,. )umais! and +. orvitz! /Personalizing Search via .utomated .nalysis of
Interests and .ctivities!0 Proc. 8Cth .nn. Int"l .;3 SI(I4 ;onf. 4esearch and )evelopment in
Information 4etrieval (SI(I4)! pp. ??B-?<=! 8>><.
FJG 3. Spertta and S. (ach! /Personalizing Search 6ased on %ser Search istories!0 Proc.
I+++:WI;:.;3 Int"l ;onf. Web Intelligence (WI)! 8>><.
F?G 6. ,an! D. Shen! and ;. Hhai! /3ining *ong-,erm Search istory to Improve Search
.ccuracy!0 Proc. .;3 SI(@)) Int"l ;onf. @nowledge )iscovery and )ata 3ining (@)))!
8>>=.
F<G @. Sugiyama! @. atano! and 3. Koshi$awa! /.daptive Web Search 6ased on %ser Profile
;onstructed without any +ffort from %sers!0 Proc. 7Jth Int"l ;onf. World Wide Web (WWW)!
8>>?.
Scope!
,o protect user privacy in profile-based PWS! researchers have to consider two
contradicting effects during the search process. &n the one hand! they attempt to improve the
search quality with the personalization utility of the user profile. &n the other hand! they need to
hide the privacy contents e'isting in the user profile to place the privacy ris$ under control. .
few previous studies ! suggest that people are willing to compromise privacy if the
personalization by supplying user profile to the search engine yields better search quality. In an
ideal case! significant gain can be obtained by personalization at the e'pense of only a small (and
less-sensitive) portion of the user profile! namely a generalized profile. ,hus! user privacy can be
protected without compromising the personalized search quality. In general! there is a tradeoff
between the search quality and the level of privacy protection achieved from generalization.