understanding cross-site linking in online social networks yang chen 1, chenfan zhuang 2, qiang cao...

21
Understanding Cross-site Linking in Online Social Networks Yang Chen 1 , Chenfan Zhuang 2 , Qiang Cao 1 , Pan Hui 3 1 Duke University 2 Tsinghua University 3 Hong Kong University of Science and Technology

Upload: winfred-jones

Post on 12-Jan-2016

223 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Understanding Cross-site Linking inOnline Social Networks

Yang Chen1, Chenfan Zhuang2, Qiang Cao1, Pan Hui3

1Duke University2Tsinghua University

3Hong Kong University of Science and Technology

Page 2: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Motivation

• A number of OSN sites with different functionality

• It is quite common for an individual user to have multiple accounts on different OSN sites.– How to efficiently manage accounts on different OSNs?

2

Interact with friends Share breaking news

Job search Location-centric social interactions

Page 3: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

The Cross-site Linking Function

The cross-site linking function allows a user to link her account on one OSN site to her accounts on other OSN sites

3

Page 4: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Cross-site Linking: Advantages

• Make cross-site content posting easy– Help Foursquare users automatically post tips to Facebook and Twitter

• Avoid repeated efforts in social connection establishment– Import the contact list from other OSNs

• Provide more information of a user– Visit the linked profiles on other OSNs

4

Page 5: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Why Foursquare?

• A representative LBSN service, one of the most popular OSN sites– Foursquare app: customized discovery and

recommendation engine– Swarm app: real-time location sharing with friends

• Supports the cross-site linking function– A user can link his profile to Facebook/Twitter

• Every Foursquare user has a public profile page (http://foursquare.com/user/ID/)

5

Page 6: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Data Collection

• Goal: analyze the entire Foursquare user base– To avoid the disadvantages of biased sampling

• Challenge: IP-based rate limiting– Crowd crawling:100 crawlers around the United States,

each crawls one chunk of IDs• Data– Collected between Jul. 22th, 2014 and Jul. 29th, 2014– Public profiles of 51.15 million Foursquare users

• Almost all (if not all) Foursquare users

6

ID Gender Tips Checkins Friends Facebook Twitter …

123 male 5 28 10 15576473 eric_c …

Page 7: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Linking Options

Twitter Facebook Linking Option PercentageYes No TW only 3.82%No Yes FB only 44.19%Yes Yes FB+TW 11.96%No No Neither 40.03%

• About 60% Foursquare users have enabled the cross-site linking function

• 56.15% users have added their Facebook accounts• 15.78% users have added their Twitter accounts

The cross-site linking function is widely used among Foursquare users7

Percentage Distribution of Linking Options

Page 8: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Group-based Analysis: Gender

Very little difference between male users and female users in terms of the distribution of linking options

Gender options: 51.52% Male 42.92% Female 5.56% Rather not say

8

Page 9: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Group-based Analysis: Country

Top four countries: USA (27.61%) Turkey (9.49%) Indonesia (8.17%) Brazil (7.04%)

9

Country BRA IDN TUR USA

Percentage 82.25% 50.51% 80.33% 50.33%

Percentage of users that have enabled cross-site linking

Page 10: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Group-based Analysis: Activity

• Two factors: social connections and location-centric activities (leaving tips, check-ins)

# of Friends # of Checkins and Tips Group Percentage=0 =0 Zombies 28.23%=0 >0 Loners 9.50%>0 =0 Watchers 14.02%>0 >0 Ordinary users 48.25%

10

Page 11: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Group-based Analysis: Activity (cont.)

Most zombies/loners have not enabled cross-site linking, as they are socially isolated. Watchers v.s. Orindary users

Watchers are less motivated to link to Twitter, as they don’t publish 73% watchers and 71% ordinary users have linked to Facebook (users from both

groups are connected with other Foursquare users)11

Page 12: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Behavioral Difference among Users with Different Linking Options

• Users who have enabled the cross-site linking function are more “active”– Cross-site linking will deliver published contents to more prospective audience

• “TW only” > “FB only”– Publicly viewable tweets can be quickly spread through the Twitter network– FB user status is only visible to friends by default (fewer possible audiences)12

Page 13: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

User Privacy Concerns & Cross-site Linking

Profile Picture

Gender Residential Location

Last Name Biography

Enabled (%) 66.99% 94.44% 91.68% 94.61% 3.29%

Disabled (%) 33.01% 5.56% 8.32% 5.39% 96.71%

In Foursquare, users can customize their profiles according to privacy concerns (a user can choose whether to enable an optional field)

13

Alice’s Foursquare Profile

Alice’s Facebook Profile

Alice’s Twitter Profile

More information about Alice!!

Intuitively, cross-site linking might cause concerns for users who care a lot about their privacy.

Page 14: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

User Privacy Concerns & Cross-site Linking (cont.)

Whether or not uploading personalized profile photo is an indicator for the adoption of cross-site linking

Enabling any of the five optional field indicates a higher probability of using the cross-site linking function

14

Page 15: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Cross-site Information Consistency

First Name Last Name Gender

Percentage of users who have entered identical information in a selected filed

89.84% 87.02% 99.30%

Cross-site Information Consistency (“Foursquare-Facebook”)

Users have a high probability to manifest cross-site information consistency

15

A user might choose to expose the same or different personal information on different sites

Page 16: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Cross-site Information AggregationUser Info

Foursquare Twitter

ID Gender Tips … ID Tweets Lists …

USER A 1 m 10 … 1982 100 17 …

USER B 2 f 20 … 34 5 0 …

USER C 5 f 0 … 19903 20 1 …

USER D 9 m 7 … 563122 7 4 …

… … … … … … … … …

• Aggregate the information of the same user from different OSN sites (learn more about a user)

• Applications: friend suggestion, point-of-interest recommendation, personalized advertising, …

• Example: gender-based analysis of Twitter16

Page 17: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Gender-based Analysis of Twitter

17

Female users publish more tweets Male users are involved in more lists

Page 18: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Gender-based Analysis of Twitter (cont.)

URL Description Location

Male 32.07% 62.82% 56.87%

Female 25.02% 64.73% 57.35%

The Use of Optional Fields (%)

Male users have a higher probability to add a URL Both male and female users have a nearly 63%

probability of adding a description, and a nearly 57% probability to add the location information.

18

Page 19: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Related Work• Cross-OSN linking papers

– Ottoni et al. [ICWSM 2014]: Pinterest-Twitter linking– Chen et al. [WOSN 2012]: Google+– Both of them used biased sampling methods, while our study is based

on the entire Foursquare user population• Wang et al. [IEEE Internet Computing, 2014] compared a series of user

activities across Foursquare, Facebook, and Twitter– We focus on the cross-site linking function

• Goga et al. [WWW 2013], Liu et al. [SIGMOD 2014] investigated how to identify accounts on different OSNs that all are owned by the same user– Useful for the OSN sites which do not support the cross-site linking

function

19

Page 20: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Conclusion

• About 60% of Foursquare users have enabled the cross-site linking function, and these users are more active than other users

• Adding contents to an optional field indicates a higher probability of activating the cross-site linking function

• If a Foursquare user has linked his account to Facebook, he will have a high chance to provide consistent information to both Foursquare and Facebook

• The use of cross-site information aggregation helps us investigate the gender difference in using Twitter

20

Page 21: Understanding Cross-site Linking in Online Social Networks Yang Chen 1, Chenfan Zhuang 2, Qiang Cao 1, Pan Hui 3 1 Duke University 2 Tsinghua University

Future Work

• Investigate cross-site links among more mainstream OSN sites– Discover general patterns to characterize cross-OSN links

• A volunteer-based study – Access the non-publicly viewable data (e.g.: check-in

history) a deeper investigation into cross-site information aggregation

• Build practical services/applications based on cross-site links– Malicious account detection

21