Crowdsourced App Review Manipulation
Shanshan Li James Caverlee Wei Niu Parisa Kaghazgaran
Department of Computer Science and Engineering, Texas A&M University
College Station, Texas, USA
{ssli,caverlee,niu,kaghazgaran}@tamu.edu
ABSTRACT
With the rapid adoption of smartphones worldwide and the reliance
on app marketplaces to discover new apps, these marketplaces are
critical for connecting users with apps. And yet, the user reviews
and ratings on these marketplaces may be strategically targeted by
app developers. We investigate the use of crowdsourcing platforms
to manipulate app reviews. We nd that (i) apps targeted by crowd-
sourcing platforms are rated signicantly higher on average than
other apps; (ii) the reviews themselves arrive in bursts; (iii) app
reviewers tend to repeat themselves by relying on some standard
repeated text; and (iv) apps by the same developer tend to share
a more similar language model: if one app has been targeted, it is
likely that many of the other apps from the same developer have
also been targeted.
KEYWORDS
app reviews; manipulation; crowdsourcing; user behavior
1 INTRODUCTION
Mobile app marketplaces like Google Play and Apple’s App Store
serve as the nexus for many of our online experiences. With the
rapid adoption of smartphones worldwide and the reliance on app
marketplaces to discover new apps, these marketplaces are critical
for connecting users with apps [
5
]. A key factor driving user en-
gagement with apps is user reviews and ratings. Indeed, previous
research has explored text mining to identify ne-grained app fea-
tures mentioned in reviews [
4
] and how user reviews can improve
app retrieval through methods that exploit reviews [
7
,
9
]. From
a software engineering perspective, researchers have developed
methods to extract informative reviews that can help developers
respond to user feedback [
2
] and observed that apps that respond
to user feedback (e.g., via bugs or desired features suggested in user
reviews) in future releases do indeed increase their ratings over
time [2, 8].
And yet, these user reviews and ratings may be strategically
targeted by app developers to articially promote their own apps
(or potentially, to demote the apps of competitors). Indeed, seminal
work by Chandy et al. [
1
] identied spam app reviews which aim
to deceive users to download harmful apps or impede them from
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than ACM
must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,
to post on servers or to redistribute to lists, requires prior specic permission and/or a
SIGIR ’17, August 07-11, 2017, Shinjuku, Tokyo, Japan
© 2017 ACM. 978-1-4503-5022-8/17/08. . . $15.00
DOI: http://dx.doi.org/10.1145/3077136.3080741
downloading benign apps. Follow on research has explored collu-
sion in app rating systems [
12
,
14
], explored the use of incentivized
review marketplaces to attack the trustworthiness of app reviews
[13], and found evidence of app popularity manipulation [15].
Amazing Party Going App!! ? ? ? ? ?
After downloading and using this app I started thinking
about all the times in the past I could have used this
amazing app. I love the party nder and text blocker
and call blocker features more than anything. Drunk
texting and dialing issues are a problem of the past.
Denitely a great phone tracker as well. Great app if
you enjoy the nightlife, drinking, pre game parties, or
the frat life. The app also helps you keep track of your
friends with the friends nder feature, and who hasn’t
lost a drunk friend a time or two. Amazing ideas and
user friendliness. I will not go out without having this
app ready again!
Figure 1: Example of a suspicious review.
In this paper, we investigate a complementary attack vector on
the trustworthiness of app reviews: crowdsourcing platforms that
allow a single manipulator to martial a crowd of human review
writers to target app reviews. Such crowd-based manipulation has
been identied as a serious threat to the viability of many systems
that rely on user-generated content [
3
,
6
,
10
,
11
], but there has
been little if any study of crowdsourced targeting of app reviews.
We present our initial investigation into the use of crowdsourcing
platforms to launch targeted review manipulation on these app
marketplaces. Our overarching goal is to study if these platforms
are susceptible to crowdsourced attacks – Do crowdsourced reviews
bypass review lters to actually be posted? Are reviews positive or
negative? Do they actually impact the aggregate ratings of apps?
Are there correlations among a developer’s apps in terms of be-
ing targeted for manipulation? Does the platform’s “related apps”
feature expose users to more targeted apps?
Toward answering these questions, we sample 100+ targeted
apps from a popular crowdsourcing platform (and a control group
of randomly selected apps from the App Store) and make the follow-
ing observations: (i) we nd that apps targeted by crowdsourcing
platforms are rated signicantly higher on average than other apps,
indicating that app manipulation is focused on app promotion,
rather than in punishing the apps of competitors; (ii) the reviews
themselves arrive in bursts, and have an immediate positive impact
on the average ratings of the apps; (iii) the patterns of linguistic
evolution suggest that app reviewers tend to repeat themselves by
relying on some standard repeated text; and (iv) apps by the same
developer tend to share a more similar language model: if one app