NetSpam:ANetwork-Based Spam Detection Framework for Reviewsin Online SocialMedia

Abstract

Nowadays,abigpartofpeoplerelyonavailablecontentinsocialmediaintheirdecisions(e.g.,reviewsandfeedback
onatopic orproduct). The possibility thatanybodycan leavea review provides a golden opportunity for
spammersto write spam reviews about products and services for
differentinterests.Identifyingthesespammersandthespamcontentisahottopicofresearch,andalthoughaconsid
erablenumberofstudies havebeen done recentlytowardthis end, but sofarthe methodologies put forth still
barely detect spam reviews, andnone of them show the importance of each extracted feature type.In this
paper, we propose a novel framework, named NetSpam,whichutilizesspamfeaturesformodeling
reviewdata setsas heterogeneous information networks to map spam detectionprocedure into a
classification problem in such networks. Usingthe importance of spam features helps us to obtain better
resultsin terms of different metrics experimented on real-world reviewdata sets from Yelp and
AmazonWeb sites. The results
showthatNetSpamoutperformstheexistingmethodsandamongfourcategoriesoffeatures,including reviewbehavioral,
user-behavioral,review-linguistic, anduser-linguistic,
thefirsttypeoffeaturesperformsbetterthantheothercategories.