Identification and Notification Process

Accountloginandselection

BullyingRisk(BullyingRankandkeymeasures)

BullyBlockeraimstobeatoolforparents’involvementintheirchildren'ssafety

Designing and Implementing a Model to Identify Cyberbullying in Facebook

Problem and Contributions• Cyberbullying is the use of online digital media to

communicate false, embarrassing, or hostileinformation about another person.

• It is the most common online risk for adolescents andwell over half of young people do not tell theirparents when it occurs.

• There have been many studies about the nature andprevalence of cyberbullying.

• Relatively less work in the area of automatedidentification of cyberbullying.

• The focus of our work is to develop an automatedmodel to identify and measure the degree ofcyberbullying in social networking sites.

• We present the design of a model for identifyingcyberbullying that builds on previous researchfindings in the areas of traditional bullying andcyberbullying in adolescents.

• We also identify challenges and opportunities tointegrate the latest results from psychology andsocial network data analysis to address a problem ofgreat social impact.

Design• BullyBlocker analyzes adolescents’ interactions with

their social network to identify:• Warning signs: e.g., number of insulting messages

and embarrassing pictures.• States of vulnerability: e.g., newcomers, age-gender

group, members of minority groups, people withdisabilities, etc.

• The integration of these factors is guided by previousresearch results in psychology.

• Bullying Rank: Estimates the probability of a minorexperiencing cyberbullying.

• The computed Bullying Rank is returned to the parentor guardian of the minor.

• The Bullying Rank Is divided into three normalizedlevels of risk intensity: Low[0-33], Medium[34-66],High[67-100]

Risk Factors

Architecture

Advisor: Yasin N. Silva Students: Chance Brown | Liz Garcia | Bryan Sawkins

Publications• Y. N. Silva, C. Rich, J. Chon, L. M.

Tsosie. BullyBlocker: An App to IdentifyCyberbullying in Facebook. The 2016IEEE/ACM International Conference onAdvances in Social Networks Analysisand Mining (ASONAM), San Francisco,CA, USA, 2016.

• Y. N. Silva, C. Rich, D. Hall.BullyBlocker: Towards the Identificationof Cyberbullying in Social NetworkingSites. The 2016 IEEE/ACMInternational Conference on Advancesin Social Networks Analysis and Mining(ASONAM), San Francisco, CA, USA,2016.

• L. M. Tsosie, Y. N. Silva. Facebully:Towards the Identification ofCyberbullying in Facebook. The GraceHopper Celebration of Women inComputing (GHC), Minnesota, USA,2013.

BullyingRank(BR)[0-100]

BR=90*WS+10*VF

WarningSigns(WS)[0-1]

WS=(DWIC^1.6)/(DWIC^1.6+.05)

VulnerabilityFactors(VF)[0-1]

VF=NSF*.33+NNF*.33+AGF*.33

AgeandGenderFactor(AGF)[0-1]

NewNeighborhoodFactor(NNF)[0-1]

NNF=1–((DNN/NNDT)^4) NewSchoolFactor(NSF)[0-1]

NSF=1–((DNS/NSDT)^4)

DaysSinceNewSchool(DNS)

NewSchoolDecayTime(NSDT)

DaysSinceNewNeighborhood

(DNN)

NewNeighborhoodDecayTime(NNDT)

DailyWeightedInsultCount(DWIC)

DWIC=TotalInsults

PhotoCommentInsults+1

FeedInsults+1

BRlevelsLowrisk:[0,33]Moderaterisk:[34,66]Severerisk:[67,100]

DataCollectionModule

PermanentStorage

CyberbullyingIdentificationModule

BullyingRankLearningModule

BullyingRankComputation

Notification(Bullyingrank,stats,resources)

ParentFeedback

FacebookGraphAPI

ChildFacebookAccount

Parent

ResourceGenerator

Warning Signs Vs. Daily Insult Count New Neighborhood/School Factor Vs. Time

FeedbackandAnti-bullyingResources

1 2 3 4 5 6 7 8 9

War

ning

Sig

ns(W

S)

Daily Weighted Insult Count