Lecture Notes in Social Networks Thirimachos Bourlai Panagiotis Karampelas Reda Alhajj   Editors Disease Control Through Social Network Surveillance Lecture Notes in Social Networks SeriesEditors RedaAlhajj,UniversityofCalgary,Calgary,AB,Canada UweGlässer,SimonFraserUniversity,Burnaby,BC,Canada AdvisoryEditors CharuC.Aggarwal,YorktownHeights,NY,USA PatriciaL.Brantingham,SimonFraserUniversity,Burnaby,BC,Canada ThiloGross,UniversityofBristol,Bristol,UK JiaweiHan,UniversityofIllinoisatUrbana-Champaign,Urbana,IL,USA RaúlManásevich,UniversityofChile,Santiago,Chile AnthonyJ.Masys,UniversityofLeicester,Ottawa,ON,Canada LectureNotesinSocialNetworks(LNSN)comprisesvolumescoveringthetheory, foundationsand applicationsof the newemergingmultidisciplinaryfield of social networks analysis and mining. LNSN publishes peer- reviewed works (including monographs, edited works) in the analytical, technical as well as the organiza- tional side of social computing, social networks, network sciences, graph theory, sociology, semantic web, web applications and analytics, information networks, theoretical physics, modeling, security, crisis and risk management, and other related disciplines. The volumes are guest-edited by experts in a specific domain. This series is indexed by DBLP. Springer and the Series Editors welcome book ideas from authors. Potential authors who wish to submit a book proposalshould contactAnneliesKersbergen,PublishingEditor,Springer e-mail:[email protected] Thirimachos Bourlai (cid:129) Panagiotis Karampelas (cid:129) Reda Alhajj Editors Disease Control Through Social Network Surveillance Editors ThirimachosBourlai PanagiotisKarampelas MultispectralImageryLab—MILAB,ECE DepartmentofInformaticsandComputers UniversityofGeorgia HellenicAirForceAcademy Athens,GA,USA AcharnesAttica,Greece RedaAlhajj DepartmentofComputerScience UniversityofCalgary Calgary,AB,Canada ISSN2190-5428 ISSN2190-5436 (electronic) LectureNotesinSocialNetworks ISBN978-3-031-07868-2 ISBN978-3-031-07869-9 (eBook) https://doi.org/10.1007/978-3-031-07869-9 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNatureSwitzerland AG2022 Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether thewhole orpart ofthematerial isconcerned, specifically therights oftranslation, reprinting, reuse ofillustrations, recitation, broadcasting, reproductiononmicrofilmsorinanyotherphysicalway,and transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar ordissimilarmethodologynowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland Preface DiseaseControl ThroughSocial Network Surveillance Thegeneraltopicofdiseasecontrolthroughsocialnetworksurveillanceiscompli- catedandrequiresanoptimalbalanceofingredients.Theintentionofthisbookwas not to generate a teaching textbook and to cover all possible areas related to this topic.Rather,weinvitedexperiencedresearcherstocoveralistofinterestingareas of novel trends and technologies related to disease control including (1) disease control and public health surveillance, (2) social networking surveillance and analysis, and (3) human authentication related studies when utilizing surveillance data and applying deep learning algorithms. These trends and technologies are included in the ten chapters listed below that aim to provide the reader with essential information,knowledge,and understandingon disease controlvia social surveillance,as well as discussion on processes related to the efficient and secure publichealthmonitoring. ChapterContributions In the first chapter, the authors discuss that the analysis of public perceptions and opinion mining (OM) has received considerable attention due to the easy availability of colossal data in the form of unstructured text generated by social media, e-commerce portals, blogs, and other similar web resources. They choose the Twitter platform in their research to study public perceptions regarding the global vaccination drive. More than 112 thousand Tweets from users of different countries around the globe are extracted based on hashtags related to the affairs of the COVID-19 vaccine. A three-tier framework is being proposed in which rawtweetsareextractedandcleanedfirst,visualizedandconvertedintonumerical vectors through word embedding and N-gram models next, and finally analyzed throughasetofmachinelearningclassifierswiththestandardperformancemetrics, v vi Preface accuracy, precision, recall, and F1-measure. The authors show that the bag-of- words(BoW) modeldevelopedachievesthe highest classification accuracy.Their conclusions are that most of the people seem to have a neutral attitude towards thecurrentCOVID-19vaccinationdriveandthatalsopeoplefavoringtheCOVID- 19 vaccination program are greater in number than those who doubt it and its consequences. In the second chapter, the authors argue that computational models for the detection and prevention of false information spreading (popularly called fake news) have gained a lot of attention over the last decade, with most proposed models identifying the veracity of information. Thus, they propose a framework basedona complementaryapproachto false informationmitigationinspiredfrom the domain of epidemiology. In such a domain, false information is analogous to infection, social network is analogous to population, and likelihood of people believing an information is analogous to their vulnerability to infection. As part of the framework, the authors propose four phases that fall in the domain of socialnetworkanalysis.Throughexperimentsonreal-worldinformationspreading networks on Twitter, they show the effectiveness of their proposed models and confirm their hypothesis that spreading of false information (fake news) is more sensitive to behavioral properties, such as trust and credibility, than spreading of trueinformation(realnews). In the third chapter, the authors work also on fake news mitigation strategies. They argue that the rapid spread of fake news during the COVID-19 pandemic has aggravated the situation and made it extremely difficult for the World Health Organization(WHO)andgovernmentofficialstoinformpeopleonlywithaccurate scientificfindings.Misinformationdisseminationhasbeensounhinderedthatsocial mediasiteshadtoultimatelyconcealpostsrelatedtoCOVID-19entirelyandallow userstoseeonlytheWHOorgovernment-approvedinformation.Thus,actionhad to be taken because newsreaders lack the ability to efficiently discern fact from fictionandtherebyindirectlyaidinthespreadoffakenewsbelievingittobetrue.In theirwork,theauthorsfocusonhelpinginunderstandingthethoughtprocessofan individualwhenreadinganewsarticle.Theyexpandthespaceofmisinformation’s impactonusersbyconductingasetofsurveystounderstandthefactorsconsumers deemmostimportantwhendecidingwhetherthe informationsomeonereceivesis true or not. Experimental results show that what people perceive to be important in deciding what is true information is different when confronted with the actual articles.Theyconcludethatpriorbeliefsandpoliticalleaningsaffecttheabilityof peopletodetectthelegitimacyoftheinformationreceived. The fourth chapter discusses the topic of trends, politics, sentiments, and mis- information during COVID-19. The authors conduct a large-scale spatiotemporal data analytics study to understand peoples’ reactions to the COVID-19 pandemic during its early stages. In particular, they analyze a JSON-based dataset that is collected from news, messages, boards, and blogs in English about COVID-19 overaperiodof4months,foratotalof5.2millionposts.Thedatawerecollected from December2019 to March 2020 from severalsocial media platformssuch as Facebook, LinkedIn, Pinterest, StumbleUpon, and VK. The study aims mainly to Preface vii understandwhichimplicationsofCOVID-19haveinterestedsocialmediausersthe most and how did they vary over time, as well as determining the spatiotemporal distributionofmisinformation,andthepublicopiniontowardpublicfiguresduring thepandemic.Theauthorsclaimthattheirresultscanbeusedbymanystakeholders (e.g.,governmentofficialsandpsychologists)tomakemoreinformativedecisions, consideringtheactualinterestsandopinionsofthepeople. In the fifth chapter, the authors present a Data-Network Science study on a datasetofpublicationsarchivedin“TheSemanticScholarOpenResearchCorpus” (S2ORC) database and categorized under the area of “Disease Control through Social Network Surveillance,” an area abbreviated from now on as “DCSNS.” In particular, their dataset consists of 10,866 documents (articles and reviews), retrieved through a Boolean search, published in the period from 1983, the first year of cataloguing such publications in S2ORC, to 2020. By retrieving the corpus of abstracts of these documents (publications) and applying the standard LDA Topic Modeling technique, the authors claim to have found the optimal number of six topics producing the maximum topic coherence score among the corresponding topic models with varying numbers of topics. In that matter, the network of their study becomes a directed citation graph of publications in the area of DCSNS, with nodes and publications labeled by the Topics. Their aim is to study global and local network properties with regards to clustering under triadic relationships among connected nodes/publications,and with regardsto the assortativityofattributesrelatedtothecontentofpublications.Theyclaimthatthey have succeeded in analyzing the interplay between semantics and structure in the area of publications on DCSNS, by examining and discovering the occurrence of certain important attributes of publications in such a way that the aggregation of publications according to these attributes is associating the meaning of attribute affiliationstocertainstructuralpatternsofclustering,exhibitedbythebibliographic citationnetworkofthecollectedpublications. Thesixthchapterfocusesonprivacyinonlinesocialnetworks.Theauthorsstart their work by arguingthat disease controlthroughonline social networks(OSNs) hasbecomeparticularlyrelevantinthepastfewmonths.Giventhesensitivenature ofthedatacollectedandmanipulatedinthatcontext,amajorconcernfor(potential) usersofsuchsurveillanceapplicationsisprivacy.Theconceptofprivacyhasbeen studiedfrommanydifferentangles,andthisworkaimstoofferageneralsystematic literature review of the area. The contributionsof their book chapter are twofold. Firstly,theyproposeasystematicmappingstudycoveringpapersrelatedtoprivacy inOSNs.Thisstudyresultsinacoarse-grainedoverviewofthelandscapeofexisting worksinthefield.Inthisfirstphase,345paperswereexamined.Thefindingsshow the characteristicsandtrendsof publicationsin the area. Theyalso emphasizethe areas where there is a shortage of publications, hence guidingresearchersto gaps in the literature. Secondly, they propose a classification framework and apply it to the subset of 108 papers that offer a solution to protect the user’s privacy.The resultsprovideawayforresearcherstopositionasolutionincomparisonwithother existing solutions and they also highlight trends in existing solutions. The main viii Preface practical implications of this book chapter are guidelines and recommendations proposedtodesignersofapplications,includingapplicationsfordiseasecontrol. Intheseventhchapter,theauthorsarguethatthehealthcrisisbroughtaboutby COVID-19hasresultedinaheightenednecessityforproperandcorrectinformation dissemination to counter the prevalence of fake news and other misinformation. Doctors are the most reliable source of information regarding patients’ health status, disease, treatment options, or necessary lifestyle changes. Prior research has tackled the problem of influence maximization (IM), which tries to identify the most influential physicians inside a physician’s social network. However, less research has taken place on solving the problem of volume maximization (VM), which deals with finding the best set of physicians that maximize the combined volume(e.g.,medicineprescribed)andinfluence(i.e.,informationdisseminated).In thischapter,theprimaryobjectiveoftheauthors’workistoaddresstheVMproblem byproposingdifferentalgorithmicframeworks,includingareinforcementlearning (RL) one. The authors compared the frameworks while using the physicianSN dataset(physiciansocialnetwork181nodesand19,026edges)andtesteddifferent algorithms. Their research highlights the utility of using reinforcement learning algorithms for finding critical physicians that can swiftly disseminate critical informationtobothphysiciansandpatients. Intheeighthchapter,theauthorsfocusonastudythatdiscussestheperceptionof IndianpopulationonCOVID-19vaccineshortageduringtheperiodofarapidhike ofcasesintheCOVID-19secondwave.UsingaTwitterAPI,46,000uniquetweets of Indian citizens have been scrapped, which include the following key words, namely“vaccine,”“secondwave,”and“COVID.”Theauthorsusedatopicmodel. Inmachinelearningandtweetprocessing,atopicmodelisatypeofstatisticalmodel fordiscoveringallabstracttopics,whichhappentoappearinacollectionoftweets (in this study of tweets relevantto the secondwave of COVID). In practice, topic modeling is a text-mining tool that is occasionally used in tweets for the purpose of discovering hidden semantic structures in the tweet text body. In this work, it was used to analyze a set of key themes based on the perception of people. The study shows thatthe Indianpopulationis concernedaboutvaccine shortage and a collective effortis recommendedto be followedfor the improvementof the well- beingoftheIndiancitizens. In the ninth chapter, the authors work on a case study relevant to a pandemic era. They aim to address the problem of face detection in the MWIR (thermal) spectrumwhenthefacesareoccludedwithfacemasks.Sincethepubliclyavailable datasets are not large enough to train original models, transfer learning is used on models trained on the COCO dataset. The models are first trained and tested on masked face images, which results in high precision and recall values for all models. Then, these models are tested on masked face images, and the precision and recall metrics drop significantly. Performance drops also when the models proposedaretestedonmaskeddatawithamarginalbutnoteworthyincreaseofthe inference time. Then, the proposed models are trained and tested on masked face data, and they yield an 89.4% precision and 92.1% recall rates, respectively. The improvedfacerecognitionresultswhenusinganefficientautomatedfacedetection Preface ix approachfurtherdemonstratetheimportanceofsuchmodelsoperatingintheMWIR spectrum.Thisworkisproposedtobefurtherextendedtoscenarioswherethedata usedarecollectedinthevisiblespectrumandunderconstrainedconditions,aswell as to scenarios where the data used are collected in both the thermal and visible bands and under outdoor, unconstrained settings. The study concludes that with sufficientrealdata,efficientunifiedmodelsforeachbandthatdetectshumanfaces ateachdistancecanbedeveloped. The tenth chapter investigates the problem of face mask compliance classifi- cation in response to the coronavirus disease (COVID-19) pandemic. Since the start of the pandemic, many governments and businesses have been continually updating policies to help slow the spread of the virus, including requiring face masks to use many public and private services. In response to these policies, many researchers have developed new face detection and recognition techniques formaskedfaces,almostexclusivelyfocusingondetectingthepresenceorabsence of someone wearing a face mask or not. In this work, the authors investigate the capability of modern classification algorithms to efficiently distinguish between maskedfaceimages,capturedinthevisibleandthermalbands,andwhichwornin complianceornotwiththesuggestedguidelinesprovidedbyhealthorganizations. The approach proposed is deep learning (DL) based and is composed of the creationofamulti-spectralmaskedfacedatabasefromsubjectswearingfacemasks or not; then, it continues with the augmentation of the generated database with synthetic face masks to simulate two different levels of non-compliant wearing of face masks; and finally, it assesses a variety of DL-based architectures, on the previous augmented database, to investigate the efficiency of different classifiers on face mask compliance when operating in either the visible or thermal bands. Experimentalresultsshowthatfacemaskcomplianceclassificationinbothstudied bandsyieldsa classification accuracythatreaches100%formost modelsstudied, when experimenting on frontal face images captured at short distances and with adequateillumination. The eleventh and final chapter discusses the effectiveness of periocular-based human authentication algorithms when wearing face masks that aim to slow the spread of viruses, such as the COVID-19. According to a study published by the National Institute of Standards and Technology (NISTIR 8311), the accuracy of facial recognition algorithmsis reduced between 5% and 50% when comparedto the accuracy yielded by the same algorithms when the subjects are not wearing face masks. The same report also states that face images of subjects wearing masks can increase the failure to enroll rate (FER) more frequently than before. In addition, it is discussed that masked face images lower the efficiency of surveillance (unconstrained) face recognition systems, which become even more pronetoerrorduetoocclusion,distance,cameraquality,outdoors,andlowlight.In this book chapter, the authors focus on the effectiveness of dual-eye periocular- based recognition algorithms when the subjects are wearing face masks under controlled and challenging conditions, and when the face images are captured in both the visible and MWIR (mid-wave infrared) bands. The authors first utilize MILAB-VTF(B),achallengingmulti-spectralfacedatasetcomposedofthermaland