designing and evaluating techniques to mitigate misinformation spread on micro-blogging web services

65
Designing and Evaluating Techniques to Mitigate Misinformation Spread on Micro-blogging Web Services Adi$ Gupta Under the Supervision of Dr. Ponnurangam Kumaraguru Indraprastha Ins9tute of Informa9on Technology, Delhi July 6, 2015

Upload: precog

Post on 19-Aug-2015

66 views

Category:

Education


0 download

TRANSCRIPT

Designing and Evaluating Techniques to Mitigate Misinformation Spread on

Micro-blogging Web Services"

Adi$  Gupta    

Under  the  Supervision  of  Dr.  Ponnurangam  Kumaraguru  

Indraprastha  Ins9tute  of  Informa9on  Technology,  Delhi  July  6,  2015  

Power of Social Media"

2  

300  hours  of  video  uploaded  every  minute  

500  million  tweets  posted  every  day  

1.44  Billion  monthly  ac$ve  

users  

60  million  photos  shared  

everyday  

*  2015  Sta9s9cs    

Real World Events"

3  

Misinformation on Social Media"

4  

Misinformation on Social Media"

5  

Misinformation on Social Media"

6  

Focus: Twitter"

7  

Profile  Photo  

Hashtag  

Followers  

Retweet  BuOon  Username  

Misinformation Tweets"

FAKE  

RUMORS  

8  

$  

Aim"  

   Designing  and  Evalua9ng  Techniques  to    Mi9gate  Misinforma9on  Spread  on    

Micro-­‐blogging  Web  Services  

9  

Proposed Solution"

10  

�  Learning  to  Rank  model  for  assessing  credibility  of  Tweets  

� Model  based  on  ground  truth  data  for  20  real  world  events  and  45  features    

�  System  evalua9on  using  year  long  real  world  experiment  

�  1800+  users  requested  for  credibility  score  of  more  than        14.2  million  tweets.  

 

TweetCred Demo"

Approach"

12  

Characterizing  Misinforma$on  and  Fake  Content  

 

Ranking  Framework  to  

Assess  Credibility  

Building  and  Evalua$ng  a  Real-­‐

$me  System    

Detec9ng  fake  images  (Hurricane  sandy)    Analyzing  rumor  propaga9on  (Boston  blasts)    Detec9ng  user  communi9es  (three  events)    Analyzing  rumors  spread  in  India  centric  events  (Mumbai  blasts  and  Assam  riots)  

14  events  data  tagging    30%  of  tweets  provide  informa9on  (17%  credible  informa9on    Linear  logis9c  regression    Present  ranking  algorithm  to  assess  credibility  in  tweets  using  pseudo  relevance  feedback  

45  features  computable  for  a  single  tweet    Live  deployment:  1,800+  TwiOer  users      Credibility  score  computed  for  14+  Million  tweets    Evaluated  TweetCred  in  terms  of  response  9me,  effec9veness  and  usability  

Data Collection"� Created  a  24*7  data  collec9on  framework  - Streaming  /  REST  APIs  - JSON  Format  - MySql  Databases    

� Collected  2+  Billion  tweets  from  2011-­‐14  

13  

Approach"

14  

Characterizing  Misinforma$on  and  Fake  Content  

 

Ranking  Framework  to  

Assess  Credibility  

Building  and  Evalua$ng  a  Real-­‐

$me  System    

Detec9ng  fake  images  (Hurricane  sandy)    Analyzing  rumor  propaga9on  (Boston  blasts)    Detec9ng  user  communi9es  (three  events)    Analyzing  rumors  spread  in  India  centric  events  (Mumbai  blasts  and  Assam  riots)  

14  events  data  tagging    30%  of  tweets  provide  informa9on  (17%  credible  informa9on    Linear  logis9c  regression    Present  ranking  algorithm  to  assess  credibility  in  tweets  using  pseudo  relevance  feedback  

45  features  computable  for  a  single  tweet    Live  deployment:  1,800+  TwiOer  users      Credibility  score  computed  for  14+  Million  tweets    Evaluated  TweetCred  in  terms  of  response  9me,  effec9veness  and  usability  

Background: Hurricane Sandy"

� Dates:  Oct  22-­‐  31,  2012  � Damages  worth  $75  billion  � Coast  of  NE  America  

15  

Faking  Sandy:  Characterizing  and  Iden9fying  Fake  Images  on  TwiOer  during  Hurricane  Sandy.  Adi9  Gupta,  Hemank  Lamba,  Ponnurangam  Kumaraguru  and  Anupam  Joshi.  Accepted  at  the  2nd  Interna9onal  Workshop  on  Privacy  and  Security  in  Online  Social  Media  (PSOSM),  in  conjunc9on  with  the  22th  Interna9onal  World  Wide  Web  Conference  (WWW),  Rio  De  Janeiro,  Brazil,  2013.  Best  Paper  Award.  

Fake Image Tweets"

16  

Data Description"

17  

Total  tweets   1,782,526  Total  unique  users   1,174,266  

Tweets  with  URLs   622,860  

Tweets  with  fake  images   10,350  

Users  with  fake  images   10,215  

Tweets  with  real  images   5,767  

Users  with  real  images   5,678  

Network Analysis"

18  

 Tweet  –  Retweet  graph  for  the  propaga9on  of  fake  images  during  first  2  hours  

Node  -­‐>  User  Id  Edge  -­‐>  Retweet      

Role of Twitter Network"� Analyzed  role  of  follower  network  in  fake  image  propaga9on  

�  Crawled  the  TwiOer  network  for  all  users  who  tweeted  the  fake  image  URLs  

19  

�  Graph  1  - Nodes:  Users,    Edges:  Retweets  

�  Graph  2  - Nodes:  Users,    Edges:  Follow  rela9onships  

Results"

20  

Total  edges  in  retweet  network   10,508  

Total  edges  in  follower-­‐followee  network   10,799,122  

Common  edges   1,215  

%age  Overlap   11%  

Classification"    5  fold  cross  valida9on  

21  

Tweet  Features  [F2]  Length  of  Tweet  Number  of  Words  

Contains  Ques9on  Mark?  Contains  Exclama9on  Mark?  Number  of  Ques9on  Marks  

Number  of  Exclama9on  Marks  

Contains  Happy  Emo9con  Contains  Sad  Emo9con  

Contains  First  Order  Pronoun  

Contains  Second  Order  Pronoun  Contains  Third  Order  Pronoun  

Number  of  uppercase  characters  

Number  of  nega9ve  sen9ment  words  

Number  of  posi9ve  sen9ment  words  Number  of  men9ons  Number  of  hashtags  Number  of  URLs  Retweet  count  

User  Features  [F1]  

Number  of  Friends  

Number  of  Followers  

Follower-­‐Friend  Ra9o  

Number  of  9mes  listed  

User  has  a  URL  

User  is  a  verified  user  

Age  of  user  account  

Classification Results"

22  

F1  (user)   F2  (tweet)   F1+F2  

Naïve  Bayes   56.32%   91.97%   91.52%  

Decision  Tree   53.24%   97.65%   96.65%  

•  Best  results  were  obtained  from  Decision  Tree  classifier,  we  got  97%  accuracy  in  predic9ng  fake  images  from  real.    

•  Tweet  based  features  are  very  effec9ve  in  dis9nguishing  fake  images  tweets  from  real,  while  the  performance  of  user  based  features  was  very  poor.    

 

Boston Blasts"�  Twin  blasts  occurred  during  the  Boston  Marathon  

-  April  15th,  2013  at  18:50  GMT  �  3  people  were  killed  and  264  were  injured  �  First  Image  on  TwiOer  (within  4  mins)    

23  $1.00  per  RT  #BostonMarathon  #PrayForBoston:  Analyzing  Fake  Content  on  TwiOer.  Adi9  Gupta,  Hemank  Lamba  and  Ponnurangam  Kumaraguru.  Accepted  at  IEEE  APWG  eCrime  Research  Summit  (eCRS),  San  Francisco,  USA,  2013.  

Sample Fake Tweets"

24  

>  50,000  RTs  

>  30,000  RTs  

Data Description"Total tweets 7,888,374

Total users 3,677,531

Time of the blast Mon Apr 15 18:50 2013

Time of first tweet Mon Apr 15 18:53 2013

25  

Geo-Located Tweets"

26  

Identifying Rumor / True tweets"�  Tagged  most  viral  20  tweet  content  

-  Rumor  /  Fake  -  True  -  Generic  (NA)    

�  Six  Rumors  -  130,690  Tweets  /  Retweets  (29%)  -  R.I.P.  to  the  8  year-­‐old  boy  who  died  in  Boston’s  explosions,  while  running  for  the  Sandy  Hook  kids.  #prayforboston    

�  Seven  True  news  -  116,454  Tweets  /  Retweets  (20%)  -  Doctors:  bombs  contained  pellets,  shrapnel  and  nails  that  hit  vicGms  #BostonMarathon  @NBC6    

�  Seven  Generic  -  206,816  Tweets  /  Retweets  (51%)  -  #PrayForBoston    

Fake Content User Profiles"

Account  1   Account  2   Account  3   Account  4  

No.  of  Followers   10   297   249   73,657  

Profile  Crea$on  Date   Mar  24  2013   Apr  15  2013   Feb  07  2013     Dec  04  2008  

Total  No.  of  Statuses   2   2   294   7,411  

No.  of  Fake  Tweets   2   2   1   1  

Current  Status   Suspended   Suspended   Suspended    Ac9ve  

28  Username:  BostonMarathons  

Temporal Patterns"

29  

Fake  content  /  rumors  becomes  viral  in  first  7-­‐8  hours  just  aoer  the  event.      

Tweet Source Analysis"

30  

76%  

16%  

8%  

Fake  

64%  

31%  

5%  True  

51%  41%  

8%  

General  

Mobile   Web   Others  

Spread of Fake Content"�  Using  linear  regression  �  Predict  how  viral  a  rumor  would  get  

-  Based  on  aOributes  of  users  who  are  propaga9ng  the  rumor  

�  Based  on:  -  Follower  -  Friends  -  Favorited    -  Status  -  Verified    

31  

Predicting Spread of Fake Content"

32  

Results  show  it  is  possible  to  predict  how  viral  a  rumor  would  become  in  future  based  on  aOributes  of  users  currently  propaga9ng  the  rumor.  

Book & Media"

33  

Approach"

34  

Characterizing  Misinforma$on  and  Fake  Content  

 

Ranking  Framework  to  

Assess  Credibility  

Building  and  Evalua$ng  a  Real-­‐

$me  System    

Detec9ng  fake  images  (Hurricane  sandy)    Analyzing  rumor  propaga9on  (Boston  blasts)    Detec9ng  user  communi9es  (three  events)    Analyzing  rumors  spread  in  India  centric  events  (Mumbai  blasts  and  Assam  riots)  

14  events  data  tagging    30%  of  tweets  provide  informa9on  (17%  credible  informa9on    Linear  logis9c  regression    Present  ranking  algorithm  to  assess  credibility  in  tweets  using  pseudo  relevance  feedback  

45  features  computable  for  a  single  tweet    Live  deployment:  1,800+  TwiOer  users      Credibility  score  computed  for  14+  Million  tweets    Evaluated  TweetCred  in  terms  of  response  9me,  effec9veness  and  usability  

Credibility  Ranking  of  Tweets  during  High  Impact  Events.  Adi9  Gupta  and  Ponnurangam  Kumaraguru,  Workshop  on  Privacy  and  Security  on  Online  Social  Media  (PSOSM),  co-­‐located  with  the  21st  Interna9onal  World  Wide  Web  Conference  (WWW),  Lyon,  France,  2012.    

Tweets about an Event"

35  

Tweets  #event  

Informa$on   No  informa$on  

Tweets  with  

informa$on  

Credible  Informa$on  

Non-­‐Credible  

Informa$on  

Fake  news  /  Rumors    Personal  Opinions  /  Spam  

No.  of  people  affected  Place  of  event  Pictures  /  videos          

36  

Architecture"

37  

Data Statistics"Events Tweets Trending Topics

UK Riots 542,685 #ukriots, #londonri- ots, #prayforlondon

Libya Crisis 389,506 libya, tripoli

Earthquake in Virginia 277,604 #earthquake, Earth- quake in SF

JanLokPal Bill Agitation 182,692 Anna Hazare, #jan- lokpal, #anna

Apple CEO Steve Jobs resigns 158,816 Steve Jobs, Tim Cook, Apple CEO

US Downgrading 148,047 S&P, AAA to AA

Hurricane Irene 90,237 Hurricane Irene, Tropical Storm Irene

Google acquires Motorola Mobility 68,527 Google, Motorola Mobility

News of the World Scandal 67,602 Rupert Murdoch, #murdoch

Abercrombie & Fitch stocks drop 54,763 Abercrombie & Fitch, A&F

Muppets Bert and Ernie were gay 52,401 Bert and Ernie

Indiana State Fair Tragedy 49,924 Indiana State Fair

Mumbai Blast, 2011 32,156 #mumbaiblast, Dadar, #needhelp

New Facebook Messenger 28,206 Facebook Messenger 38  

Annotation"�  Step  1  

-  R1.  Contains  informa9on  about  the  event  -  R2.  Is  related  to  the  event,  but  contains  no  informa9on  -  R3.  Not  related  to  the  event  -  R4.  Skip  tweet    

�  Step  2  

-  C1.  Definitely  credible  -  C2.  Seems  credible  -  C3.  Definitely  incredible  -  C4.  Skip  tweet.        

39  

Annotation Results"

40  

�  Each  tweet  annotated  by  3  people    

�  Inter-­‐annotator  agreement  (Cronbach  Alpha)  =  0.748    

�  30%  of  tweets  provide  informa9on  (17%  credible  informa9on)  and  14%  was  spam  

Feature Sets"

41  

Message based features

Length of the tweet

Number of words

Number of unique characters

Number of hashtags

Number of retweets

Number of swear language words

Number of positive sentiment words

Number of negative sentiment words

Tweet is a retweet

Number of special symbols [$, !]

Number of emoticons [:-), :-(]

Tweet is a reply

Number of @- mentions

Number of retweets

Time lapse since the query

Has URL

Number of URLs

Use of URL shortener service

Message based features

Length of the tweet

Number of words

Source based features

Registration age of the user

Number of statuses

Number of followers

Number of friends

Is a verified account

Length of description

Length of screen name

Has URL

Ratio of followers to followees

Source based features

Registration age of the user

Number of statuses

Number of followers

Evaluation Metric"

42  

Evalua9on  Metric:  NDCG  (Normalized  Discounted  Cumula9ve  Gain)          NDCG  is  the  standard  metric  used  to  evaluate  “graded”  results  

Ranking Results"

43  

•  Tweet  and  user  based  features  contribute  in  determining  the  credibility  –  it  maOers  “what  you  post  and  who  you  are”    

PRF"� PRF  (Pseudo  Relevance  Feedback)    - Extract  k  ranked  documents  and  then  re-­‐rank  those  documents  according  to  a  defined  score    - Re-­‐ranking  based  on  ‘top  words’  of  an  event      - Top  n  unigrams  based  on  BM25  ranking  func9on  

44  

Algorithm"

45  

SVM-­‐Rank  

T1  .  .  .  .  

Tn  

T’1  .  .  T’k  .  

T’n  

Extract  top  unigrams  per  

event  

PRFRank  (similarity  metric)  

T’’1  .  .  

T’’k  

Ranking Results"

46  

PRF  ranking  greatly  enhances  the  performance  (upto  .74  NDCG)  

Approach"

47  

Characterizing  Misinforma$on  and  Fake  Content  

 

Ranking  Framework  to  

Assess  Credibility  

Building  and  Evalua$ng  a  Real-­‐

$me  System    

Detec9ng  fake  images  (Hurricane  sandy)    Analyzing  rumor  propaga9on  (Boston  blasts)    Detec9ng  user  communi9es  (three  events)    Analyzing  rumors  spread  in  India  centric  events  (Mumbai  blasts  and  Assam  riots)  

14  events  data  tagging    30%  of  tweets  provide  informa9on  (17%  credible  informa9on    Linear  logis9c  regression    Present  ranking  algorithm  to  assess  credibility  in  tweets  using  pseudo  relevance  feedback  

45  features  computable  for  a  single  tweet    Live  deployment:  1,800+  TwiOer  users      Credibility  score  computed  for  14+  Million  tweets    Evaluated  TweetCred  in  terms  of  response  9me,  effec9veness  and  usability  

TweetCred:  Real-­‐Time  Credibility  Assessment  of  Content  on  TwiOer.  Adi9  Gupta,  Ponnurangam  Kumaraguru,  Carlos  Cas9llo  and  Patrick  Meier.  Proceedings  of  the  6th  Interna9onal  Conference  on  Social  Informa9cs  (SocInfo),  Barcelona,  Spain,  2014.  Honorable  Men$on  for  Best  Paper.      

TweetCred"� Available  as  a  Chrome  Extension  � Rest  API  

Features for Real-time Analysis"

49  

Feature  set      Features  (45)    

Tweet  meta-­‐data     Number  of  seconds  since  the  tweet;  Source  of  tweet  (mobile  /  web/  etc);  Tweet  contains  geo-­‐coordinates  

Tweet  content  (simple)    

Number  of  characters;  Number  of  words;  Number  of  URLs;  Number  of  hashtags;  Number  of  unique  characters;  Presence  of  stock  symbol;  Presence  of  happy  smiley;  Presence  of  sad  smiley;  Tweet  contains  `via';  Presence  of  colon  symbol  

Tweet  content  (linguis9c)    Presence  of  swear  words;  Presence  of  nega9ve  emo9on  words;  Presence  of  posi9ve  emo9on  words;  Presence  of  pronouns;  Men9on  of  self  words  in  tweet  (I;  my;  mine)  

Tweet  author     Number  of  followers;  friends;  9me  since  the  user  if  on  TwiOer;  etc.  

Tweet  network     Number  of  retweets;  Number  of  men9ons;  Tweet  is  a  reply;  Tweet  is  a  retweet  

Tweet  links     WOT  score  for  the  URL;  Ra9o  of  likes  /  dislikes  for  a  YouTube  video  

Training Data"� 500  Tweets  per  event  � Used  CrowdFlower  service  

50  

Event   Tweets   Users  Boston  Marathon  Blasts  (2013)   7,888,374   3,677,531  

Typhoon  Haiyan  /  Yolanda  (2013)   671,918   368,269  

Cyclone  Phailin  (2013)   76,136   34,776  Washington  Navy  yard  shoo9ngs  (2013)   484,609   257,682  

Polar  vortex  cold  wave  (2014)   143,959   116,141  

Oklahoma  Tornadoes  (2013)   809,154   542,049  

 Total       10,074,150   4,996,448  

Annotation"�  Step  1  

-  R1.  Contains  informa9on  about  the  event  -  R2.  Is  related  to  the  event,  but  contains  no  informa9on  -  R3.  Not  related  to  the  event  -  R4.  Skip  tweet  

45%  (class  R1),  40%  (class  R2),  and  15%  (class  R3)      

�  Step  2  -  C1.  Definitely  credible  -  C2.  Seems  credible  -  C3.  Definitely  incredible  -  C4.  Skip  tweet.    52%  (class  C1),  35%  (class  C2),  and  13%  (class  C3)     51  

Ranking Model Evaluation"

52  

AdaRank  Coord.  Ascent   RankBoost  

SVM-­‐rank  

NDCG@25   0.6773   0.5358   0.6736   0.3951  NDCG@50   0.6861   0.5194   0.6825   0.4919  NDCG@75   0.6949   0.7521   0.689   0.6188  NDCG@100     0.6669   0.7607   0.6826   0.7219  

Time  (training)   35-­‐40  secs   1  min   35-­‐40  secs   9-­‐10  secs  Time  (tes$ng)   <1  sec   <1  sec   <1  sec   <1  sec  

Top Ten Features"� No.  of  characters  in  tweet    � Unique  characters  in  tweet    � No.  of  words  in  tweet  � User  has  loca9on  in  profile    � Number  of  retweets  � Age  of  tweet  � Tweet  contains  URL  � Tweet  contains  via  � Statuses  /  Followers  � Friends  /  Followers    

53  

Implementation"

Feedback by Users"

55  

Usage Statistics"

Date  of  launch  of  TweetCred    27  Apr,  2014  

Credibility  score  requests  received   14,234,131  

Unique  TwiOer  users   1,808  

Feedback  was  given  for  tweets   1,654  

Unique  users  who  gave  feedback   364  

56  *  Data  as  on  April’15  

Users of TweetCred"Sample  users:  - Emergency  responders  - Firefighters  - Journalists  /  news  media  - General  users  - Researchers  (Requested  API  tokens)  

57  

System Evaluation"� Usability  Evalua9on  - System  Usability  Scale  (SUS):  70  

� Response  Time  

58  

v

Media"

Limitations & Future Work"� Current  research  focuses  on  TwiOer,  we  would  like  analyze  credibility  of  content  on  different  social  media  using  similar  framework    

� We  would  like  to  enhance  the  current  system  to  indicate  tweets  that  are  9mely,  factual,  well-­‐wriOen,  etc.  

60  

Contributions Summary"�  Analyzed  how  real  and  fake  content  is  propagated  through  the  TwiOer  network,  with  the  purpose  of  assessing  the  reliability  of  TwiOer  as  an  informa9on  source  during  real-­‐world  events.      

�  Proposed  a  learning-­‐to-­‐rank  framework  for  assessing  credibility  of  content  on  TwiOer  using  a  combina9on  of  content,  meta-­‐data,  network,  user  profile  and    temporal  features.    

�  Evaluated  and  deployed  a  novel  framework  for  providing  indica9on  of  trustworthiness  /  credibility  of  tweets  posted  during  events.  

61  

Real world Impact"  �  The  real-­‐9me  system  TweetCred  built  to  assess  credibility  of  content  on  TwiOer  is  used  by  1,808  real  TwiOer  users  to  obtain  credibility  scores  for  more  than  14.2  million  tweets.      

�  A  unique  data  set  of  thousands  of  fake  images,  rumor  tweets  and  malicious  profiles  for  25+  real-­‐world  events.        

 

62  

Publications"�  Peer  Reviewed  Publica9ons  

-  TweetCred:  Real-­‐Time  Credibility  Assessment  of  Content  on  TwiOer.  Adi9  Gupta,  Ponnurangam  Kumaraguru,  Carlos  Cas9llo  and  Patrick  Meier.  Proceedings  of  the  6th  Interna9onal  Conference  on  Social  Informa9cs  (SocInfo),  Barcelona,  Spain,  2014.  Honorable  Men9on  for  Best  Paper.    

-  $1.00  per  RT  #BostonMarathon  #PrayForBoston:  Analyzing  Fake  Content  on  TwiOer.  Adi9  Gupta,  Hemank  Lamba  and  Ponnurangam  Kumaraguru.  Accepted  at  IEEE  APWG  eCrime  Research  Summit  (eCRS),  San  Francisco,  USA,  2013.  

-  Faking  Sandy:  Characterizing  and  Iden9fying  Fake  Images  on  TwiOer  during  Hurricane  Sandy.  Adi9  Gupta,  Hemank  Lamba,  Ponnurangam  Kumaraguru  and  Anupam  Joshi.  Accepted  at  the  2nd  Interna9onal  Workshop  on  Privacy  and  Security  in  Online  Social  Media  (PSOSM),  in  conjunc9on  with  the  22th  Interna9onal  World  Wide  Web  Conference  (WWW),  Rio  De  Janeiro,  Brazil,  2013.  Best  Paper  Award.  

-  Iden9fying  and  Characterizing  User  Communi9es  on  TwiOer  during  Crisis  Events.  Adi9  Gupta,  Anupam  Joshi  and  Ponnurangam  Kumaraguru.  Workshop  on  Data-­‐driven  User  Behavioral  Modeling  and  Mining  from  Social  Media  (UMSOCIAL),  Co-­‐located  with  21st  ACM  Interna9onal  Conference  on  Informa9on  and  Knowledge  Management  (CIKM),  Hawaii,  USA,  2012.  

-  Credibility  Ranking  of  Tweets  during  High  Impact  Events.  Adi9  Gupta  and  Ponnurangam  Kumaraguru,  Workshop  on  Privacy  and  Security  on  Online  Social  Media  (PSOSM),  co-­‐located  with  the  21st  Interna9onal  World  Wide  Web  Conference  (WWW),  Lyon,  France,  2012.  

-  Beware  of  What  You  Share:  Inferring  Home  Loca9on  in  Social  Networks.  Ta9ana  Pontes,  Gabriel  Magno,  Marisa  Vasconcelos,  Adi9  Gupta,  Jussara  Almeida,  Ponnurangam  Kumaraguru  and  Virgilio  Almeida,  Privacy  in  Social  Data  (PinSoda),  in  conjunc9on  with  Interna9onal  Conference  on  Data  Mining  (ICDM)  (2012).  

63  

Publications"�  Peer  Reviewed  Publica9ons  (Posters)  

-  Analyzing  and  Measuring  Spread  of  Fake  Content  on  TwiOer  during  High  Impact  Events.  Adi9  Gupta,  Hemank  Lamba,  Ponnurangam  Kumaraguru.  Security  and  Privacy  Symposium  IIT,  Kanpur,  2014.  Best  Poster  Winner.  

-  Twit-­‐Digest  Version  2:  An  Online  Solu9on  for  Analyzing  and  Visualizing  TwiOer  in  Real-­‐Time.  Adi9  Gupta,  Mayank  Gupta,  Ponnurangam  Kumaraguru.  Security  and  Privacy  Symposium  IIT,  Kanpur,  2014.  

-  Twit-­‐Digest:  Real-­‐9me  TwiOer  search  portal  for  extrac9ng,  tracking  and  visualizing  informa9on.  Adi9  Gupta,  Akshit  Chhabra  and  Ponnurangam  Kumaraguru.  IBM  ICARE  2012.  2nd  Runner’s  Up  prize  Best  Poster.    

-  U2P2:  Understanding  User  Privacy  Percep9ons,  Niharika  Sachdeva,  Ponnurangam  Kumaraguru  and  Adi9  Gupta,  Poster  at  IBM-­‐ICARE,  2011.  

�  Book  Chapter  -  Misinforma9on  on  TwiOer  during  Crisis  Events.  Encyclopedia  of  Social  Network  Analysis  and  Mining  (ESNAM).  Adi9  Gupta,  Ponnurangam  Kumaraguru.  Book  Chapter.  Springer  publica9ons.  2012.  

64  

Thank  you!      

hOp://twitdigest.iiitd.edu.in/TweetCred/  cerc.iiitd.ac.in