Crowdsourcing Opportunity or New Threat Major Area Exam

Crowdsourcing: Opportunity or New Threat? Major Area Exam: June 12 th Gang Wang Committee: Prof. Ben Y. Zhao (Co-chair) Prof. Heather Zheng (Co-chair) Prof. Christopher Kruegel

Why Crowdsourcing • Software automation replaces the role of human in many areas – Store and retrieve large volumes of information – Perform calculation • Human still outperform computer in many ways 1

Searching for Jim Gray (2007) • Jim Gray, Turing Award winner • Missing with his sailboat outside San Francisco Bay, Jan 2007 • No result from searches of coastguard and private • Use satelliteplanes image to search for Jim Gray’s sailboat • Problem: the search cannot be automated by computer • Solution – Split the satellite image into many small images – Volunteers look for his boat in each image 100, 000 tasks completed in 2 days 2

Traffic Monitoring (2012) Traffic Jam! • Google Map traffic monitoring – Large area – Real-time update • Previous approach – Deploy sensors on the road – Expensive equipment No Traffic Here! 3

User-driven traffic monitoring – 200 million Google Map users (mobile) 1 – Report location while driving – Integrate the traffic map in real-time Newsflash: Apple also builds crowdsourced map system in i. OS 6 this fall 1. http: //techcrunch. com/2011/05/25/google-maps-for-mobile-stats 4

Crowdsourcing: a process that enlists a crowd to do micro-work to solve problems that software cannot do Tasks/Pr oblems Solution Effective when a big problem can be decomposed into small tasks that are easy for individuals to solve 5

Crowdsourcing Workflow • Requester – Submit tasks, integrate results • Platform – Manage tasks and workers – E. g. Amazon Mechanical Turk • Workers – Work on tasks, return results – Large number of human users Requester Submit Tasks Collect Results Platform Distribute Tasks Return Results Workers 6

State of the Art • Popular crowdsourcing services – Amazon Mechanical Turk, Free. Lancer – Tasks: translation, transcription, product survey, etc. • Other success stories – Wikipedia – Protein folding • However, there are problems Unfolded Folded 7

Misuse of Crowdsourcing “Over 40% of New Mechanical Turk Jobs Involve Spam” “In a Race to Out-Rave, 5 -Star Web Reviews Go for $5” “Dairy Giant Mengniu in Smear Scandal” “Hacked Emails Reveal Russian Astroturfing Program” • Difficult to detect – High quality spam – Deceptive product reviews – Realistic fake accounts • Emerging threat to online communities 8

Crowdsourcing Research Economics • Business models • Market survey • Labor economics Social Science • Social, cultural, and ethical issues Computer Science System/Applications Security HCI NLP Fake Review Sybils IR DB Social Spam SEO 9

Outline • • Introduction Overview of Crowdsourcing Applications Research Challenges Security and Crowdsourcing 10

Crowdsourcing Applications Categorize applications based on human intelligence • Natural language processing (NLP) – Data labeling [Snow 2008] [Callison-Burch 2009] – Searching results validation [Alonso 2008] – Database query: Crowd. DB [Franklin 2011], Qurk [Marcus 2011] • Image processing – Image annotation [Ahn 2004] [Chen 2009] – Image search [Yan 2010] • Content generation/knowledge sharing – Wikipedia, Quora, Yahoo! Answers, Stack. Overflow – Real-time Q&A: Vizwiz [Bigham 2010], Mimir [Hsieh 2009] • Human sensor – Google Map traffic monitoring – Twitter earthquake report [Sakaki 2010] 11

Crowdsourcing Applications Categorize applications based on human intelligence • Natural language processing (NLP) – Data labeling [Snow 2008] [Callison-Burch 2009] – Searching results validation [Alonso 2008] – Database query: Crowd. DB [Franklin 2011], Qurk [Marcus 2011] • Image processing – Image annotation [Ahn 2004] [Chen 2009] – Image search [Yan 2010] • Content generation/knowledge sharing – Wikipedia, Quora, Yahoo! Answers, Stack. Overflow – Real-time Q&A: Vizwiz [Bigham 2010], Mimir [Hsieh 2009] • Human sensor – Google Map traffic monitoring – Twitter earthquake report [Sakaki 2010] 12

Data Annotation • Natural Language Processing (NLP) problems – Evaluating machine translation quality [Callison-Burch 2009] – Labeling text content (e. g. emotions) [Snow 2008] • Challenges – Difficult for software automation – Experts are expensive and slow • Benefits of using crowdsourcing – Non-experts are cheap and fast – Non-expert results (processed) are as good as experts 13

Image Search Crowd. Search [Yan 2010] • Accurate image searching for mobile devices by combining – Automated image searching – Human validation of searching results via crowdsourcing Automated Image Search Query Image Crowd Validation Candidate Images Only 25% accuracy Res ult Accuracy > 95% Use human intelligence to improve automation 14

Question and Answer Viz. Wiz [Bigham 2010] • Help blind people • Answer questions • Near real-time “Take a photo” “Record the question” “Right side” “Which item is corn? ” • Example – Shopping scenario Server • Use the crowd to help people in need • Replace an “expensive” personal assistant 15

Outline • • Introduction Overview of Crowdsourcing Applications Research Challenges Security and Crowdsourcing 16

Challenges in Crowdsourcing • Quality control – High diversity in worker background and expertise • Incentives – Encourage participation – Improve work quality • Task management – Perform complex/real-time tasks – Coordinate workers and requesters • Security – Spammy/cheating workers, fraud requesters – Using crowdsourcing systems for malicious attacks 17

Quality Control • Fundamental problem: the crowd is not reliable – [Oleson 2011] [Snow 2008] [Callison-Burch 2009] [Yan 2010] [Franklin 2011] – Workers make mistakes – Workers spam the system • Existing strategies – Majority voting – Pre-screening to test workers – Statistic models to clear data bias Yes Yes No Ground Truth Screenin g Test 18

Incentives • Basic questions: how to set the right price of the tasks? – Can you improve work quality by raising payment? – Can you attract more workers by raising payment? • Empirical study on worker incentives [Mason 2010] [Hsieh 2010] – High payment helps to recruit workers faster and increase participation – Money does not improve quality – Punishment/bonus based quality control • Pay the minimum $0. 01 for all workers and $0. 01 for bonus • Common problem for all applications 19

Task Management Example: Writing a travel book for New York City • Crowdsource complex tasks [Kittur 2011] Example: use bubble sort algorithm to sort pictures – Partition tasks Partition the complex … Brief History Attractions (outline)execute each work flow – Parallel – Integrate results … Task 2 Task 1 Task 3 • Map (gather facts) Implement algorithms • Reduce (collect text) Open problems … … on the crowd Task 1 Task 2 [Little 2010] one is better? unit – Regard the. Which crowd as computation – Design/organize the tasks in a way to run algorithms Paragraph – Real-time crowdsourcing, parallel tasks execution, synchronization 20

Security Challenges • Attacks inside crowdsourcing systems – Spammy workers give random/bad answers – Dishonest requesters • Using crowdsourcing system to carry out malicious campaigns – Real-user can perform all kinds of malicious tasks – Crowdsourcing makes it possible to scale • • • Write fake reviews Create fake accounts (Sybils) Generate social network spam Solve CAPTCHA Give biased voting Build back links (SEO) 21

Outline • • Introduction Overview of Crowdsourcing Applications Research Challenges Security and Crowdsourcing – Malicious Crowdsourcing Systems – Fake Reviews Generation by the Crowd – Detecting Sybils in Online Social Networks 22

Threat Map Sybils Spam OSNs Greyhat SEO Search Engine Fake Reviews Online Review 23

Dark Side of Crowdsourcing • Crowdsourcing – Large number of workers – Easy, cheap, fast – Real users can do bad jobs 5 star rating and positive review Use different IP addresses Bypass existing spam filter 24

Measuring Malicious Crowdsourcing • Crowdsourcing malicious tasks – Generate Spam, solve CAPTCHA, create fake accounts, Greyhat SEO • Scale and economics – Zhubajie (China): malicious jobs 10 K/month, with $1 M/month [Wang 2012 b] – Free. Lancer (US): malicious jobs 140 K/7 years [Motoyama 2011] • Emerging threat – International work force – Growing exponentially Buyers Figure from [Motoyama 2011] Workers 25

Outline • • Introduction Overview of Crowdsourcing Applications Research Challenges Security and Crowdsourcing – Malicious Crowdsourcing Systems – Fake Reviews Generation by the Crowd – Detecting Sybils in Online Social Networks 26

Online Reviews: Why Important? • 80% of people will check online reviews before purchasing products/travel online. 1 • Independent restaurants: a one-star increase in Yelp rating leads to a 9% increase in revenue. 2 1 2 http: //www. coneinc. com/negative-reviews-online-reverse-purchase-decisions Michael Luca. Reviews, reputation, and revenue: The Case of Yelp. com. Harvard Business School Working Paper, 2011 27

Detecting Fake Reviews • Detecting review spam [Jindal 2008] – Duplicated/Near-duplicated reviews Dataset Reviewer Products Reviews Spam Reviews Amazon 2, 146, 048 1, 195, 133 5, 838, 032 55, 319 • Detecting review spammers Lower bound of spam reviews – Classify rating/review behaviors [Lim 2010] – Detect synchronized reviews in groups [Mukherjee 2012] • Deception models [Ott 2011] – Content classification using trained data – Psycholinguistic deception detection Human accuracy 60% Classifier accuracy 90% 28

Challenges to Detect Fake Reviews • Current review spam detection solution is limited – Assume a few attackers control many accounts – Crowdsourcing can break these assumptions • Deception model (NLP approach) has limitations – Domain specific – Content analysis still has high false positive (10%) • Detecting fake reviews is an open problem – Low false positive – Real-time 29

Outline • • Introduction Overview of Crowdsourcing Applications Research Challenges Security and Crowdsourcing – Malicious Crowdsourcing Systems – Fake Reviews Generation by the Crowd – Detecting Sybils in Online Social Networks 30

Social Network Sybils • Sybils in Online Social Networks (OSNs) – Cheating in social games – Spreading spam/malware – [Thomas 2011], [Gao 2010], [Nazir 2010] • Challenges to detect Sybils in the wild – Various/adaptive Sybil behavior patterns/attack strategies – Increasingly sophisticated/realistic Sybil account profiles – Automated mechanisms losing effectiveness • Use crowdsourcing for Sybil detection 31

Crowdsourced Sybil Detection • Basic idea: build a crowdsourced Sybil detector – Resilient to changing attacker strategies • Question: can human identify Sybil profiles? (answer: user study) – Ground truth datasets of full user profiles • 200 real + 180 fake accounts (Renren, Facebook-India) – Segmented user groups • Renren users (Chinese), Facebook (US), Facebook (Indian) • Experts (conscientious, motivated), Turkers (paid per profile, $driven) • High level results – Experts are accurate; both experts and turkers have near-zero false positives – Quality control can improve turker accuracy ~ experts – Accurate, scalable, cost-effective 32

Conclusion • An alternative solution to various problems – Difficult to be automated by software – Can be decomposed into small tasks • Many challenges in crowdsourcing system – Quality control against unreliable workers – Task management for complex tasks – Incentive models to reduce cost and optimize performance • Malicious crowdsourcing and related attacks – Serious threat to existing security mechanisms – Measurement study to understand the problem – Defense is still an open problem 33

Possible Research Areas • Defend against malicious crowdsourcing systems – Attacking malicious crowdsourcing systems – Detecting crowdsourcing campaigns in real-time • Spot fake reviews/reviewers – Resilient to changing behaviors – Real-time – Scalability • Using crowdsourcing to solve security problems – Crowdsourcing to detect social Sybils 34

Thank you! Questions? 35