Summary of "Week 6.2 eCrime on Online Social Media"

This lecture, part of a course on Privacy and Security in Online Social Media, focuses on e-crimes related to Link Farming and spam on Twitter. Building on the previous session’s overview of e-crimes on social networks, this part delves deeper into specific problems, research findings, and data analysis related to Spam Campaigns and manipulation techniques in Twitter’s social graph.

Main Ideas and Concepts

Recap of e-Crimes on Social Networks
- Different types of crimes on social networks affect social reputation.
- Malicious users exploit social interactions for harmful purposes.
PageRank and Link Farming
- PageRank: A ranking algorithm used by search engines like Google, where a page’s rank increases with the number and quality of inbound links (in-degree).
- Link Farming: The creation of artificial or non-legitimate reciprocal links between websites to inflate PageRank. This is considered spamming (also called spamdexing or spamexing).
- Link Farming differs from legitimate linking as it involves creating links that wouldn’t naturally exist to manipulate rankings.
Link Farming on Twitter
- Twitter acts as a "web within the web" with real-time news and massive data generation.
- Twitter search results are influenced by factors like follower count, verified status, and social connections, similar to PageRank principles.
- Link Farming on Twitter involves spammers following many users to increase their follower count (in-degree), hoping for reciprocal follows to boost their influence and visibility in search results.
- Legitimate popular users also benefit from high in-degree, blurring lines between spam and genuine influence.
Social Reputation and Influence Metrics
- Number of followers is a proxy for social reputation and influence.
- Klout Score: A metric that aggregates online social media presence into a numerical influence score (1-100 scale). Used in research to measure user influence.
Differences Between Link Farming on Web and Twitter
- On the web, hyperlinks are unidirectional and reciprocal linking is not guaranteed.
- On Twitter, there is a higher probability of reciprocal following due to social norms and platform mechanics.
Research on Spam and Automation on Twitter
- Spam Campaigns: Large-scale operations controlling hundreds of thousands of accounts that persist for months.
- Twitter suspended over 1.1 million accounts in 7 months for disruptive/spam activities.
- About 77% of spam accounts get suspended within a day of their first tweet.
- Approximately 8% of URLs shared on Twitter link to phishing, malware, or scams.
- Twitter’s click-through rates on spam links are higher than email spam due to trust in social connections.
- Around 16% of active Twitter accounts show high automation, with some bots spoofing browser sources to disguise automation.
Data Set and Analysis from 2009 Twitter Data
- Dataset: 54 million users, 1.9 billion links, collected in 2009.
- Definitions:
  - Follower: User B follows user A.
  - Following: User A is followed by user B.
  - Spam Targets: Users targeted by spam.
  - Spam Followers: Accounts that follow spam accounts.
- 379,340 accounts suspended between August 2009 and February 2011 due to spam or inactivity.
- 41,352 suspended accounts posted at least one blacklisted URL shortened by services like Bitly or TinyURL.
- Overlap: 82% of spam followers also appear as spam targets, indicating reciprocal spam relationships.
Key Findings on Spammers’ Influence
- Spammers appear among highly ranked users by PageRank/in-degree:
  - 7 spammers in the top 10,000 users.
  - 304 spammers in the top 100,000 users.
  - 2,131 spammers in the top 1 million users.
- This shows spammers successfully increase their in-degree, leveraging Link Farming to boost visibility.

Methodology / Instructions Highlighted

Understanding PageRank:
- Count inbound links (in-degree).
- Higher in-degree → higher PageRank → better search result ranking.
Identifying Link Farming:
- Look for unnatural reciprocal linking among websites or Twitter accounts.
- On Twitter, track mass following and reciprocal follows to detect Link Farming.
Measuring Influence:
- Use metrics like follower count and Klout Score.
- Analyze user rank by PageRank/in-degree.
Detecting Spam and Automation:
- Analyze account suspension data.
- Monitor URLs posted, especially shortened.