Summary of "Week 8.2: Anonymous Networks"

Summary of Week 8.2: Anonymous Networks

This lecture segment from the "Privacy and Security in Online Social Media" course focuses on anonymous social networks, their characteristics, user behavior, and how they differ from traditional social networks like Facebook and Twitter.

Main Ideas and Concepts

Definition of Anonymous Networks: Networks where users post content without revealing their identity, making it difficult or impossible to attribute posts to specific individuals.
Examples of Anonymous Networks:
- 4chan
- Whisper
- Secret
- Yik Yak
- Wickr
Need for Anonymous Networks:
- Growing awareness and demand for privacy online.
- Incidents like Snowden revelations (PRISM project) and viral posts with real-world consequences motivate users to seek anonymity.
- Traditional networks clearly attribute posts to user identities, which anonymous networks avoid.
Whisper as a Case Study:
- A confessional app encouraging anonymous posting of secrets or confessions.
- Posts ("whispers") are often merged with stock photos, creating meme-like content.
- Users can change usernames frequently, enhancing anonymity.
- No collection or archiving of personal user information or persistent social links.
- Interactions include "hearts" (likes) and replies, all anonymous.

Key Terminologies

Whispers: Anonymous posts on Whisper.
Replies: Responses to whispers, similar to comments on Facebook or Twitter.
Hearts: Equivalent of likes, given anonymously.
GUID (Global User Identifier): Unique user ID used in data collection but removed by Whisper after privacy concerns.

Research Questions Addressed

How do users interact in an anonymous environment like Whisper?
Do users form communities similar to those in traditional social networks?
Does anonymity eliminate strong ties (close relationships) between users?
Does anonymity reduce "stickiness" or long-term user engagement?

Data Collection and Analysis

Data from Whisper (2014):
- 9 million whispers
- 15 million replies
- 1 million unique user IDs (GUIDs, later removed by Whisper)
Posting and Reply Behavior:
- 55% of whispers receive no replies.
- 25% have at least two replies.
- 54% of replies arrive within one hour; 94% within one day.
- If a Whisper doesn’t get a reply within a day, it likely won’t get one later.
User Activity Patterns:
- 80% of users post fewer than 10 whispers/replies combined.
- 15% only post replies (no original whispers).
- 30% only post whispers (no replies).

Network Analysis Comparison (Whisper vs Facebook vs Twitter)

Metric Whisper Facebook Twitter Number of nodes 690,000 707,000 4,317,000 Average degree 9.47 1.78 3.93 Clustering coefficient 0.033 0.059 0.048 Average path length 4.28 10.13 5.52 Assortativity coeff. -0.011 (higher) (higher)

Interpretations: - Whisper users interact with many strangers (high average degree). - Low clustering coefficient indicates users’ contacts are unlikely to know each other. - Short average path length suggests a random graph structure, unlike the "small world" phenomenon in Facebook/Twitter. - Negative assortativity confirms randomness in connections.

Content Deletion and Moderation

18% of whispers are deleted (compared to 4% on Twitter).
Deletion mostly occurs within the first week, peaking between 3-9 hours after posting.
Categories with highest deletion rates: sexting, selfies, chat topics.
Categories with lowest deletion rates: emotion, religion, entertainment, life stories, work, politics.
Deletion indicates users’ concern over content even in anonymous settings.

User Interaction and Geography

90% of interacting user pairs are located within the same state.
75% are within 40 miles of each other, showing localized interaction despite anonymity.
Higher user population density in an area correlates with more interactions.
Users who post more are more likely to interact.