Table 2 Statistics of the synthetic decentralized social media Dataset, including number of users, posts, modality breakdown, and threat type frequencies.

From: Privacy-preserving cyberthreat detection in decentralized social media with federated cross-modal graph transformers

Statistic

Value

Number of Users

5,000

Number of Instances (Servers)

50

Total Posts

7,20,000

Text Posts

3,20,000

Image Posts

2,20,000

Audio Clips

1,80,000

Avg. Posts per User

144

% Multi-modal Posts

41%

Threat Event Frequency

4.5% of total posts

Major Threat Types

Cyberbullying (28%), Hate Speech (21%), Misinformation (24%), Coordinated Harassment (17%), Scam/Phishing (10%)

Avg. Degree (Social Graph)

17.6

Clustering Coefficient

0.18

% Nodes in Largest Component

93%