One plus one makes three (for social networks).

Area of Science:

• Social Network Analysis
• Information Security
• Machine Learning

Background:

• Social network platforms encourage users to share private information for access to services.
• This sharing creates a trade-off between privacy and platform utility.
• The extent of privacy sacrifice is often underestimated.

Purpose of the Study:

• To investigate if combined contact data can infer relationships between non-members.
• To quantify the privacy risks associated with sharing contact information on social networks.

Main Methods:

• Utilized machine learning algorithms to predict relationships.
• Combined data included confirmed member contacts and member email contacts of non-members.
• Evaluated prediction accuracy using the area under the receiver operating characteristic curve (AUC).

Main Results:

• Achieved an AUC of at least 0.85 in predicting connections between non-members.
• This prediction accuracy was maintained even with conservative estimates of user disclosure.
• The combination of contact knowledge is sufficient to infer a substantial proportion of non-member relationships.

Conclusions:

• Sharing contact lists on social media poses significant privacy risks for non-members.
• Even partial disclosure of contacts can lead to the inference of social connections.
• The study highlights the need for greater user awareness regarding data privacy implications.