Social media client fingerprinting

US12197473B1 · kind B1 · utility

Assignee

• The Government of the United States, as represented by the Director, National Security Agency · US

Inventors

• Brian M. Baughman · Silver Spring, US
• Rodrigo Gomez · Columbia, US
• Ryan L. Kaliszewski · Columbia, US
• Shawn L. Robertson · Wellington, NZ

Key dates

• Filing date: Nov 4, 2021
• Grant date: Jan 14, 2025
• Expiry date: Aug 22, 2042

Classification

• Technology area (CPC G): Physics
• CPC primary: G06N20/00
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

A method is disclosed for identifying related social media accounts of a social media platform, based on social media clients used to interact with the platform. The method includes obtaining a first dataset of information including, for each interaction between the accounts and the platform, the client used. The method also includes using the first dataset to generate a pruned dataset including accounts, clients, and numbers of interactions by using an unsupervised machine learning algorithm to identify and exclude outlier accounts and outlier clients. The method also includes normalizing the pruned dataset across accounts. The method also includes identifying a connected sub-component of the pruned dataset. The method also includes performing topic modeling of the sub-component to generate a matrix W associating the accounts of the sub-component with topics. The method also includes determining, using the matrix W, that a first account A is associated with a second account B.