DeeProBot: a hybrid deep neural network model for social bot detection based on user profile data.

Area of Science:

• Artificial Intelligence
• Computer Science
• Social Media Analysis

Background:

• Online social networks (OSNs) facilitate global communication but are exploited by bot accounts spreading misinformation.
• Distinguishing human users from automated bot accounts on platforms like Twitter is a critical research challenge.

Purpose of the Study:

• To propose and evaluate DeeProBot, a novel deep learning framework for classifying Twitter accounts as human or bot.
• To leverage user profile metadata for efficient and accurate bot detection, reducing feature engineering complexity.

Main Methods:

• Developed DeeProBot, a hybrid deep learning model utilizing user profile metadata (description, follower/tweet counts).
• Employed Global Vectors (GloVe) for word representation of textual profile descriptions.
• Integrated Long Short-Term Memory (LSTM) units and dense layers to process mixed data types (numerical, binary, text).

Main Results:

• Achieved a high Area Under the Curve (AUC) of 0.97 on labeled datasets.
• Demonstrated model generalizability by testing across different datasets, showing robust performance.
• Profile-based features proved effective, reducing overhead compared to timeline-based analysis.

Conclusions:

• DeeProBot offers an effective and efficient deep learning approach for Twitter bot detection using profile information.
• The framework's hybrid nature and generalizability make it a valuable tool for combating misinformation on social media.
• Accurate bot detection is crucial for maintaining the integrity of online discourse and public opinion.