A novel approach toward cyberbullying with intelligent recommendations using deep learning based blockchain solution

Area of Science:

• Computer Science
• Artificial Intelligence
• Cybersecurity

Background:

• Cyberbullying is a growing problem on social media platforms.
• Existing solutions often lack robust privacy features and effective detection mechanisms.
• There is a need for decentralized, privacy-preserving AI solutions for online safety.

Purpose of the Study:

• To introduce a novel approach for combating cyberbullying using Blockchain and Federated Learning (FL).
• To develop a decentralized AI solution that prioritizes user privacy and enhances online safety.
• To formally model social connections and analyze cyberbullying patterns.

Main Methods:

• Integration of Blockchain technology with Federated Learning (FL) for decentralized AI.
• Utilizing Alloy Language for formal modeling of social connections.
• Employing a two-phase approach with Long Short-Term Memory (LSTM) for feature development and Dynamic Bayesian Networks (DBN) for relation testing on blockchain data.

Main Results:

• The proposed method demonstrates effective cyberbullying detection through a novel DBN and LSTM integration.
• Formal modeling of social connections provides insights into cyberbullying dynamics.
• Performance evaluation against previous research using standard metrics indicates the efficacy of the approach.

Conclusions:

• The combination of Blockchain and FL offers a transformative vision for decentralized cyberbullying defense.
• The developed method enhances user privacy and fosters a safer online environment.
• This research provides a benchmark for real-world applications in healthcare and predictive modeling for online safety.