AI-POWERED MINDSHIELD MODEL FOR ANALYZING TOXICITY AND THREATS IN ONLINE INTERACTIONS

Authors

  • Shristi Verma Sharda University, School of Computing Science and Engineering Author
  • Pranshi Goyal Sharda University, School of Computing Science and Engineering Author
  • Kalidindi Sowmya Sharda University, School of Computing Science and Engineering Author

Keywords:

Bad online socialization, AI system, cyberbullying detection, natural language processing, semantic embeddings, transparent AI, internet safety.

Abstract

Cyberbullying, hate speech and harassment have been reinforced by the rise of internet platforms. speech, and harassment have been enhanced by the booming nature of internet platforms. These are very difficult roles psychologically, morally and socially. the roles that are very difficult psychologically, morally and socially. In this article, the authors present the system known as MindShield, an artificial intelligence-based system, whose goal is to identify, classify and study dangerous activities online. paper, the authors present the system known as the MindShield that is the artificial intelligencebased system, the purpose of the system is to identify, classify, and study dangerous online activities. Theprogram involves natural language processing (NLP), semantic integration as well as deep learning classifiers to identify the presence of toxic behaviors in text communication. natural language processing (NLP), semantic embeddings as well as deep learning classifiers in identifying presence of the toxic behaviors within the textbased communication. Regarding the mind shield, compared to its conventional moderation mechanisms, MindShield should prioritize the elements of explainability, risk identification and adaptive learning, thanks to which it is able to adapt to the reaction to new categories of cyber threats. as compared to its conventional moderation mechanisms, MindShield ought to prioritize the elements of explainability, risk identification and adaptive learning, whereby it is able to adjust to reaction to new categories of cyber threats. The experimental results are characterized by high performance compared to the benchmark datasets, as well as strong improvement in precision, recall and interpretability. comparison with benchmark datasets, as well as high improvement in precision, recall, and interpretability.

 

Downloads

Published

13-03-2026

Conference Proceedings Volume

Section

Articles

How to Cite

Verma, S. ., Goyal, P. ., & Sowmya, K. . (2026). AI-POWERED MINDSHIELD MODEL FOR ANALYZING TOXICITY AND THREATS IN ONLINE INTERACTIONS. DMPedia Lecture Notes in Computer Science & Engineering, IMPACT26, 210-222. https://digitalmanuscriptpedia.com/conferences/index.php/DMP-LNCSE/article/view/147