MTikGuard System: A Transformer-Based Multimodal System for Child-Safe Content Moderation on TikTok

TL;DR

MTikGuard addresses the challenge of detecting child-harmful content on fast-moving TikTok streams by extending the TikHarm dataset to 4,723 labeled videos and employing a transformer-based multimodal architecture that fuses visual frames, audio transcripts, and scene text. The system uses TimeSFormer for video and multilingual text models (Multilingual DistilBERT and ViSoBERT), with late or attention-based fusion, and is deployed in a scalable streaming pipeline built on Apache Kafka, Spark, Docker, and Airflow. It achieves state-of-the-art performance with an overall accuracy of 89.37% and a macro-F1 of 89.45% on the test set, with robust generalization when evaluated on the Final Combined Dataset. The approach demonstrates practical impact for real-time, multilingual content moderation on large-scale platforms and lays groundwork for extending to other short-video ecosystems through modular deployment and multimodal fusion enhancements.

Abstract

With the rapid rise of short-form videos, TikTok has become one of the most influential platforms among children and teenagers, but also a source of harmful content that can affect their perception and behavior. Such content, often subtle or deceptive, challenges traditional moderation methods due to the massive volume and real-time nature of uploads. This paper presents MTikGuard, a real-time multimodal harmful content detection system for TikTok, with three key contributions: (1) an extended TikHarm dataset expanded to 4,723 labeled videos by adding diverse real-world samples, (2) a multimodal classification framework integrating visual, audio, and textual features to achieve state-of-the-art performance with 89.37% accuracy and 89.45% F1-score, and (3) a scalable streaming architecture built on Apache Kafka and Apache Spark for real-time deployment. The results demonstrate the effectiveness of combining dataset expansion, advanced multimodal fusion, and robust deployment for practical large-scale social media content moderation. The dataset is available at https://github.com/ntdat-8324/MTikGuard-System.git.

Figures (9)

• Figure 1: The dataset extending process.
• Figure 2: Proposed multimodal harmful content detection architecture.
• Figure 3: Confusion matrix of the best-performing model on the test set.
• Figure 4: Overview of the MTikGuard system.
• Figure 5: An example of the Harmful Content Detection task.