PREDICTING INDONESIAN ISP COMPLIANCE WITH NEGATIVE CONTENT DATA SYNCHRONIZATION: A SUPERVISED LEARNING APPROACH

Abstract

The Indonesian government, through the Ministry of Communication and Information Technology (Kominfo), mandates that all Internet Service Providers (ISPs) actively filter negative content, creating a significant regulatory and technical challenge. This study presents a data-driven approach to predict ISP compliance with these mandates using supervised machine learning. Leveraging a dataset of operational metrics from 606 Indonesian ISPs for April 2025, this research defines compliance based on an ISP's foundational capability to perform network-level blocking, proxied by the implementation of Remotely Triggered Black Hole (RTBH) filtering. Key performance indicators, including DNS Queries Per Second (QPS), a DNS trustworthiness score (TNG), and Transactions Per Second (TPS), were used as predictive features. Five machine learning algorithms Logistic Regression, Support Vector Machine (SVM), Decision Tree, Naive Bayes, and Random Forest were trained and evaluated. The results demonstrate high predictive accuracy, with the Decision Tree and Random Forest models achieving 97% accuracy. Feature importance analysis, using the Gini impurity method, revealed that QPS was the most significant predictor, serving as a latent indicator of network