A DEEP REINFORCEMENT LEARNING APPROACH FOR PROACTIVE ATTACK MITIGATION AND EFFICIENT FORWARDING IN NAMED DATA NETWORKING
DOI: https://doi.org/10.65725/JCISE/1/2/004
JOURNAL OF COMPUTATIONAL INTELLIGENCE SCIENCE AND ENGINEERING (JCISE)
ISSN: 3107-8168
Volume 1 Issue 2, October 2025

Abstract

Named Data Networking (NDN) presents a paradigm shift from host-centric to data-centric internet architecture, inherently offering advantages like built-in caching and multicast delivery. However, its core mechanisms, the Interest and Data packets, introduce new vulnerabilities, most notably Interest Flooding Attacks (IFA) and Content Poisoning Attacks (CPA), which can cripple network performance and integrity. Simultaneously, the efficient management of the Pending Interest Table (PIT) is critical for scalability. This paper proposes a novel, integrated framework that leverages Machine Learning (ML) to holistically enhance NDN’s security and scalability. We introduce a dual-ML model architecture: a Deep Reinforcement Learning (DRL) agent for dynamic, state-aware forwarding strategy to mitigate IFA and optimize PIT utilization, and a supervised Deep Learning model based on a Convolutional Neural Network (CNN) for real-time detection of CPA by analyzing content checksum patterns and request anomalies. Simulation results on the ndnSIM platform demonstrate that our proposed framework achieves a 98.5% detection rate for CPA and reduces the impact of IFA by over 95%, while also improving effective data retrieval rates by 30% under attack conditions, showcasing a robust and scalable enhancement to future internet architecture.

Authors: A.Anbarasi, A.Gobinath, and A.Afrin Safna

Keywords: Named Data Networking; Deep Reinforcement Learning; Interest Flooding Attack; Content Poisoning Attack; Pending Interest Table; Future Internet Architecture; Network Security