Lightweight Neural Anomaly Detection for Resource-Constrained Edge-ICN Environments: A Systematic Literature Review

Edge computing and Information-Centric Networks (ICN) are converging to support low-latency, scalable, and content-aware network infrastructures. This shift introduces new challenges for real-time anomaly detection on resource-constrained edge devices. While prior studies have addressed ICN security and lightweight detection models in Internet of Things (IoT) settings, few have systematically explored lightweight neural models for anomaly detection at the edge. This paper conducts a Systematic Literature Review (SLR) that examines general edge-based approaches and extracts transferable insights for ICN anomaly contexts. We answer four key research questions: (1) the types of lightweight neural models used for anomaly detection on edge devices, (2) how these models are adapted for edge resource constraints, (3) the strategies used for integrating detection within ICN architectures, and (4) the main research challenges and future directions in this space. We analyzed 71 peer-reviewed studies (2017–2025) to identify trends in model architectural, optimization techniques, deployment strategies, and performance evaluation. Our synthesis highlights the predominant use of CNN-based and hybrid models optimized via pruning, quantization, and knowledge distillation. Special focus is given to content-centric anomalies and caching-related threats unique to ICN. We identify several research gaps, including scarcity of ICN-specific datasets, underutilization of explainable AI techniques, and limited focus on real-time and energy-efficient deployments. This review provides a structured foundation that may inform future research on secure, intelligent, and efficient anomaly detection systems in next-generation ICN-edge infrastructures.