Abstract
This article provides a comprehensive overview of the "Computer Networks and IoT" section of the First International Artificial Intelligence Conference (IAIC 2023). It examines the integration of Artificial Intelligence (AI) into networking infrastructures, focusing on the emergence of intelligent edge computing, enhanced security protocols, and domain-specific hardware optimizations. The synthesis highlights a significant shift toward decentralized intelligence, where AI is utilized for real-time resource management and the protection of complex, abstract environments.
Introduction
The rapid proliferation of Internet of Things (IoT) devices and the escalating complexity of global communication networks have surpassed the capabilities of traditional manual management and static, rule-based security systems. The First International Artificial Intelligence Conference (IAIC 2023) addressed these challenges by showcasing research at the intersection of AI, computer networks, and IoT. This volume specifically highlights the transition toward "intelligent edges," where AI is no longer just a tool for back-end data analysis but a fundamental component in optimizing network security, task scheduling, and hardware efficiency.
Addressing the Complexity of Modern Networks
Researchers in this volume tackle three primary challenges inherent in modern networking:
- Decentralized Complexity: Managing resources across the "Cloud-Edge-End" spectrum is a critical hurdle. The goal is to achieve seamless resource allocation without incurring prohibitive latency or operational overhead.
- Security and Trust: As environments become more abstract—such as the Metaverse or passive IoT networks—traditional firewalls prove insufficient. New research focuses on protecting servers and users through more dynamic, AI-driven security frameworks.
- Domain-Specific Optimization: There is a concerted effort to move AI from controlled laboratory settings into niche, real-world applications. These include monitoring physiological data during exercise, improving sports analytics (e.g., basketball shooting accuracy), and forecasting pharmaceutical sales.
Methodological Frameworks
The methodologies presented represent a diverse "toolbox" of modern engineering techniques:
- Reinforcement Learning and Optimization: Techniques such as Proximal Policy Optimization (PPO) and the NSGA-II genetic algorithm are employed for dynamic task deployment and channel allocation in RFID systems.
- Computer Vision at the Edge: Real-time sports analytics and monocular visual odometry for navigation leverage models like YOLOv8 and optical flow matching.
- Privacy-Preserving Architectures: To aggregate data streams without compromising individual user identities, researchers explore techniques like Local Differential Privacy.
- Hardware-Level AI: Innovation extends to the physical layer, with work on high-power amplifiers for the BDS-3 (Beidou) terminal using InGaP/GaAs HBT technology.
Key Insights: Intelligence at the Edge and Trust Management
A central theme of the proceedings is that the "Edge" is becoming significantly smarter. Research into task offloading and dynamic deployment suggests that networks are evolving into intelligent, self-organizing entities. Furthermore, the conceptualization of "Trust Management" for RFID applications in the cloud indicates a shift in security philosophy: security is increasingly treated as a dynamic, measurable metric rather than a binary state.
The reach of AI is also expanding into the social and public health domains. Studies within this volume investigate the impact of smartphones on the health of rural older adults, the dynamics of rumor-spreading in social networks, and the use of clustering algorithms to map autism research trends within China.
Results and Future Outlook
The IAIC 2023 conference maintained a competitive selection process, accepting 86 papers across its various tracks. Key results include:
- Efficiency Gains: New methods for Deep Neural Network (DNN) inference partitioning using "Graph-Cut" techniques optimize task distribution between devices.
- Hardware Precision: The development of broadband power amplifiers based on "Normal Distribution Real Frequency Matching" has pushed the boundaries of communication hardware.
Despite these advancements, several limitations remain. Many edge solutions have only been tested in simulated environments, leaving their performance in massive, congested global networks as a subject for future study. Additionally, the inherent tension between data privacy (e.g., via Differential Privacy) and the high accuracy required for AI forecasting continues to be a primary challenge. Finally, optimizing sophisticated models like YOLOv8 for low-power, passive IoT devices remains a critical area for ongoing research.
Conclusion
The research presented at IAIC 2023 underscores a transformative period for computer networks and IoT. By embedding AI directly into the network fabric and hardware, researchers are paving the way for more resilient, efficient, and intelligent global infrastructures.
This summary is based on the proceedings of the First International Artificial Intelligence Conference (IAIC 2023), Part III.
Summary updated