The human-level fluency achieved by large language models in text generation has intensified the challenge of distinguishing between human-written and AI-generated texts. While current fine-tuned detectors exist, they often lack robustness against adversarial attacks and struggle with out-of-distribution topics, limiting their practical applicability. This study introduces AIDER, a robust and topic-independent AI-generated text detection framework. AIDER leverages the ALBERT model for topic content disentanglement, enhancing transferability to unseen topics. It incorporates an augmentor that generates robust adversarial data for training, coupled with contrastive learning techniques to boost resilience. Comprehensive experiments demonstrate AIDER’s significant superiority over state-of-the-art methods, exhibiting exceptional robustness against adversarial attacks with minimal performance degradation. AIDER consistently achieves high accuracy in non-augmented scenarios and demonstrates remarkable generalizability to unseen topics. These attributes establish AIDER as a powerful and versatile tool for LLM-generated text detection across diverse real-world applications, addressing critical challenges in the evolving landscape of AI-generated content.

The advancement of the fifth generation (5G) mobile communication and Internet of Things (IoT) has facilitated the development of intelligent applications, but has also rendered these networks increasingly complex and vulnerable to various targeted attacks. Numerous anomaly detection (AD) models, particularly those using deep learning technologies, have been proposed to monitor and identify network anomalous events. However, the implementation of these models poses challenges for network operators due to lacking expert knowledge of these black-box systems. In this study, we present a comprehensive review of current AD models and methods in the field of communication networks. We categorize these models into four methodological groups based on their underlying principles and structures, with particular emphasis on the role of recent promising large language models (LLMs) in the field of AD. Additionally, we provide a detailed discussion of the models in the following four application areas: network traffic monitoring, networking system log analysis, cloud and edge service provisioning, and IoT security. Based on these application requirements, we examine the current challenges and offer insights into future research directions, including robustness, explainability, and the integration of LLMs for AD.

Programmable Logic Controllers (PLCs) are critical components in Industrial Control Systems (ICSs). Their potential exposure to external world makes them susceptible to cyber-attacks. Existing detection methods against controller logic attacks use either specification-based or learnt models. However, specification-based models require experts’ manual efforts or access to PLC’s source code, while machine learning-based models often fall short of providing explanation for their decisions. We designSRLR— aSymbolic Regression based Logic Recoverysolution to identify the logic of a PLC based only on its inputs and outputs. The recovered logic is used to generate explainable rules for detecting controller logic attacks. SRLR enhances the latest deep symbolic regression methods using the following ICS-specific properties: (1) some important ICS control logic is best represented in frequency domain rather than time domain; (2) an ICS controller can operate in multiple modes, each using different logic, where mode switches usually do not happen frequently; (3) a robust controller usually filters out outlier inputs as ICS sensor data can be noisy; and (4) with the above factors captured, the degree of complexity of the formulas is reduced, making effective search possible. Thanks to these enhancements,SRLRconsistently outperforms all existing methods in a variety of ICS settings that we evaluate. In terms of the recovery accuracy,SRLR’s gain can be as high as 39% in some challenging environment. We also evaluateSRLRon a distribution grid containing hundreds of voltage regulators, demonstrating its stability in handling large-scale, complex systems with varied configurations.

The reliability of routing algorithms is critical for network operations. Deep Reinforcement Learning (DRL)-based routing algorithms are emerging as promising solutions for achieving next-generation intelligent network routing. However, existing DRL-based routing algorithms struggle to handle topological changes such as link failures due to the restriction of traditional neural networks. In this paper, we propose a novel routing algorithm named PPO-TEGN for enhancing robustness against topological changes. PPO-TEGN leverages Proximal Policy Optimization (PPO) framework for training, and incorporates a meticulously designed Transformer-based Edge-Enhanced Graph Neural Network (TEGN) to extract graph information for probabilistic routing scenarios. Additionally, motivated by the analysis of traffic distribution, we propose a subgraph-enhancement mechanism to improve the ability of perceiving potential congestion at the source and destination nodes. We conduct experiments on simulated and real-world topologies in different traffic patterns. Experimental results reveal that PPO-TEGN outperforms baselines including the traditional routing algorithm and DRL-based routing algorithms with basic GNN in reducing end-to-end (E2E) delay in networks with link failures, implying the strong robustness of the proposed algorithm.

With the rapid advancement of communication networks, wireless traffic prediction plays a pivotal role in resource allocation and energy management. However, due to the complex spatio-temporal characteristics inherent in real-world traffic data, capturing its intrinsic features accurately has proven to be challenging, leading to unsatisfactory prediction accuracy. In this paper, we propose a Graph Convolution Network based State Space Model (GSSM) aimed at predicting call detail records wireless traffic. We utilize both fixed and adaptive weighted matrices to learn the complex spatial dependencies among base stations. The graph convolution network is conducted to accurately estimate the parameters of the state space model. Additionally, we employ a mixture Gaussian hypothesis to provide more flexibility for the state space model when predicting traffic data. Our experimental results demonstrate that the proposed method outperforms the baseline methods. Furthermore, we conduct extensive visualization, ablation and parameter analysis experiments to confirm the effectiveness of our proposed approach.

Fueled by the explosive growth of ultra-low-latency and real-time applications with specific computing and network performance requirements, the computing force network (CFN) has become a hot research subject. The primary CFN challenge is to leverage network resources and computing resources. Although recent advances in deep reinforcement learning (DRL) have brought significant improvement in network optimization, these methods still suffer from topology changes and fail to generalize for those topologies not seen in training. This paper proposes a graph neural network (GNN) based DRL framework to accommodate network traffic and computing resources jointly and efficiently. By taking advantage of the generalization capability in GNN, the proposed method can operate over variable topologies and obtain higher performance than the other DRL methods.

With the rapid development of Mobile Internet, social media platforms explosively grow and greatly facilitate people to obtain and exchange information. Since any users can post arbitrary information on social media, some people with ulterior motives try to create and spread various misinformation for their own benefits, which might damage citizen rights and disturb social order. In this article, we propose a novel rumor detection model based on Graph Convolution Networks (GCNs). Based on the real dataset from social media, we consider not only the static features such as users's basic information and text contents, but also the dynamic features such as rumor propagation relations. The GCN-based model is used to represent the spreading structure of rumors, with graph convolution operator for node vector updating. We also optimize the feature fusion module and pooling module to make our model have better performance. Experiments on Sina Weibo dataset validate the performance of the proposed GCN-based model for rumor detection.

In modern times, mobile online communication through Instant Messaging is a part of many people's daily lives, which in turn forms a very complex and huge social network on Internet. In this paper, based on real traffic flow data collected from operational networks of Mobile Internet Service Provider, we analyze the traffic of several popular Mobile Instant Messaging Service in China, i.e. WECHAT, QQ, FETION. By constructing flow graphs of IM services, we list Top-five busiest IM servers, and investigate the statistical characteristics of the IM applications and servers. The traffic data is divided into data samples by a second, we aggregate packets for each group and sum packets number in different time scale, and the Self-similarity in the IM services' traffic are proposed, then the existence of Self-similarity in the IM traffic is studied and compared. Our extensive researches of the mobile traffic show that the traffic of Instant Messaging services has the characteristics of self- similarity, especially WECHAT.

Internet structure possesses many properties of complex networks. However, existing studies are often constrained to deriving web connections based on partial data collected, and the actual Internet traffic and user behaviors are far from being understood. With detailed traffic flow records collected through powerful hardware-based monitors, we study from the perspective of complex network the characteristics of four types of traffic: P2Pdownload, HTTP, Instant Messaging and overall traffic. Based on the data analysis and comparison of different applications, we confirm that both the distributions of node degree and strength of nodes/edges follow power law but they have significant different exponents. Specifically, taking advantage of the strict timing of the records, we study the dynamics of flow graphs. The growth of edges upon nodes is nonlinear. Edges formed between existing nodes, instead of the ones arriving with new nodes dominate the growth. We also observe linear preferential attachment behaviors in the flow graphs.

IEEE 802.16 is a promising technology in broadband wireless access area. In multiservice IEEE 802.16 network, connection admission control (CAC) plays a critical role for resource management and QoS guarantee. In this paper, a statistical CAC mechanism is proposed for IEEE 802.16 network. In order to avoid the QoS degradation, the proposed CAC mechanism considers the traffic variability and overflow. Furthermore, a model of traffic and air interface capacity is provided to make the CAC mechanism easy to be implemented. Then, a performance analysis model based on Markov chains is proposed. Numerical and simulation results show that the proposed CAC mechanism can prevent the traffic overflow and achieve a good packet level QoS.