Graph Neural Networks (GNNs) have demonstrated remarkable success in various scenarios. However, their impressive performance is under the assumption of class balance (i.e., equal training sample distribution across various categories). Once trapped in the class-imbalanced issue, the GNN-based models typically under-represent the minority ones, resulting in decreased performance compared to balanced graphs. A promising solution is to balance the graph in a generative manner. However, the existing studies overlook the consistency between the synthesized sample and its corresponding class. Furthermore, the homophily assumption (i.e., like attracts like) undermines the topological diversity of graphs, thereby complicating the capability of models to capture the true distribution and boundaries of the categories. To this end, we propose aConsistency-Aware andLooseHomophily guided generative method for class-imbalanced graphs, namelyGraphCALH. Specifically, we design a consistency-aware feature synthesis method to balance the node- wise characteristics and the class- wise commonality for the synthesized samples. Moreover, we devise a loose homophily guided topology modeling method to enrich the topological diversity and simplify category boundaries. The experimental results on eleven class-imbalanced datasets demonstrate that the proposed GraphCALH outperforms ten state-of-the-art methods. The source code of this work will be uploaded to Github.

Jamming attacks present a significant security threat to wireless networks by exploiting the open wireless medium, causing not only a denial-of-service, but also overloading the network and interfering with signals. The jamming attack can enable more sophisticated disruptions, such as protocol-aware and learning-based jamming, where adversaries transmit selectively upon detecting legitimate network activity, and this selective transmission conserves attacker resources and complicates detection. Channel hopping is widely adopted as a mitigation strategy, allowing networks to move away from interfered channels dynamically. However, recent studies have demonstrated that intelligent jammers, such as the Multi-Armed Bandit (MAB)-based attack, can effectively learn and predict channel hopping patterns, thereby continuously jamming communications and severely degrading network performance. Designing robust countermeasures against such intelligent jamming remains an open and critical challenge. In this paper, we analyze a kind of MAB-based attack methodology and propose two refined variants employing online and offline paradigms. To counteract these advanced threats, we introduce three defense strategies: (i) speeding up, which modifies the frequency of channel-hopping decisions to avoid the jammer’s attack, (ii) deploying a helper node, diversify the communication patterns to affect the attacker’s learning process, and (iii) employing a mirror Multi-Armed Bandit (mirror MAB) approach to predict and bypass channels of highest jamming probability. Comprehensive evaluations demonstrate the effectiveness of our proposed defenses, significantly mitigating MAB-based attacks. Specifically, our strategies achieve a PDR exceeding 90% under persistent attack conditions, while effectively preserving robust and dynamic channel hopping behaviors.