In the years since its initiation, the software-defined network paradigm has evolved into a distinguished networking technology by causing a revolution in separating the control logic from physical devices and centralizing software-based controllers. Despite its indisputable benefits compared with the traditional network, the SDN raises the challenge of scalability with its physically centralized control. The only potential solution is to transform it into physically distributed SDN control. However, this solution requires the interoperability between SDN controllers, and the consistency of network state being distributed across these controllers. Although some east–west interfaces that help SDN controllers exchange network information have been released, they reveal several drawbacks. First, they cannot support a heterogeneous SDN system where SDN controllers are developed by different providers. Secondly, their consistency solution is simple in disregarding the tradeoff between the consistency level and the performance of SDN networks. This paper proposes an east–west interface, called SINA, to provide the interoperability of a heterogeneous and distributed SDN network. In addition, a novel reinforcement-learning-based consistency algorithm is introduced for an adaptive, quorum-based replication mechanism. The experimental results showed that SINA successfully connects heterogeneous and distributed SDN domains and balances the consistency and network performance.

Encrypted network traffic classification has become a critical task with the widespread adoption of protocols such as HTTPS and QUIC. Deep learning-based methods have proven to be effective in identifying traffic patterns, even within encrypted data streams. However, these methods face significant challenges when confronted with new applications that were not part of the original training set. To address this issue, knowledge transfer from existing models is often employed to accommodate novel applications. As the complexity of network traffic increases, particularly at higher protocol layers, the transferability of learned features diminishes due to domain discrepancies. Recent studies have explored Deep Adaptation Networks (DAN) as a solution, which extends deep convolutional neural networks to better adapt to target domains by mitigating these discrepancies. Despite its potential, the computational complexity of discrepancy metrics, such as Maximum Mean Discrepancy, limits DAN’s scalability, especially when applied to large datasets. In this paper, we propose a novel DAN architecture that incorporates Smooth Characteristic Functions (SCFs), specifically SCF-unNorm (Unnormalized SCF) and SCF-pInverse (Pseudo-inverse SCF). These functions are designed to enhance feature transferability in task-specific layers, effectively addressing the limitations posed by domain discrepancies and computational complexity. The proposed mechanism provides a means to efficiently handle situations with limited labeled data or entirely unlabeled data for new applications. The aim is to limit the target error by incorporating a domain discrepancy between the source and target distributions along with the source error. Two statistics classes, SCF-unNorm and SCF-pInverse, are used to minimize this domain discrepancy in traffic classification. The experimental results demonstrate that our proposed mechanism outperforms existing benchmarks in terms of accuracy, enabling real-time traffic classification in network systems. Specifically, we achieve up to 99% accuracy with an execution time of only three milliseconds in the considered scenarios.

Today, the Software-defined Network, with its advantages such as greater reliability via automation, more efficient network management, cost-savings, and faster scalability, is increasingly being deployed in many network systems and network operators. The most common deployment architecture is a distributed system with the existence of many independent domains, each controlled by an SDN controller. One of the well-known applications in SDN is server selection and routing. However, deploying server and route selection in distributed and heterogeneous SDN networks faces two issues. First, the lack of global views of the whole system is because the inter-communication between SDN domains has not been standardized for the distributed and heterogeneous SDN network. To solve this issue, we use our previous work, an open East-West interface called SINA, to adaptively guarantee the network state consistency of the distributed SDN network with multiple domains. Secondly, selecting the path for packet transmission based only on the current network states of a local SDN domain is ineffective as it can bring over-utilization to several links and under-utilization to others. Predicting the link cost of the whole path from the source to the destination is necessary. Therefore, this paper proposes an LSTM-based link cost prediction for the server and route selection mechanism in a distributed and heterogeneous SDN network. The experimental results show that our proposal improves up to 15% of link utilization, reduces 10% of packet loss, and obtains the lowest servers’response time compared to benchmarks

Traffic classification plays a crucial role in optimizing network management, enhancing security, and enabling intelligent resource allocation in distributed network systems. However, traditional Federated Learning (FL) approaches struggle with domain heterogeneity, as network traffic characteristics vary significantly across different domains due to diverse infrastructure, applications, and usage patterns. This results in degraded performance when applying a single global model across all domains. To overcome this challenge, we propose KD-PFL-TC, a Knowledge Distillation-based Personalized Federated Learning framework for Traffic Classification, aimed to balance global knowledge sharing with personalized model adaptation in heterogeneous network environments. Our approach leverages knowledge distillation to enable collaborative learning without directly sharing raw data, preserving privacy while mitigating the negative effects of domain shifts. Each domain refines its local model by integrating insights from a global model and peer domains while maintaining its unique traffic distribution. To further enhance performance, we introduce an adaptive distillation strategy that dynamically adjusts the influence of global, peer, and local knowledge based on the similarity between traffic distributions, ensuring optimal knowledge transfer designed to each domain’s characteristics. Extensive experiments on real-world traffic datasets show that KDPFLTC maintains 88.0% accuracy under high heterogeneity (vs. 75.0% for FedAvg) while reducing communication overhead by 60%, delivering an efficient and robust solution for large-scale, heterogeneous networks.

To enhance the efficiency and resource utilization of a computer network, it is imperative to classify network traffic and implement distinct priority policies. Network traffic classification plays a pivotal role across various domains, including network administration, cybersecurity, and network resource optimization. As encrypted network data undergoes diverse evolution, as evident in datasets from tech giants like Google, Facebook, and YouTube, traditional traffic classification methods have given way to machine learning-based approaches. Given that computer networks are primarily deployed as distributed multi-domain systems, employing machine learning for traffic classification becomes challenging when a new network domain appears with a limited dataset. One potential remedy is to employ transfer learning, allowing knowledge transfer from a pre-trained model in an established domain to a new one. In this paper, we present two contributions. First, a novel algorithm called Multi-class TrAdaBoost-CNN is introduced to tackle the challenge of cross-domain classification in encrypted network services. This algorithm extends the Multi-class TrAdaBoost approach by incorporating a Convolutional Neural Network (CNN) as a weak learner. Secondly, extensive experiments are conducted on two distinct domains characterized by imbalanced data distributions to assess the efficacy of our proposed method. The experimental results clearly demonstrate that our algorithm outperforms the traditional CNN model, achieving remarkable accuracy improvements of up to 16%, even when dealing with extremely limited data.

Nowadays, Software-Defined Networking (SDN) with a separation between control and data plane, is a potential solution to overcome network issues caused by the explosive growth of network equipment and services. Two services that have attracted considerable attention in the computer network research community are server selection and routing. However, this research is only limited to a single SDN network, which is not a popular inter-SDN domain network where each SDN domain is dedicated to an Internet Service Provider (ISP). This limitation is mainly due to the following two reasons: 1) First, each SDN domain does not have a global view of other SDN domains, so routing policies are not optimal for the entire network; 2) Second, without the global view, ISPs select the shortest path to forward packets in their networks, often bringing over-utilization to several links and under-utilization to other links. This leads to a degradation of network performance and Quality of Service (QoS). Therefore, in this paper, we propose a QoS-based server and route selection for inter-SDN domains to optimize network performance. To overcome the first limitation, we take advantage of our recent work, SINA (SDN Inter-Domains Network Application), and an adaptive consistency mechanism to guarantee information consistency between controllers of inter-SDN domains and intra-SDN domains. To deal with the second one, we present a QoS-based server and route selection mechanism to select server and routing paths with optimized cost in order to optimize network performance. The experiments show that our proposal provides a good trade-off between QoS metrics (improvements of 15 percent of link utilization 13 percent of loss) and greatly reduced overhead and response time, compared with benchmarks.