Recently, the energy consumption of WiFi in smart buildings has reached extreme levels. Here, SDN leads intelligent energy management by decoupling the control and physical layers without touching network equipment. DT also enables continuous monitoring and synchronized management across the topology by combining real and virtual data in the central controller. DT-enabled SDN is proposed to optimize timeliness while predicting access point states to minimize energy efficiency in time series analysis. There are many ML models for time-series problems in the literature, but they do not solve the trade-off between prediction performance and model training time. Therefore, we've proposed an Autogluon-based energy management framework to determine the best ML model in a short DT-based SDN period. To execute it, a novel energy efficiency metric is defined in an optimization formula. To solve this optimization with low cost, we proposes AP state prediction and QoS-aware energy management algorithms that determine AP state configurations and embed OpenFlow rules. According to the results, the proposed method outperforms the conventional approaches with an energy gain of 48.54%.

Today, the number of latency-sensitive tasks in Industry 4.0 is growing rapidly. With the emergence of smart factories, optimal task scheduling plays an important role in software-defined data centers (SDDC). There are many studies in the literature to solve this NP-hard problem, which do not consider the collective of data from the physical layer and optimal task scheduling in cloud servers. In addition, the intelligent transformation of the manufacturing industry by mapping the algorithm decision with real-world requirements has a major synchronization problem. In terms of CPU and memory utilization, the proposed solution should be easy to implement in real-life scenarios to improve the scalability of the server. The response time of task scheduling is also very important for latency-sensitive tasks in Industry 4.0. Therefore, we've proposed a novel Digital Twin (DT)-based SDDC platform that provides weighted task optimization. Thanks to virtual replicas of industrial objects in the physical layer, it optimizes task scheduling by monitoring and integrating virtualization and synchronization with the digital twin layer. A novel algorithm based on particle swarm optimization is proposed to maximize the profit of clustered tasks in different virtual machines of DT-based SDDC. The performance evaluation is compared with conventional dynamic programming and greedy approaches in terms of memory utilization, CPU utilization, response time, and profit maximization as the number of tasks increases. Our results show that the proposed approach achieves acceptable response time with higher profits.

Nowadays, new developments in artificial intelligence (AI) and the Internet of Things (IoT) have made it easier to continuously improve smart Industry 4.0 applications. These can be exemplified as smart manufacturing, environmental monitoring and management, industrial IoT (IIoT) platforms, augmented and virtual reality (AR/VR) for maintenance and training. In the literature, the digital twin (DT)-based network management is alleviated for these applications by replicating the physical equipment with an AI-based virtual object that periodically takes data from the real industrial world. Moreover, the accuracy of AIbased predictions in DT is extremely high due to the centralized view and aggregate huge amount of data to the cloud. However, centralized DT has main challenges such as data privacy concerns, the lack of trust in sharing sensitive data, and having a high potential for exposure to many types of attacks during data sharing between IoT devices and cloud. Therefore, we've proposed DT-based edge network (DTEN) to integrate DT at the edge to overcome the aforementioned challenges. Moreover, Federated Learning (FL) is a promising approach to address these concerns by ensuring privacy and trust in DTEN. Here, a novel framework that integrates FL with DT-enabled Industry 4.0 combines the advantages of both to keep accuracy of AI based applications in acceptable level, improve scalability and adaptability while maintaining privacy. Our practical analysis shows that the DTEN based on the federated approach not only protects privacy instead of sharing raw data, but also achieves acceptable accuracy results nearly as centralized AI-based models. It is believed that the proposed approach offers a promising way to revolutionize more use cases such as Industry 4.0, healthcare, smart cities.

In 5G and beyond, the end-user broadcasts data into the radio domain; therefore, protecting the physical layer security (PLS) is a more challenging issue. Here, Sparse Code Multiple Access (SCMA) handles PLS naturally. Because multiple physical resources are encoded in different codes, these codes cannot be easily decoded without sharing a mother codebook and mapping matrix. However, there are a few studies about security implementation in the physical layer against the challenges of the three subdomains of end users, base stations, and the core network. Moreover, there are two new research questions for resource allocation in SCMA: How do codebooks affect the Bit Error Rate (BER) performance of end-users? In the presence or absence of a jammer or interference, how should codebooks be assigned to users? We believe that it can only be overcome by using some new paradigms together: Software-defined radio (SDR), Software-defined networking (SDN), and network function virtualization (NFV). We consider the PLS implementation on SCMA-based 5G by investigating codebook assignment in the presence of noise/jammer with a novel Quality of Service (QoS) parameter$(\phi)$, which is the ratio of the received power by the user to overloading$(\lambda)$. In the SDN/NFV controller, the QoS metric-based codebook assignment algorithm prioritizes users into newly defined PEACE, NORMAL, and WAR states. In the OpenFlow table of SDR, grant-free access can be easily implemented. As a result, the BERs for different overloading are the same for the same ratio of energy rate per bit and spectral noise density. The BER is lower when the mother codebook is 150% than when the codebook is 200%. An existing codebook of one eMBB can be utilized to serve one URLLC user and two MTC users with an acceptable BER.

Due to increased mMTC devices in 5G, OFDMA has not been a solution for resource allocation in SDR. Therefore; SCMA which is one of the code domain NOMA has been using in such overloaded cases. As distinct from OFDMA, it uses also the constellation diagram as an addition to frequency and time domains in resource mapping to the mobile user equipment. The current studies do not consider UE prioritization while code allocation in SCMA is believed to enhance QoS of the whole topology. Moreover, due to the nature of SCMA, protecting the privacy of UE has become harder than conventional allocation strategies. The implementation details of secure SCMA does not consider by handling low OPEX/CAPEX. Therefore, in this paper, we introduce a novel SDN/NFV based SDR by handling these 5G paradigms into the same platform to overcome the aforementioned challenges. With SDN, we newly define a QoS metric; i.e. received power over overloading of SDR, for UE prioritization such as Micro and Macro UE in an SDR. With NFV, SDR can have two virtual roles shown to UEs by Macrocell and Microcell virtual network functions. With SDR, the SCMA code allocation is performed according to defined Macro and Micro codes are taken from generated codebooks. Afterward, it performs the asymmetric encryption algorithm to provide security of code allocations and the privacy of the user. According to performance evaluation; when overloading is 150%, the Micro UEs have 10x better BER than Macro UEs where the QoS is acceptable for the whole topology.

In Software Defined-Content Delivery Networks (SD-CDN), the policies of tenants such as Youtube, Netflix, Office 365, etc. are not the same due to having different 5G traffic requirements for such contents as enhanced Mobile Broadband (eMBB), ultra-reliable low-latency (URLLC) and massive machine-type communication (mMTC). This leads SD-CDN multi-tenant slicing to provide different services with limited network resources, where each tenant can functionally manage their own virtual slice of a physical component according to service level agreements (SLAs). However, they are not permitted to dynamically configure their own components. Therefore, the physical end-to-end configuration of all edge devices causes extra hardware and bandwidth costs. Although software as a service (SaaS) is more preferred to handle cost-efficiency on a switch configuration that increases forwarding throughput (Mbps) with a less number of physical components in SD-CDN, the edge devices can be only served as infrastructure as a service (IaaS) currently. This motivation leads us to serve the switch as a service that includes both IaaS and SaaS characteristics. Therefore, we propose an OpenFlow as a service (OFaaS) design where each tenant has flow management and switch configuration permissions on their own virtual slice. In flow management, we define a novel Service Oriented Architecture (SOA) to orchestrate OFaaS driven topology by isolating each tenant from physical complexity. Here, each tenant can dynamically alter QoS on a flow and load balance between a sub-set of contents via OpenFlow protocol. In switch configuration; a novel OFaaS Management Algorithm for a multi-tenant slicing increases the number of tenants served per OpenFlow switch thanks to OFaaS design. It enables an end-to-end configuration via the NETCONF protocol with a novel YANG model of OFaaS. According to performance evaluation; OFaaS has the same forwarding throughput as conventional IaaS based OpenFlow switch for a homogenous content, whereas it has 71% more forwarding throughput (Mbps) and it has 40% more cost-efficient than conventional one for a heterogeneous content with $17 savings per tenant.

The high bandwidth and moderate latency requirements of enhanced Mobile BroadBand (eMBB) content in 5G are handled by Content Delivery Network (CDN) components such as customer premises equipment (CPE) in edge, branches in fog, and servers in the cloud. The indispensable goal of 5G network providers is meeting these requirements on the same physical component that reduces operational and capital expenditures (OPEX/CAPEX). However, we've believed that different configuration policies of providers can be only handled by Software-Defined CDN (SD-CDN). Moreover; due to the flow characteristics of eMBB and handling packet forwarding by only software-based component in an OpenFlow (OF) switch, obtained packet processing delay in conventional CPE has become challenging. Therefore, we propose SD-vCPE with a novel physical block diagram which includes hardware and software-based OF tables. The analytical models of conventional and SD-vCPE are also derived by using G/G/1 Markov queues, considering general eMBB flow characteristics, and SpeedUp parameters in the software part. According to analytical and simulation results, as the resolution of video content in eMBB is increased, the conventional CPE cannot handle delay in the microsecond level; whereas, proposed SD-vCPE outputs approximately 1-millisecond decrease on packet delay which differentiates according to matching rules in TCAM or CPU. Namely, the proposed analytical model for SD-vCPE results in a delay in an acceptable rate while comparing the simulation results performed in MATLAB.

The tremendous increase on mobile data traffic has stressed conventional cellular networks recently. In order to handle this rapid increase, Software Defined Networking (SDN) is proposed as one of the novel approach that makes it easier to orchestrate physical devices in Data Plane with its centralized control fashion. On one hand, SDN provides scalability and flexibility on network management with dummy OpenFlow(OF) switches and its nature of centralized authority; on the other hand, these properties cause spatial and temporal complexity in the Data Plane. Spatial complexity, described as Memory Usage in an OF switch, should be minimized by removing redundancy on OF switch flow table pipeline. Temporal complexity, defined as Flow Forwarding Delay, should also be reduced by lowering number of comparison in OF switch pipeline to enhance Quality of Service (QoS) of a flow. Therefore, in this paper, a novel Multi-Stage OF (MsOF) Switch model is proposed and examined considering Queuing Theory in the light of spatial complexity and temporal complexity parameters. According to performance evaluation results, specially in urban areas, MsOF has much less spatial and temporal complexity considering conventional OF switch model as the number of input ports in an OF switch (N) increases. MsOF gains much memory space by deploying more tables with less memories, total of (n · k). Furthermore, with such a Multi-Stage deployment of flow tables, MsOF provides approximately 7 times less flow forwarding delay, compared with a conventional OF Switch for a network load more than %60.

The tremendous increase in the mobile data traffic has led the network operators to use offloading solutions in Radio Access Technologies (RATs) such as WiFi or smallcells. However, these solutions are not efficient since none of them controls the offloading from a centralized global network view. Current offloading methods are mostly user-based and they do not completely consider the user satisfaction, and more significantly they do not consider the Quality of Service (QoS) of the whole network system. These aforementioned motivations have led us to design a software based orchestration model for Heterogeneous Networks (HetNets) which is provided by our SDoff framework in a Software-Defined Network (SDN) fashion. This paper presents a novel SDN-based control of smallcell offloading in order to enhance the satisfaction of users and QoS of the overall system by considering a dissatisfaction parameter (ψ). The proposed framework analyzes the distance of each user to the all HetNet base stations in the topology by considering different type of users. After that, the offloading decision algorithm resolves which user must be offloaded to which base station by identifying the most dissatisfied user in the topology according to (ψ), by evaluating the QoS matrix and user types. Moreover, a new beacon format is proposed in order to maintain a standardized communication between the SDoff plane and virtual representatives (Avatars) of the Data Plane elements. We showed the overall system dissatisfaction difference between SDoff controlled offloading and on-the-spot offloading in the evaluation.

The deployment of Heterogeneous Networks (Het-Nets), specifically the smallcell networks ranging from picocells, microcells and femtocells, have been increased dramatically with the proliferation of next generation wireless devices and broad traffic demands. Embedding the promising Cognitive Radio (CR) technology into the smallcells have brought complementary increases in spectrum utilization as well as the topological flexibility. This paper deals with the topological control of such CR based smallcell networks by considering the spectrum utilization, packet losses, delay and jitter parameters. The proposed framework analytically models the spectral topology assignment taking into account the analysis of the CR user requests and decides to assign the proper spectrum to the CR smallcell users by dividing the topology in an optimal manner. This optimality is achieved by taking into account different Quality of Service (QoS) parameters like throughput, delay, jitter and packet losses. The proposed frameworks' mathematical model also provides optimal QoS parameter values for more efficient and fair topology assignment, considering the heterogeneous background CR traffic in the surrounding. The proposed method is evaluated for different types of CR traffics and thorough simulation results show that the proposed topology control mechanism provides an optimum number of smallcell CR networks to maintain fair and efficient QoS requirements in the unlicensed bands.