Collaboration opportunities for devices are facilitated with Federated Learning (FL). Edge computing facilitates aggregation at edge and reduces latency. To deal with model poisoning attacks, model-based outlier detection mechanisms may not operate efficiently with hetereogenous models or in recognition of complex attacks. This paper fosters the defense line against model poisoning attack by exploiting device-level traffic analysis to anticipate the reliability of participants. FL is empowered with a topology mutation strategy, as a Moving Target Defence (MTD) strategy to dynamically change the participants in learning. Based on the adoption of recurrent neural networks for time-series analysis of traffic and a 6G wireless model, optimization framework for MTD strategy is given. A deep reinforcement mechanism is provided to optimize topology mutation in adaption with the anticipated Byzantine status of devices and the communication channel capabilities at devices. For a DDoS attack detection application and under Botnet attack at devices level, results illustrate acceptable malicious models exclusion and improvement in recognition time and accuracy.

The explosive growth of user devices and emerging applications is driving unprecedented traffic demands, accompanied by stringent Quality of Service (QoS) requirements. Addressing these challenges necessitates innovative service orchestration methods capable of seamless integration across the edge-cloud continuum. Terrestrial network-based service orchestration methods struggle to deliver timely responses to growing traffic demands or support users with poor or lack of access to terrestrial infrastructure. Exploiting both aerial and terrestrial resources in service composition increases coverage and facilitates the use of full computing and communication potentials. This paper proposes a service placement and composition mechanism for integrated aerial-terrestrial networks over the edge-cloud continuum while considering the dynamic nature of the network. The service function placement and service orchestration are modeled in an optimization framework. Considering the dynamicity, the Aerial Base Station (ABS) trajectory might not be deterministic, and their mobility pattern might not be known as assumed knowledge. Also, service requests can traverse through access nodes due to users' mobility. By incorporating predictive algorithms, including Deep Reinforcement Learning (DRL) approaches, the proposed method predicts ABS locations and service requests. Subsequently, a heuristic isomorphic graph matching approach is proposed to enable efficient, latency-aware service orchestration. Simulation results demonstrate the efficiency of the proposed prediction and service composition schemes in terms of accuracy, cost optimization, scalability, and responsiveness, ensuring timely and reliable service delivery under diverse network conditions.

Federated Learning (FL) enables training of a global model from distributed data. However, the singular-model based operation of FL is open with uploading poisoned models compatible with the global model structure and can be exploited as a vulnerability to conduct model poisoning attacks. This paper proposes a multi-model based FL as a proactive mechanism to enhance the opportunity of model poisoning attack mitigation. A master model is trained by a set of slave models. To enhance the opportunity of attack mitigation, the structure of client models dynamically change and the supporter FL protocol is provided. For a MEC system, the model selection problem is modeled as an optimization to minimize loss and recognition time, while meeting a robustness confidence. A deep reinforcement learning based model selection is proposed. For a DDoS attack detection scenario, results illustrate a competitive accuracy gain under poisoning attack with the scenario that the system is without attack, and also a potential of recognition time improvement.

Network function virtualization leverages programmable data plane switches to deploy in-network implementable functions, to improve QoS. The memories of switches can be extended through remote direct memory access to access external memories. This paper exploits the switches external memories to place VNFs at time intervals with ultra-low latency and high bandwidth demands. The reconfiguration decision is modeled as an optimization to minimize the deployment and reconfiguration cost, while meeting the SFCs deadlines. A DRL based method is proposed to reconfigure service chains adoptable with dynamic network and traffic characteristics. To deal with slow convergence due to the complexity of deployment scenarios, static and dynamic filters are used in policy networks construction to diminish unfeasible placement exploration. Results illustrate improvement in convergence, acceptance ratio and cost.

Offering intelligent services with ultra low latency and high reliability is one of the main objectives of 6G networks. Federated Learning (FL) is a solution to enhance the security of data and the accuracy, in comparison with local training of data in devices. The transmission cost in conventional FL is high. Performing FL using edge infrastructure is a solution. However, edge servers might not be available at every location or the communication with edge resources may prolong the learning process. This paper proposes a collaborative federated learning approach to provide intelligent services through collaboration of various learning levels including central cloud level, edge cloud level, and device level. Computational capabilities of neighbourhood devices are exploited to provide a fast recognition via 6G D2D communication. The learning is modeled as an optimization that performs trade-off between recognition accuracy and response time of recognition for devices. Considering the dynamicity in communication and computation status of the network/devices, a deep reinforcement learning method is proposed to decide about the collaboration of learning levels, and performing the appropriate trade-off. For a DDoS attack detection scenario, the evaluation results show improvement in the gained rewards, the attack detection accuracy, the response time of recognition, and the accumulation of accuracy and response time.

Multi-Access Edge Computing (MEC) is considered as one of the key enablers of Unmanned Aerial Vehicles (UAVs) use cases. However, the envisioned MEC deployments introduce new challenges related to the management of the mobility of services across the distributed MEC hosts, following the UAVs movements and possible handovers to ensure sustainable Quality-of-Service (QoS). A major challenge for MEC service mobility is the decision-making on where and when to relocate services. In this paper, we motivate the use of the predefined flight plans of UAVs for devising proactive relocation strategies that can deal efficiently with realistic asynchronous relocation processes. Moreover, we formulate the Proactive Service Relocation for UAV (PSRU) problem using linear programming, and we validate the gains introduced by the proactive relocation strategy and the use of the predefined flight plans of UAVs.

With the explosive demands for data, content delivery networks are facing ever-increasing challenges to meet end-users' quality-of-experience requirements, especially in terms of delay. Content can be migrated from surrogate servers to local caches closer to end-users to address delay challenges. Unfortunately, these local caches have limited capacities, and when they are fully occupied, it may sometimes be necessary to remove their lower-priority content to accommodate higher-priority content. At other times, it may be necessary to return previously removed content to local caches. Downloading this content from surrogate servers is costly from the perspective of network usage, and potentially detrimental to the end-user QoE in terms of delay. In this paper, we consider an edge content delivery network with vehicular nodes and propose a content migration strategy in which local caches offload their contents to neighboring edge caches whenever feasible, instead of removing their contents when they are fully occupied. This process ensures that more contents remain in the vicinity of end-users. However, selecting which contents to migrate and to which neighboring cache to migrate is a complicated problem. This paper proposes a deep reinforcement learning approach to minimize the cost. Our simulation scenarios realized up to a 70% reduction of content access delay cost compared to conventional strategies with and without content migration.

Slices were introduced in 5G to enable the co-existence of applications with different requirements on a single infrastructure. Slices may be delay-constrained for mission-critical applications such as Tactile Internet applications. When delay-constrained slices are implemented as collections of virtual network function (VNF) chains, a key challenge is to place the VNFs and route the traffic through the chains to meet a strict delay constraint. Parallel VNF processing has been proposed as a promising approach. However, this approach increases the number of physical nodes in the chains, and thus decreases the reliability, which is also critical for Tactile Internet applications. Furthermore, the cost depends upon the specific VNF placement and traffic routing, as nodes and links are heterogeneous. This article tackles the issues of reliability and cost when embedding delay-constrained slices. We model the problem as an optimization problem that minimizes reliability degradation and cost while ensuring the strict delay constraint when a parallel VNF processing approach is used. Due to the complexity of the formulated problem, we also propose a Tabu search-based algorithm to find sub-optimal solutions. The results indicate that our proposed algorithm can significantly improve cost and reliability while meeting a strict delay constraint.

Fog-based Content Delivery Networks (CDNs) distribute contents from origin servers to cloud replica servers and to fog caches. Some of these fog caches may be located on Set-top Boxes (STBs) close to end users, assuming that the STBs can host micro-caches. This can greatly reduce Latency. Network function virtualization (NFV) is a technology that can be used in fog systems. NFV-enabled STBs can therefore host micro-caches implemented as virtual network functions (VNFs). Appropriate placement mechanisms are needed to place these micro-caches to improve end-users' QoS in terms of Latency while still minimizing cost. In this paper, this placement problem is modeled as an optimization problem. A bee colony-based algorithm is suggested to find the placement that minimizes an aggregation of Latency and micro-cache storage cost. The simulation results show an improvement in the aggregated Latency and cost when the proposed algorithm is deployed.

The fifth generation (5G) utilizes Network Functions Virtualization (NFV) and Software Defined Network (SDN) to provide applications. Among them, there are deadline constrained applications. When the traffic generated by applications traverses Virtual Network Functions (VNFs) chains, it may not be possible to meet the deadlines when the traffic is processed sequentially; even if very fast servers are provisioned to host the VNFs. We propose a parallel VNF processing approach for the traffic. This is done through pools of candidate servers that host VNFs which process the traffic. Considering the routing policy provided by the SDN routing application, the server selection problem with the aim of deadline satisfaction and cost minimization is modeled as a non-linear binary optimization problem. An iterative search algorithm is proposed to solve this problem. Simulations show enhancement in deadline satisfaction as well as cost reduction.

Multi-stream transport protocols, such as SCTP and QUIC, tackle challenges such as Head of Line (HoL) faced by TCP. They are getting more and more deployed and currently carry more than 7% of the global Internet traffic. However, since these protocols have limited control over the actual routes the streams will take in the network, the streams are generally forwarded through a single and same path, even though there are generally multiple paths available in the network. Consequently, the traffic streams do not fully benefit from the available bandwidth and performance improvement remains limited. Software Defined Networks (SDN) separate control planes and data planes. It offers a complete view of the network to applications and enables network programmability through flexible rules. These rules may be used to ensure that the different streams generated by multi-stream transport protocols follow multiple paths in the network. In this paper, we propose an SDN-based framework for multi-stream transport protocols in multipath networks. The proposed framework provides an interface for applications to specify multi-stream rules. Based on these rules, the framework uses the services offered by the SDN Controller to ensure that the multiple streams go over multiple paths in the network. Experiments performed show that our proposal improves the QoS offered to the end users.

Fog computing extends the cloud to the edge of the network, close to the end-users enabling the deployment of some application component in the fog while others in the cloud. Network Functions Virtualization (NFV) decouples the network functions from the underlying hardware. In NFV settings, application components can be implemented as sets of Virtual Network Functions (VNFs) chained in specific order representing VNF-Forwarding Graphs (VNF-FG). Many studies have been carried out to map the VNF-FGs to cloud systems. However, in hybrid cloud/fog systems, an additional challenge arises. The mobility of fog nodes may cause high latency as the distance between the end-users and the nodes hosting the components increases. This may not be tolerable for some applications. In such cases, a prominent solution is to migrate application components to a closer fog node. This paper focuses on application component migration in NFV-based hybrid cloud/fog systems. The objective is to minimize the aggregated makespan of the applications. The problem is modeled mathematically, and a heuristic is proposed to find the sub-optimal solution in an acceptable time. The heuristic aims at finding the optimal fog node in each time-slot considering a pre-knowledge of the mobility models of the fog nodes. The experiment's results show that our proposed solution improves the makespan and the number of migrations compared to random migration and No-migration.

Fog computing reduces the latency induced by distant clouds by enabling the deployment of some application components at the edge of the network, on fog nodes, while keeping others in the cloud. Application components can be implemented as Virtual Network Functions (VNFs) and their execution sequences can be modeled by a combination of sub-structures like sequence, parallel, selection, and loops. Efficient placement algorithms are required to map the application components onto the infrastructure nodes. Current solutions do not consider the mobility of fog nodes, a phenomenon which may happen in real systems. In this paper, we use the random waypoint mobility model for fog nodes to calculate the expected makespan and application execution cost. We then model the problem as an Integer Linear Programming (ILP) formulation which minimizes an aggregated weighted function of the makespan and cost. We propose a Tabu Search-based Component Placement (TSCP) algorithm to find sub-optimal placements. The results show that the proposed algorithm improves the makespan and the application execution cost.

Value-added Services (VASs) (e.g. dynamic site acceleration, media management) play a critical role in Content Delivery Networks (CDNs). Network Functions Virtualization (NFV) enables the agile provisioning of VASs. In NFV settings, VASs are provisioned as ordered sets of Virtual Network Functions (VNFs), forming VNF-Forwarding Graphs (VNF-FG) which are deployed in the CDN infrastructure. The CDN VAS VNF-FGs have a specific characteristic: they have one end-point (corresponding to the content server) that is unknown, prior to their placement. The proposals for CDN VAS VNF-FG placement, so far, have only considered offline placement, where the VNF-FGs are placed before end-user traffic steers into the network. However, in concrete cases, a change in service usage patterns might occur, a situation that could require a VNF-FG placement in an online manner. This paper tackles the problem of online VNF-FG placement for VASs in CDNs, taking into account the eventual reuses and migrations of already-deployed VNFs. A cost model is considered, including multiple costs; i.e. new VNF instantiations, migration, hosting and routing costs. The objective is to optimally place the VNF-FGs such that total reconfiguration costs are minimized while QoS is satisfied. An Integer Linear Programming (ILP) formulation is provided and evaluated in a small-scale scenario.

Applications are sets of interacting components that can be executed in sequence, in parallel, or by using more complex constructs such as selections and loops. They can, therefore, be modeled as structured graphs with sub-structures consisting of these constructs. Fog computing can reduce the latency induced by distant clouds by enabling the deployment of some components at the edge of the network (i.e., closer to end-devices) while keeping others in the cloud. Network Functions Virtualization (NFV) decouples software from hardware and enables an agile deployment of network services and applications as Virtual Network Functions (VNFs). In NFV settings, efficient placement algorithms are required to map the structured graphs representing the VNF Forwarding Graphs (VNF-FGs) onto the infrastructure of the hybrid cloud/fog system. Only deterministic graphs with sequence and parallel sub-structures have been considered thus to date. However, several real-life applications do require non-deterministic graphs with sub-structures as selections and loops. This paper focuses on application component placement in NFV-based hybrid cloud/fog systems, with the assumption that the graph representing the application is non-deterministic. The objective is to minimize an aggregated weighted function of makespan and cost. The problem is modeled as an Integer Linear Programming (ILP) and evaluated over small-scale scenarios using the CPLEX optimization tool.

Large scale distributed systems employ thousands of resources which inevitably suffer from the unavailability issue. Serious side effects like unexpected delay or failure in the application execution are probable in case of such an issue. The imposed outcome might then be catastrophic consequences for real time applications or penalties for the service providers. Better prediction of the resource unavailability helps diminishing the undesired outcomes. This paper proposes a resource availability prediction algorithm for the mentioned goal. The resource availability variation is modeled as a stochastic process. By analyzing the availability information of NDU resources and both physical and virtual machines of the PlantLab, we found that the transition probabilities among the availability levels are non-stationary. To cope with this characteristic, we introduce Availability Transition Patterns (ATPs); the ATPs are dynamically constructed and the transitions between them are modeled by a Markov chain. The future ATP is then predicted based on the constructed Markov chain, according to which the resource availability-level is predicted. Experimental results confirm the efficiency of the proposed prediction algorithm.