Abstract

     The ongoing expansion of the Industrial Internet of Things (IIoT) is enabling the possibility of effective Industry 4.0, where massive sensing devices in heterogeneous environments are connected through dedicated communication protocols. This brings forth new methods and models to fuse the information yielded by the various industrial plant elements and generates emerging security challenges that we have to face, providing ad-hoc functions for scheduling and guaranteeing the network operations. Recently, the large development of Software-Defined Networking (SDN) and Artificial Intelligence (AI) technologies have made feasible the design and control of scalable and secure IIoT networks. This paper studies how AI and SDN technologies combined can be leveraged towards improving the security and functionality of these IIoT networks. After surveying the state-of-the-art research efforts in the subject, the paper introduces a candidate architecture for AI-enabled Software-Defined IIoT Network (AI-SDIN) that divides the traditional industrial networks into three functional layers. And with this aim in mind, key technologies (Blockchain-based Data Sharing, Intelligent Wireless Data Sensing, Edge Intelligence, Time-Sensitive Networks, Integrating SDN&TSN, Distributed AI) and improve applications based on AI-SDIN are also discussed. Further, the paper also highlights new opportunities and potential research challenges in control and automation of IIoT networks.

Introduction

     Currently, we are witnessing an industrial revolution ranging term as Industry 4.0. Although this term may not be fully correct and involves a certain level of hype, the generalized implementation of autonomous decision making systems has become a new industry per se. The industry 4.0 is built by integrating networking capabilities that interconnect Artificial Intelligence (AI) agents in industrial plants, leading to the paradigms of Industrial Internet of Things (IIoT) and smart manufacturing [1,2]. These subjects are attracting continuous attention from both academia and industry, and have been categorized in various prototype models [3,4].

     Fig. 1 displays a typical architecture for IIoT networks, integrating four functional layers: data sensing, data transfer, data processing, and application layers [5,6]. The data sensing layer utilizes information acquisition and sensing technologies to collect heterogeneous data with different industrial attributes, and uploads them to the data transfer layer through advanced communication technologies, e.g., WiFi, Bluetooth, Internet of Things (IoT), 5G, etc. [7,8]. As the intermediated layer, the data transfer layer (based on the Internet backbone) aims at uploading the industrial data to the data processing layer by deploying Traffic Engineering (TE) or data routing policies, according to the given requirements and constraints (e.g., Quality of Service (QoS), costs, lags, etc.) of the industrial data delivery services [9]. Based on the dedicated data computing schemes (e.g., cloud/edge computing, distributed computing), the data processing layer adopts data fusion algorithms to process the data in real-time, and it provides an open interface for the application layer, such that the diverse categories of industrial applications can be edited and deployed at the application layer according to the current needs [10].

Conclusion

     Although the paradigm of IIoT networks promotes the progress of smart manufacture towards achieving reduced costs and increased efficiency, IIoT networks present the constraints of network heterogeneity and scalability. Current network architectures integrate computing units of low-intelligence, which cannot support the scheduling for diverse and complex industrial flow – especially for case scenarios with delay-sensitivity requirements. In this paper, we argued that these issues can be addressed by utilizing an AI-enabled SDN technique, and we have presented a formal architecture named by AI-SDIN to achieve this vision. Within the presented architecture, the IIoT networks are divided into three functional layers: data layer, network layer, and control layer. In the latter, advanced machine intelligence can be deployed to improve the security and functionality of the network. Then, we have surveyed possible key technologies towards actual implementation – e.g., Blockchain-based data sharing, intelligent wireless data sensing, edge intelligence, TSN&SDN, distributed AI, etc. We have also presented potential enhanced industrial applications based on AI-SDIN ranging from logistics to agriculture. Finally, the paper has indicated challenges and areas where further research is required. Our paper aims at revealing that the AI-SDIN leads the future of smart manufacture to be clear and bright. With AI-SDIN, industrial manufacture will be supermatic with less manufacture time and intelligent logistics, supporting fast production and marketing.

Abstract

1. Introduction

2. Related work

3. Methodology

4. Evaluation

5. Conclusion

Declaration of competing interest

Acknowledgments

References

Abstract

     Dynamic traffic in a software-defined network (SDN) causes explosive data to flow from one system to another. The explosive data affects the functionality of system parameters, network-level configuration, routing parameters, network characteristics, and system load factors. Adapting to the traffic flow is a key research area in SDN in today’s big data world. Load balance vehicular sensor accessibility reduces delays, lowers energy consumption, and decreases the execution time. This paper combines the entropy-based active learning model to identify intrusion patterns efficiently, which is a packet-level intrusion detection model. The developed afterload balancing model can track the attack on the network. We then proposed a load balancing algorithm that optimizes the vehicular sensor usability by using sensor computing capability and source needs. We make use of a convergence-based mechanism to achieve high resource utilization. We then perform experiments on the state-of-the-art intrusion detection dataset. Our experimental results show that the load balancing mechanism can achieve  in performance improvements compared to traditional approaches. Thus, we can see that the designed model can help improve the decision boundary by increasing the training instance through pooling strategy and entropy uncertainty measure.

Introduction

     The application of the Internet of things (IoT) and distributed computing has enabled a massive growth of heterogeneous applications. The future of IoT will connect many heterogeneous devices with the ability to communicate with the network directly [1]. Billions of objects (i.e., sensors network) will be connected to the Internet in the next generation of networks. This will result in extensive amounts of data that give rise to data delivery issues. Objects may include home application, traffic flow analysis, irrigation systems—these objects are usually equipped with several sensors or nodes. The role of these sensors/nodes is to gather and analyze real-time environments.

     Connected sensors with autonomous vehicles will also grow and become the future of intelligent transportation systems. Travel comfort, road security and safety will depend upon high data rates and reliable connectivity among autonomous vehicles. Such a transformation will increase the need for a safe and convenient network environment from transportation and transport infrastructure. These sensors are designed to acquire real-time data, and efficient processing is required for better performance. The gathered data is then being used by cloud computing-based applications [2]. The applications can store, process, and update the data in real-time. The applications have centralized data centers that are distributed geographically. Fog computing-based application services are handled at the network’s edge.

Conclusion

     This study presented a novel load balancing algorithm to balance the load of SDN among different vehicular sensors by using network and application information. The proposed model can balance the batch of applications among the vehicular sensor connected over the SDN. We tested our approach on different datasets, and the outcome of the proposed model has been compared with well-known heuristic-based models. Moreover, we used an entropy-based active learning approach to classify intrusion attacks. The developed model can achieve high accuracy to identify the patterns in terms of sparse and dense datasets. The entropy-based active learning-based method significantly increases training instances for the deep feedforward model. In the future, the network will be optimized tuned to apply the active learning mechanism. A weighted-based method for each class sub-sample selection can also be considered as a further extension.

1- Introduction

2- Related work

3- Flow-aware multi-topology adaptive routing

4- Implementation environment

5- Router components

6- Tests of the FAMTAR router

7- Conclusion

References

Abstract

Flow-Aware Multi-Topology Adaptive Routing (FAMTAR) is a new approach to multipath and adaptive routing in IP networks which enables automatic use of alternative paths when the primary one becomes congested. It provides more efficient network resource utilization and higher quality of transmission compared to standard IP routing. However, thus far it has only been evaluated through simulations. In this paper we share our experiences from building a real-time FAMTAR router and present results of its tests in a physical network. The results are in line with those obtained previously through simulations and they open the way to implementation of a production grade FAMTAR router.

Introduction

FAMTAR (Flow-Aware Multi-Topology Adaptive Routing) is a new approach to routing in IP networks, which provides multipath and adaptive capabilities. It is based on flows, notion currently incorporated in many architectures, including Software-Defined Networking (SDN). Unlike many SDN solutions, FAMTAR is a fully distributed mechanism and does not depend on a central entity. This is a key characteristic, as it significantly improves scalability and resilience. Our aim was to test the efficiency of FAMTAR in real network conditions and validate the results obtained earlier through network simulations. In order to do that, we needed to build a FAMTAR router. In this paper we present our experiences in building a FAMTAR router.

1- Introduction

2- Ethanol architecture

3- Ethanol implementation

4- Case studies

5- Other uses of Ethanol

6- Related work

7- Conclusions

References

Abstract

Wireless Networks have become ubiquitous to support the growing demand from mobile users, and other devices, like in the Internet of Things (IoT). This article proposes a Software-Defined Wireless Networking architecture specialized in 802.11 Wireless LANs, called Ethanol, which provides a more fine-grained control. Ethanol is the first wireless SDN architecture that extends the control to the user devices. Further, Ethanol allows intelligent white-box control with finer grain than the state of the art, since it is optimized for WiFi. The proposed architecture is evaluated on a prototype over three use cases. Ethanol can be deployed in any Access Point (AP) running embedded Linux, since it has a negligible overhead — up to 1% in memory and 0.1% in CPU usage. Our results show that Ethanol dynamically alter the throughput of an application up to 3x during a prioritization period, returning bandwidth to other applications outside this period. Only by controlling the best time to perform the handover, our results show about 45% improvement over the traditional signal-based handover process.

Introduction

Wireless networks have become ubiquitous and dense. More devices will become connected to the network because of IoT. Cisco predicts that 50 billion things will connect to the Internet by 2020.1 Emerging high-speed network standards (e.g. 802.11ad) are migrating to higher frequencies, which are absorbed by walls, thus requiring the densification of APs. Thus, a more refined control of all the network devices is needed to increase future wireless networks scalability in these environments. Current management architectures for Wireless LANs (WLANs) employ proprietary controllers. These controllers perform network-wide optimizations such as adjustment of transmission power at each AP, selection of best operational channels, faster client mobility, and enhanced traceability, as well as Quality of Service (QoS) policy (rate limiting) enforcement, and security. However, those controllers only manage compatible devices, usually from a single manufacturer, since they rely on proprietary interfaces and Management Information Bases (MIBs). Furthermore, current wireless network controllers for WLAN remain (mostly) closed [1] and the industry has little incentive to change. This article presents Ethanol, an Software Defined Networking (SDN) architecture for IEEE 802.11 WLANs. Ethanol refactors the control plane functionalities between the APs, and the controller, creating hooks in the AP implementation that trigger events to be treated by the controller. Also, the controller can use getter/setter methods to change the AP behavior, controlling features such as client mobility, association and disassociation, QoS, link-level parameters and current state, and virtual APs. The key benefit of Ethanol is its flexibility, and the use cases it enables, for example those shown in Sections 4 and 5.

Software Defined Networking (SDN) has focused enormous attractiveness in changing conventional network by means of offering flexible and dynamic network management. It has drawn important concentration of the researchers from together academia and industries. Mainly, integrating SDN in Wireless Body Area Network (WBAN) applications specifies capable results in terms of handling with the issues like traffic management, security, energy efficiency etc. Recent improvements in miniaturization and energy efficient physiological sensor designs in SDN based Wireless Body Area Networks (WBANs) paved the way for health monitoring systems for collection and processing the real-time physiological data. The collection of signals from different sensor allows reliable diagnosis in heterogeneous than in homogeneous WBANs. Inspired by the evolutions of heterogeneous WBANs, a study on Wireless Electroencephalography Sensor Networks (WESNs) is carried out under distributed signal processing. The distributed WESNs are designed under two different hierarchy i.e. Hierarchical Fully-Connected Topology (HFCT) and Ad-Hoc Nearest-Neighbor Topology (ANNT) to improve the energy-efficiency using distributed Multi-channel Weighted Weiner Filter design (MW2F). Here, each module transmits linear combination of local channels with other modules. The power efficiency is improved Journal Pre-proof in MW2F signal processing algorithm by avoiding centralization of EEG data. A case study is carried out to test the reduced energy consumption after the removal of eye blink artifacts and it is tested with centralized counterparts. The MW2F is evaluated in both topologies against centralized environments and significant reduction of eye blink artifacts improves the energy efficiency in HFCT than other topologies.

Introduction

The energy efficient and miniaturization is the most hot research area in health monitoring systems, where the body is placed with several sensor nodes. These nodes can collect the EEG data and process it in real time and measures are considered to improve the energy efficiency. This system is called as Wireless Body Area Networks (WBANs) [1], and it offers higher demand in the area of physiological monitoring in hospital environment. The WBAN system is composed of several nodes with physiological sensors, wireless communication devices and a signal processing unit. These nodes can communicate with other sensor nodes placed over the body or it can be connected to a fusion center or a storage device. The signals collected from various resources are analyzed jointly in heterogeneous WBANs in real time environment for medical diagnosis.

1. Introduction

2. 5G service quality and business requirements

3. 5G network softwarization and slicing: Concepts & use cases

4. 5G network slicing enabling technologies

5. State-of-the-Art: 5G network slicing architectures and implementations

6. Open source orchestrators, proof of concepts & standardization efforts

7. 5G network slicing orchestration and management

8. Future challenges and research directions

9. Conclusion

Declaration of Competing Interest

Acknowledgment

Appendix A. Supplementary materials

Research Data

References

Abstract

The increasing consumption of multimedia services and the demand of high-quality services from customers has triggered a fundamental change in how we administer networks in terms of abstraction, separation, and mapping of forwarding, control and management aspects of services. The industry and the academia are embracing 5G as the future network capable to support next generation vertical applications with different service requirements. To realize this vision in 5G network, the physical network has to be sliced into multiple isolated logical networks of varying sizes and structures which are dedicated to different types of services based on their requirements with different characteristics and requirements (e.g., a slice for massive IoT devices, smartphones or autonomous cars, etc.). Softwarization using Software-Defined Networking (SDN) and Network Function Virtualization (NFV)in 5G networks are expected to fill the void of programmable control and management of network resources.

In this paper, we provide a comprehensive review and updated solutions related to 5G network slicing using SDN and NFV. Firstly, we present 5G service quality and business requirements followed by a description of 5G network softwarization and slicing paradigms including essential concepts, history and different use cases. Secondly, we provide a tutorial of 5G network slicing technology enablers including SDN, NFV, MEC, cloud/Fog computing, network hypervisors, virtual machines & containers. Thidly, we comprehensively survey different industrial initiatives and projects that are pushing forward the adoption of SDN and NFV in accelerating 5G network slicing. A comparison of various 5G architectural approaches in terms of practical implementations, technology adoptions and deployment strategies is presented. Moreover, we provide a discussion on various open source orchestrators and proof of concepts representing industrial contribution. The work also investigates the standardization efforts in 5G networks regarding network slicing and softwarization. Additionally, the article presents the management and orchestration of network slices in a single domain followed by a comprehensive survey of management and orchestration approaches in 5G network slicing across multiple domains while supporting multiple tenants. Furthermore, we highlight the future challenges and research directions regarding network softwarization and slicing using SDN and NFV in 5G networks.

Introduction

The exponential growth of mobile video services (e.g., YouTube and Mobile TV) on smart devices and the advances in the Internet of Things (IoT) have triggered global initiatives towards developing the fifth-generation (5G) mobile and wireless communication sys-tems [1,2,3]. The increasing number of smart devices (e.g., tablets and smartphones) and the growing number of bandwidth-hungry mobile applications (e.g., live video streaming, online video gaming) which demand higher spectral efficiency than that of 4G systems are posing significant challenges in 5G. The Cisco Visual Networking Index (VNI) Forecast [4] predicts that IP video traffic will be 82% of all consumer Internet traffic by 2022, up from 75% in 2017. Mobile video traffic alone will account for 78% of the global mobile data traffic. While the traffic for virtual/augmented reality (VR/AR) will increase at a Compound Annual Growth Rate (CAGR) of 82% between 2017 and 2022, the traffic growth rates of TVs, tablets, smartphones, and M2M modules will be 21%, 29%, 49%, and 49%, respectively. Such a tremendous growth will be the result of 12.3 billion mobile-connected devices, which is expected to even exceed the world’s projected population of 8 billion by 2022. A 5G connection is expected to generate 4.7 times more data than that of 4G [4].

Access control technologies are fundamental for addressing the security and privacy requirements of the Internet of Things (IoT). This paper proposes an access control solution for Constrained Application Protocol (CoAP)- based IoT services. The proposed solution considers a network of a single provider that interconnects various IoT endpoints. It leverages the Software-Defined Networking (SDN) paradigm and implements application aware policy enforcement at the network level. All operations are transparent to the IoT endpoints and no modifications are required to the IoT communication protocol. Furthermore, our solution is built on standard OpenFlow, hence it is realistic and it can be easily deployed to an existing network. We prove the feasibility of our solution through a proof of concept implementation using network emulation.

Introduction

Nowadays, many aspects of our life are controlled–or assisted–by cyber-physical systems. The so-called Internet of Things (IoT) is already used in many domains, including agriculture, patient monitoring, home automation, well-being, smart cities, and many others. The IoT is mainly composed of devices which may be deprived of computational power, continuous network connectivity, energy, or even physical security. Therefore, it comes as no surprise that applying security solutions in this environment is a challenging problem. In this paper, we focus on a particular aspect of security, that is access control. We consider the case of a network of a single operator that interconnects various IoT devices. These devices provide resources or actuation services, and can be accessed using the Constrained Application Protocol(CoAP) [1]. In order to motivate our solution, we consider the use case of a smart city management system. This system is composed of IoT sensors (e.g., temperature sensors) and actuators (e.g., switches). Our goal is to enable system administrators to define context-aware access control policies that will mediate access to the IoT devices. More formally, we want to provide a Mandatory Access Control (MAC) solution where policies are centrally defined by the system administrators and cannot be modified or overridden by end users. An example of such policy, in our reference system, is the case of a switch that turns on and off street lights; in that case the system administrator could create an access control policy that defines that “street lights switches can be turned on after 8pm and turned off after 6am, and all operations should originate from the management center building”.

As Software-Defined Networking (SDN) is getting popular, its security issue is being magnified as a new controversy, and this trend can be found from recent studies of presenting possible security vulnerabilities in SDN. Understanding the attack surface of SDN is necessary, and it is the starting point to make it more secure. However, most existing studies depend on empirical methods in different environments, and thus they have stopped short of converging on a systematic methodology or developing automated systems to rigorously test for security flaws in SDNs. Therefore, we need to disclose any possible attack scenarios in diverse SDN environments and examine how these attacks operate in those environments. Inspired by the necessity for disclosing the vulnerabilities in diverse SDN operating scenarios, we suggest an SDN penetration tool, DELTA, to regenerate known attack scenarios in diverse test cases. Furthermore, DELTA can even provide a chance of discovering unknown security problems in SDN by employing a fuzzing module. In our evaluation, DELTA successfully reproduced 26 known attack scenarios, across diverse SDN controller environments, and also discovered 9 novel SDN application mislead attacks.

Introduction

Security has been a subject of controversy in many newly emerged networked systems, such as peer-to-peer networks and cloud networks. After their appearance, researchers and practitioners have examined their security issues from various angles to verify their safeness, and this process makes them more secure so that they can be adapted in a real-world system. Software-Defined Networking (SDN), which manages a network in a centralized way, is a recently proposed networking technology, and now it is endorsed by both industry and academia. As SDN technology is getting popular, its security problem is being at issue, and thus researchers are investigating its security issues as they have conducted in other networked systems [26, 46, 39, 20, 25, 1]. Such security-critical reviews of SDNs offer a view into various breaches, but overall, the attack surfaces thus far explored have been quite limited to either highly targeted exploits, such as ARP spoofing or specific vulnerabilities that arise in various SDN components. Each previous result may not be applicable to other SDN environments (e.g., different control planes). Hence, operators seeking to assess security issues in their SDN environments need to survey existing SDN security-related studies and determine relevance on a case-by-case basis. Furthermore, an operator may have to adapt or redesign deployment-specific security test suites. This paper introduces a new SDN security evaluation framework, called DELTA, which can automatically instan- ∗Corresponding author ORCID(s): tiate attack cases against SDN elements across diverse environments, and which may assist in uncovering unknown security problems within an SDN deployment. Motivated by security testing tools in the traditional network security domain [45, 15], DELTA represents the first security assessment tool for SDN environments. Furthermore, we enhanced our tool with a specialized fuzzing module [30] to exploit opportunities for discovering unknown security flaws in SDNs.

Wireless devices have a plethora of technologies at their disposal to connect to the Internet and other services. Management and control of each technology are traditionally isolated, and coordination between technologies is nearly non-existent. This isolation leads to poor resource usage, which in turn reduces performance and service guarantees. To satisfy growing user demands, we need to leverage the different service guarantees offered by each technology. Additionally, we need to improve orchestration between technologies to increase performance and flexibility while offering a more extensive range of service guarantees and maximizing resource utilization across networks and users. In this work, we present the general challenges one encounters when managing heterogeneous wireless networks. We argue that the primary challenge is the heterogeneity itself, the number of different devices and technologies, the different service requirements, and the increasing complexity as a consequence. However, technology abstraction can overcome these challenges. We provide an overview of state of the art commercial and scientific solutions and show their strengths and weaknesses. Based on this, we discuss the current status and what future challenges still await to provide full seamless heterogeneous wireless network management.

Introduction

Today’s devices, such as phones, wireless access points (APs), sensors, and other machines, are often equipped with multiple networking technologies to enable them always to stay connected. This connectivity allows users to use a plethora of Internet and other services through technologies like Digital Subscriber Line (DSL), Long-Term Evolution (LTE), or IEEE 802.11 with even more technologies in the future such as IEEE 802.11ax/ay or 5G solutions [1–3]. We expect this trend to continue, with a further increase in available technologies. From this follow two scenarios: (i) technologies cover similar scenarios but do not share spectrum, such as LTE and IEEE 802.11 bring Internet access to users, (ii) technologies cover different scenarios but share the spectrum, such as IEEE 802.11 and Bluetooth. Currently, technologies are isolated, and applications or the operating system takes care of technology selection. This isolation leads to inefficient use of each technology; for example, one technology is congested while another has plenty of free resources. To truly achieve the high bandwidth and low latency requirements of today’s services, orchestration across technologies needs to be in place. Only a holistic approach allows optimizing the performance of services in these heterogeneous networks.

1- Introduction

2- Related work

3- System framework

4- System model and problem formulation

5- The heuristic deployment algorithm

6- Performance evaluation

7- Conclusion

References

Abstract

The emergence of Software Defined Networking (SDN) and Network Function Virtualization (NFV) enables flexible service provisioning and deployment. However, with the continuous expansion of network scale and sharp increasing of end users, how to flexibly provide network services across multiple SDN domains for users is becoming a critical issue. A major challenge in the multi-domain network service provisioning is the network service deployment method taking into account energy efficiency. In this paper, we study the problem of how to optimally deploy network services across multiple SDN domains with the target of saving energy while achieving the load balancing of multi-domain networks. Specifically, firstly, we propose a novel multi-domain network service deployment framework by integrating SDN architecture and NFV technology, which can intelligently deploy virtual network functions (VNFs) into multi-domain networks. Secondly, we formulate this problem as a multi-objective optimization model to achieve the minimization of energy consumption and load balancing of multi-domain networks. Furthermore, we present a heuristic network service deployment algorithm to solve it. Finally, simulation results demonstrate that the proposed heuristic service deployment algorithm is efficient and outperforms comparison algorithms in terms of energy consumption and load balancing degree.

Introduction

In recent years, a large number of middleboxes (e.g., Network Address Translator (NAT), Firewall (FW) and Intrusion Detection System (IDS)) have been developed and deployed in current networks to provide various network services. However, it is inefficient for service provisioning due to the ossified architecture of traditional network and middleboxes’s heavy dependence on specific hardware equipment [1]. Fortunately, the emergence of Network Function Virtualization (NFV) brings new opportunities to address the challenge [2,3]. It implements network functions in software rather than hardware devices, running on virtual machines hosting on commodity servers [4]. One network service in NFV can be represented as one Service Function Chain (SFC), which is composed of a series of Virtual Network Functions (VNFs) in a given order. To provide efficient service, the SFC is deployed in several different servers and traffic is steered to pass through a set of VNFs in a specific sequence. NFV supports flexible service provisioning, significantly reducing Capital Expenditures (CAPEX) and Operational Expenditures (OPEX). On the other hand, Software Defined Networking (SDN), as an emerging networking paradigm, achieves the separation of control plane and data plane, and provides logically centralized control capacity as well as powerful programmable capability [5–10].

I- INTRODUCTION

II- RELATED WORK

III- MACHINE LEARNING FOR SDN SECURITY

IV- MACHINE LEARNING FOR TRAFFIC CLASSIFICATION

V- CONCLUSIONS

REFERENCES

Abstract

As a new network architecture, software defined network (SDN) separates the control plane from the forwarding plane which enables administrators to define and control the network through the method of software programming, provides a new research direction for the next generation of network architecture. At the same time, the machine learning technology has been developed rapidly in recent years and some studies have begun to introduce machine learning methods into SDN to improve the efficiency of network management and conformity, or to solve problems that cannot be solved easily by traditional methods. The paper analyses, summarizes and introduces these researches which used the supervised learning, unsupervised learning or semi-supervised learning methods to solve some specific problems on SDN, and it will help later researchers understand the filed more quickly and promote the development of the machine learning technology in SDN.

INTRODUCTION

The machine learning is an important branch of artificial intelligence research area, and various machine learning algorithms such as Support Vector Machine (SVM) [1], KNearest Neighbor (KNN) [2], Logistic Regression (Logistic Regression) [3], Boosting [4], etc. have been widely used to solve complex problems in engineering and science fields. The emergences of big data and GPU technology provide more powerful support for the development of machine learning technology. The deep learning [5] proposed by Geoffrey Hinton et al. in 2006 pushed the machine learning to a new climax, and made machine learning rapidly develop into an independent area and be applied to various fields, such as pattern recognition, data mining, bioinformatics and autonomous driving, etc. Clark proposed a network architecture of “A Knowledge Plane for the Internet” in 2003, which relies on machine learning and cognitive technology to manipulate the network [6]. The knowledge plane (KP) would bring many benefits to the network and change the way we operate, optimize and troubleshoot the network. But the distributed network architecture results in that each node (i.e., switches, routers) only has a partial view of the entire system, which makes it a huge challenge to apply machine learning to the network. Logical centralized control will alleviate the complexity of learning in a distributed environment.

1- Introduction

2- Background and related work

3- The SLDP protocol

4- Simulation results and discussions

5- Conclusion and future work

References

Abstract

In Software Defined Networks (SDNs), the global view of the underlying network topology is created and maintained at the logically centralized controller. SDN achieves it by decoupling the data plane from the control plane. The up-to-date global view at SDN controller enables the applications (running on top of it) to innovate through dynamic network programmability. To establish a global view, a controller needs to discover a physical topology of the underlying SDN network infrastructure, which is challenging due to various reasons such as the lack of SDN protocols standardization and authentication mechanisms, use of sub-optimal link discovery protocols (e.g., OFDP and LLDP), dynamic topology due to movement of virtualized data centers, switches, and multi-tenant cloud networks, and lack of integration of security schemes for the topology discovery. In this paper, we propose a SDN Link Discovery Protocol (SLDP) for efficient discovery and extraction of topology information in SDN networks. The design of SLDP is motivated from the need of a secure, lightweight, and efficient link discovery protocol in SDN. SLDP aims to prevent, detect, and mitigate various security threats such as poison, replay, and flooding attacks, which are due to lack of source authentication, lack of packet integrity checks, and reuse of static packets. SLDP creates and maintains the global network topology at SDN controller by using smaller size and lower number of SLDP packets during the topology discovery process. Thus, it significantly minimizes the topology discovery overhead in the network. We implemented SLDP on Mininet emulator, and the results show the effectiveness and correctness of SLDP concerning topology discovery time, CPU computational time, and bandwidth overheads, when compared with the traditional OpenFlow Link Discovery Protocol (OFDP). Additionally, SLDP successfully prevent, detect, and mitigate various attacks (e.g., poison, replay, and flooding) in different SDN scenarios.

Introduction

For any data center, the essential requirements are robustness and manageability. Software Defined Network (SDN) offers programmability, flexibility, and openness to ensure these requirements [1,2]. Due to the separation of data plane and control plane, the SDN controller exhibits a global view of the underlying network topology, which enables it to make the optimal decisions for various applications that runs on top of the controller. For instance, applications such as load-balancing and shortest pathfinder, uses the global view to function efficiently. The view construction and maintenance require the discovery of underlying network topology, which consists of switches and links. The existing SDN controllers use OpenFlow Discovery Protocol (OFDP) with Link Layer Discovery Protocol (LLDP) packets for link discovery, which is prone to various security threats. In SDN, the global view is generated by performing the switch discovery, the link discovery, and sometimes the host discovery. Once an OpenFlow-enabled switch connects to the network, it performs a TCP three-way handshake with a pre-stored remote socket residing at the SDN controller. After successful handshaking, both negotiate on the OpenFlow version. Subsequently, the switch is asked for its capabilities and ports status. These steps help controller to discover the switch with available ports. To perform various topology-aware activity, link discovery is mandatory. Most of the SDN controllers use OpenFlow Link Discovery Protocol (OFDP) and Link Layer Discovery Protocol (LLDP) for the discovery process.