No abstract

Convolutional neural networks (CNNs) are the specific architecture of feed-forward artificial neural networks. It is the de-facto standard for various operations in machine learning and computer vision. To transform this performance towards the task of network anomaly detection in cyber-security, this study proposes a model using one-dimensional CNN architecture. The authors' approach divides network traffic data into transmission control protocol (TCP), user datagram protocol (UDP), and OTHER protocol categories in the first phase, then each category is treated independently. Before training the model, feature selection is performed using the Chisquare technique, and then, over-sampling is conducted using the synthetic minority over-sampling technique to tackle a class imbalance problem. The authors' method yields the weighted average f-score 0.85, 0.97, 0.86, and 0.78 for TCP, UDP, OTHER, and ALL categories, respectively. The model is tested on the UNSW-NB15 dataset.

No abstract

The robustness of wireless sensor networks (WSNs) has broadened the use of wireless communication systems for modern machine-to-machine (M2M) communication. Internet-of-Things (IoT) being one of the most explored technologies has gained wide attention due to low-cost WSN communication systems. IoT is one of the dominant technologies used in Smart City implementation demands better communication paradigm including mobility assisted transmission paradigms; however, native IEEE 802.15.4 standard does not have mobility provision that limits its use for real-time applications. On contrary, implementing mobility with classical WSN might impose significantly great topological changes and node or network condition variations which cannot be dealt with the traditional reactive routing approach. Incorporating node and/or network awareness with proactive network management can enable classical WSN to support mobility, which can be vital for IoT based Smart City Planning and Management (SCPM). With this motivation, in this study, a robust Cross-Layer Architecture based WSN Routing Protocol for Event-Driven M2M Communication in SCPM (CWSN-eSCPM) is developed. CWSN-eSCPM encompasses Proactive Node Management Strategy, Data-Centric Service Differentiation and Fair Resource Scheduling (DCSDFRS), Packet Velocity Estimation, Cumulative Congestion Estimation, Dynamic Link Assessment that enables optimal Best Forwarding Node selection for deadline sensitive and reliable data communication. CWSN-eSCPM applies dynamic link quality, cumulative congestion degree, and packet velocity of a node to enable optimal routing decision. DCSDFRS enables optimal resource provision to the real-time data while assuring maximum possible resource for non-real-time data that assures Quality of Service provision for event-driven critical communication in SCPM. Noticeably, CWSN-eSCPM protocol has been applied on top of IEEE 802.15.4 protocol standard, while preserving backward compatibility feature, it can be used for WSN assisted communication purposes. The authors' proposed protocol has performed better in terms of packet delivery ratio, packet loss ratio and deadline miss ratio for both real-time data as well as non-real-time data.

The exponential rise in the demands of the wireless communication system has alarmed industries to achieve more efficient and quality‐of‐service (QoS) centric wireless communication networks. The decentralised and infrastructure‐less nature of wireless sensor networks (WSNs) enable it to be one of the most sought and used wireless network globally. Its cost‐efficiency and functional robustness towards low‐power lossy networks make it suitable for internet‐of‐things (IoT) applications. In recent years, IoT technologies have been used in diverse applications, including Smart City Planning and Management (SCPM). Although, mobile‐WSN has played a decisive role in IoT enabled SCPM, its routing optimality and power transmission have always remained challenging. Noticeably, major existing researches address mainly on routing optimisation and very few efforts are made towards dynamic power management (DPM) under non‐linear network conditions. With this motive, in this study, a highly robust and efficient QoS – centric reinforcement learning‐based DPM model has been developed for mobile‐WSN to be used in SCPM. Unlike classical reinforcement learning methods, the authors’ proposed advanced reinforcement learning‐based DPM model exploits both known and unknown network parameters and state‐activity values, including bit‐error probability, channel state information, holding time, buffer cost etc. to perform dynamic switching decision. The key objective of the proposed model is to ensure optimal QoS oriented DPM and adaptive switching control to yield reliable transmission with the maximum possible resource utilisation. To achieve it, they proposed model has been developed as a controlled‐Markov decision problem by applying hidden Markov model it obtains known and unknown parameters, which are subsequently learnt using an enhanced reinforcement learning to yield maximum resource utilisation while maintaining low buffer cost, holding cost and bit‐error probability to retain the QoS provision.