Now showing 1 - 6 of 6
  • Publication
    Human Behavior Prediction Though Noninvasive and Privacy-Preserving Internet of Things (IoT) Assisted Monitoring
    (IEEE, 2019-04-18) ;
    As the supporting technologies for Ambient Assistant Living (AAL) in Internet of Things (IoT) domain have become more powerful and more attractive, the related systems will be widely deployed and put into action. With all associated embedded IoT sensing devices, how to maintain users’ privacy and data security is a highly concerned problem. There are generally two approaches to protect privacy. One is to implement complex security protocols to guarantee the safety of sensing, storage and data transmission. Another one is to prevent the privacy issues and concerns from the source. This proposed research will provide a concept of a framework that can support behaviour monitoring through noninvasive and privacy preserving sensing. The data collected, transmitted and used for analysing in this framework is sensing information with low richness. This framework aims to increase the users’ perceived privacy in existing monitoring systems to avoid data over-collection and over exposure problems.
  • Publication
    Internet of Too Many Things in Smart Transport: The Problem, The Side Effects and The Solution
    (IEEE, 2018-10-24) ;
    The Internet of Things (IoT) involves embedding electronics, software, sensors, and actuators into physical devices, such as vehicles, buildings, and a wide range of smart devices. Network connectivity allows IoT devices to collect and exchange data. The prevalence of IoT devices has increased rapidly in last five years, driven by cheaper electronics and a desire to monitor and control the physical world. We introduce the concept of IoT flood to describe the increased use of IoT devices. Just like the data deluge, the IoT flood has potential benefits and risks. This paper focuses on the hidden side effects of the increased usage of IoT, such as energy consumption, physical pollution, radiation, and health issues. We indicate that an evaluation system with carefully designed metrics reflecting the impact of IoT flood with input from academic, industry, and government is required. We propose some practical measures that can reduce the IoT flood, such as common platforms and data sharing to reduce the side effects. This paper demonstrates the IoT flood problem and potential solutions by examining the intelligent transport system domain where IoT is being deployed to solve problems related to time efficiency and energy consumption through smart mobility.
      392Scopus© Citations 21
  • Publication
    Spreading the load in a tree type routing structure
    Many routing protocols have advanced the Wire- less Sensor Network (WSN) paradigm with each new method offering unique ways to maximise Quality of Service (QoS) while minimising energy cost to the network. Tree routing is a well examined method with a proven record in offering a high level of service with trees constructed to fulfil a particular routing objective defined by a chosen metric. The tree structure can be maintained with low overall network overhead but exhibits a weakness with regard to load balancing. Without effective load balancing particular nodes in the network may be subject to excessive network load, leading to uneven energy consumption in the network. This in turn can lead to an unwanted scenario of premature node failure. Neighbourhood metrics is presented in this paper as a means to preserve network objectives while achieving improved load distribution in a tree type routing structure. Neighbourhood metrics offer a framework for expand- ing on currently used metrics to include information on the quality of a nodes neighbourhood in addition to the current forwarding route. Neighbourhood metrics is compared to the current state of the art in the form of the Routing Protocol for Low Power and Lossy Networks (RPL) implementation using the direct Expected Transmissions (ETX) metric. Neighbourhood metrics exhibits improved load distribution in a number of open public testbeds.
      478Scopus© Citations 9
  • Publication
    Towards Smart Networking through Context Aware Traffic Identification Kit (TriCK) in 5G
    (IEEE, 2018-11-12) ;
    In order to distribute diverse traffic flow into proper network interfaces, Access Network Discovery and Selection Function (ANDSF) is proposed by 3GPP, which can distribute every traffic flow to a preferred network interface according to several observed features from that flow. However, the static policies in ANDSF can neither understand the context nor adapt to real time changes. In order to address that problem, in our previous work, we have proposed a server-client based Context aware Traffic identification Kit (TriCK) to dynamically identify traffic, which can extend the functionalities of 3GPP ANDSF. It can classify traffic data not only based on its own characteristics, but also the real time network conditions and the current context. In this paper, we provide an implementation for the network selection component in TriCK based on clustering techniques, with a complexity of O(n). A static version and a dynamic version of the implementation are analysed. The static approach is easy to implement and comprehend. The static solution can distribute the traffic flow according to the traffic characteristics and the network context. The dynamic approach can further balance the traffic load between different network interfaces and therefore provide an overall better transmission quality.
      334Scopus© Citations 2
  • Publication
    Explainable Text-Driven Neural Network for Stock Prediction
    It has been shown that financial news leads to the fluctuation of stock prices. However, previous work on news-driven financial market prediction focused only on predicting stock price movement without providing an explanation. In this paper, we propose a dual-layer attention-based neural network to address this issue. In the initial stage, we introduce a knowledge-based method to adaptively extract relevant financial news. Then, we use an input attention to pay more attention to the more influential news and concatenate the day embeddings with the output of the news representation. Finally, we use an output attention mechanism to allocate different weights to different days in terms of their contribution to stock price movement. Thorough empirical studies based upon historical prices of several individual stocks demonstrate the superiority of our proposed method in stock price prediction compared to state-of-the-art methods.
      730Scopus© Citations 27