Showing papers presented at "IEEE International Conference on Advanced Infocomm Technology in 2020"

Graph and Natural Language Processing Based Recommendation System for Choosing Machine Learning Algorithms

Abstract: Machine Learning is a subset of Artificial Intelligence(AI). It provides the systems with the ability to learn automatically and perform independently without being programmed. When concentrating on the machine learning project choosing an algorithm is one of the most important steps. Machine learning algorithms are the math and logic that adjust the training of the model and the performance. When it comes to a machine learning project the choosing a correct algorithm for implementing the model is a big task. If the developer doesn’t choose the suitable and most efficient algorithm for the program the accuracy of the program can be decreased. When we look at these machine learning algorithms there is no solution or no approach that fits for all. Several factors affect when choosing a machine learning algorithm such as the number of data, type of the problem ..etc. Most of the time a developer chooses the machine learning algorithm using his prior experience or analyzing several past similar projects. However, when it comes to a beginner it can be a tough experience. Most of the time beginners try with several algorithmic approaches to implement their model without any understanding. However, to avoid this there is no proper solution.Therefore here in this research, proposed an NLP and graph analytics-based approach for recommending a machine learning algorithm for a project. The summary of the proposed solution is here it analyzes past algorithms used in machine learning projects which are stored in a graph using NLP based keyword analysis and recommend the most suitable algorithmic approach. When the user inputs his project idea by using the Natural Language processing it generates keywords for the project description. Thereafter the system analyzes the graph which stores past machine learning projects and used technologies to find the most suitable algorithmic approach to the users’ project. Moreover, it shows how the proposed algorithms were used in past similar projects. Therefore by using this system the developers can get a clear idea of the algorithm approach that they need to choose.

Simulative Assessments of Credit-Based Shaping and Asynchronous Traffic Shaping in Time-Sensitive Networking

Abstract: To extend industrial Ethernet to guarantee real-time quality-of-service (QoS), the IEEE 802.1 Time-Sensitive Networking (TSN) task group has been developing a series of specifications, including the flow control series, of which Credit-Based Shaping (CBS) and Asynchronous Traffic Shaping (ATS) were defined in IEEE 802.1 Qav standard and IEEE 802.1 Qcr draft respectively. While CBS is a per-class scheduling strategy, ATS is a per-flow scheduling. To analyze performance of these flow control strategies, comparison among CBS, ATS and Strict Priority Queuing (SPQ) is performed using a suite of self-developed models. Models were implemented and operated in the OMNeT++ simulation platform. Moreover, CBS and ATS are implied in different network load. And the results show that ATS flow control strategy can provide better real time performance guarantee for aperiodic traffics than CBS in heavy network load.

An Uncertainty-Based Traffic Training Approach to Efficiently Identifying Encrypted Proxies

Abstract: Encrypted proxies, such as Shadowsocks and v2ray, are increasingly used to reserve user privacy and circumvent censorship. However, they are also widely misused by attackers to carry out illegal activities like malware downloading, information theft. Therefore, identifying encrypted proxies is a fundamental task concerning cyber security for network administrators. Existing studies focus on traffic feature engineering and designing the classification model. Although indispensable, they do not consider the training efficiency problem, thereby unable to approach the best possible performance when the number of affordable training samples is limited due to resource constraint. In this paper, we propose an uncertainty-based traffic sample selection strategy to boost traffic training of encrypted proxies. The proposed strategy allows one to use fewer samples to quickly learn diverse traffic characteristics. Through experiments, we demonstrate that our strategy significantly outperforms random sample selection, and hence substantially improves identification performance.

An Edge Server Placement Algorithm in Edge Computing Environment

Abstract: With the rapid development of the Internet of things technology, the big data generated at the edge of the network prevents the centralized cloud computing center from efficiently processing massive amounts of data, and the real-time requirements of users are difficult to guarantee. In order to meet the above challenges, edge computing technology came into being. It sinks cloud computing capabilities from the cloud to the edge of the network to ensure real-time business needs. Focusing on the placement problem of edge servers in the edge computing environment, this paper designs an effective edge server placement method to reduce access latency and optimize load balancing. Firstly, the edge server placement model in the mobile edge computing environment is established. Secondly, the immune optimization algorithm is used to solve the problem. Finally, the effectiveness and feasibility of the algorithm are verified based on simulation experiments.

High-Order Harmonic Generations in Epsilon-Near-Zero Aluminum-Doped Zinc Oxide Nanopyramid Array

Abstract: We numerically demonstrate the high harmonic generations (HHGs) up to the seventh order in epsilon-near-zero (ENZ) aluminum-doped zinc oxide (AZO) nanopyramid array Pumped by a 1-kHz repetition rate 50-mW 100-fs laser source at the telecommunication wavelength of 1550 nm, the conversion efficiencies of 141×10−5, 364×10−6, and 209×10−7 are achieved for third-, fifth-, and seventh-harmonic generations, respectively, which are comparable with the experimental results of indium tin oxide and indium-doped cadmium oxide The electric field distributions, localized resonances, and emission patterns at HHG wavelengths are also discussed The results of this work can be potentially useful in designing novel multiwavelength light sources for optical communication and on-chip light circuits

A Beamforming Method Based on Polarization Matching

Abstract: By analyzing the polarization-sensitive array model and combining the polarization information of different elements with space-time processing, a beamforming method based on polarization matching is proposed, which realizes the high-performance beamforming of the polarization-sensitive antenna array. The simulation performance comparison with traditional spatial beamforming confirms the high direction finding accuracy and high signal-to-noise ratio robustness of this algorithm.

Link Failure Analysis and Routing Planning for Fiber Network of Smart Grid

Abstract: Diverse and robust data transmission requirement decide the design for fiber-optic communication network of smart grid A well-designed route scheme is significant to facilitate communication efficiency Aiming at the problem of choosing an optimal transmission path, a link failure monitoring method based on fuzzy logic is designed Complex network theory converts fiber-optic communication network into a concrete model Model analysis and simulations show that the method to evaluate link failure is effective, and routing planning is restricted by many factors The analysis also provides new perspective to combine integrated network of sensing and communication with fiber-optic sensing technology

Fiber Tip Temperature Sensor Based on PVA Filled Silica Tube Fabry-Perot Interferometer

Abstract: Optical fiber temperature sensors have attracted much attention owing to their outstanding intrinsic merits of electromagnetic immunity, remote operation capability, compact size and high sensitivity. In the techniques of optical fiber sensors, the combination of Fabry-Perot interferometer (FPI) with polymer filled resonance cavity shows an easy and feasible way to realize high-sensitive temperature monitoring. In this paper, an optical fiber tip FPI based on polyvinyl alcohol (PVA) filled silica capillary is proposed and experimentally demonstrated for temperature sensing application. The PVA in silica tube is shaped as a cylinder to form a Fabry-Perot resonance cavity, the fiber/PVA facet and PVA/air facet act as two reflection mirrors. When ambient temperature changes, the cavity length and effective refractive index of PVA varies, therefore induces a spectral evolution. By monitoring the interference dip wavelength, dip intensity and fringe visibility, the proposed sensor enables accurately online detect the ambient temperature change. The sensitivities are 90.4 pm/°C, −0.117 dB/°C and 2.4×10−3 /°C, respectively, in the test range of 50−70 °C. The proposed fiber tip sensor can be applied for temperature monitoring in space-limited environment with simple fabrication, low cost and good robustness.

An Approach to All-Optical Wavelength Conversion Based on XPM Effect in HNLF for 16QAM/OFDM Signal

Abstract: In recent years all-optical wavelength conversion (AOWC) technology has researched widely Based on four-wave mixing (FWM) in semiconductor optical amplifier (SOA) to realize AOWC have been demonstrated But the design of cross phase modulation (XPM) based wavelength converter can increase system spectral efficiency at higher data rate In this paper, an AOWC based on XPM effect in highly nonlinear fiber (HNLF) for 56 Gbps 16-Orthogonal amplitude modulation(16QAM)/orthogonal frequency division multiplexing (OFDM)signal approach has been proposed The performance evaluation for this experiment focuses on the system bandwidth utilization presented in terms of optical spectrum and the quality of service (QoS) measured in terms of 16QAM signal constellation diagram, Q-factor and bit-error-rate (BER) at the receiver The observed numerical results demonstrate that this novel HNLF scheme effectively improves Q-factor gain up to 36 dB

Frequency Estimation Based on MLE-M-Rife Algorithm

Abstract: The M-Rife algorithm is widely used in frequency estimation field. By using the interpolation and spectrum shift it can achieve high-precision frequency estimation. In this paper, a new frequency estimation method based on the Multi-M-Rife algorithm is proposed. First M-Rife algorithm is used to estimate the intra pulses’ rough-estimated frequency, then an inter-pulse sinusoidal wave sequence is constructed, after that by using maximum likelihood estimation (MLE) the frequency offset between signal and sequence has been detected. Experimental results show that by using the new algorithm the frequency estimation accuracy of pulse signals is better than 0.3Hz, and the algorithm can also achieve a high detection rate under the condition of low signal-to-noise ratio.

Generalized Spatial Multiplexing for Optical Wireless Communication Systems

Abstract: Spatial Multiplexing (SMP) is one of the most commonly used multiple-input multiple-output (MIMO) techniques in optical wireless communication (OWC) systems Although SMP can achieve high spectral efficiency, it suffers from severe inter-channel interference (ICI) In this paper, we for the first time propose a novel MIMO technique, ie, generalized SMP (GSMP), for intensity modulation/direct detection (IM/DD) OWC systems Differing from conventional SMP which activates all the transmitters to transmit signals, GSMP selects a variable number of transmitters for signal transmission The information bits of GSMP can be carried by both the spatial index symbols via variable-number transmitter selection and the constellation symbols transmitted by the activated transmitters Compared with SMP, GSMP has the advantages of reduced ICI and enhanced spectral efficiency The obtained analytical and simulation results show that GSMP can achieve greatly improved bit error rate performance than conventional SMP to achieve the same spectral efficiency in indoor 2 × 2 and 4 × 4 MIMO-OWC systems

Contribution to Data Compression and Transmission Techniques: Application in the Field of Computer Science and Distance Education

Abstract: The emergence of network interconnection technology promotes collaboration and the sending of large amounts of data into the network In fact, we are at a time when entities are forced to share their data in a collaborative working environment and mutualisation Data exchanged through collaborative networks is constantly growing As part of a virtual organization, as in the case of distributed networks or digital universities, actors often have to exchange large data in the form of images, sounds and videos This plethora of large data creates slowdowns within the network and dramatically slows the flow Data compression is a mandatory solution This article proposes to provide a solution on compression and data transmission techniques The compression technique consists of using a polynomial formula to compress the data to the maximum before its transmission This makes it possible to send large amounts of data faster, to gain bandwidth and to increase the fluidity of transmissions So the knots will be less loaded

Multimedia Data Encryption and Compression via Hybrid Chaotic Compressive Sensing and Latent Vector Transmission

Abstract: We propose a digital transmission system guaranteeing both encryption and compression of multimedia data Encryption is performed via chaotic compressive sensing, while compression is ensured via latent vector transmission based on a stacked autoencoder (StAE) We demonstrated successful compression in eight cases up to a 003 compression ratio corresponding to 80 pixels transmitted A compression ratio below 04 with StAE can achieve peak signal-to-noise ratio (PSNR) and mutual information (MI) between original and reconstructed images comparable to compression ratios between 06 to 08 without StAE, meaning same performances for about half of the pixels Moreover, StAE can guarantee at least a 78,4% slower decrease of PSNR per pixel transmitted and at least 813% slower decrease of MI per pixel, averaging at 816% and 9396%, respectively, for the four tested images StAE can hence push the compression ratio of secure multimedia data while preserving the quality of the received image at receiver side

A Hyper-Converged Video Management System Based on Object Storage

Abstract: In recent years, online education, video surveillance industry and live streaming platform have encountered huge evolution Massive video capture, storage, and access are facing increasing demands But in large-scale video scenario, traditional video management systems encounter the challenges of reliability, extendibility and fault-tolerance In this paper, a novel hyper-converged massive video management system based on object storage is designed and realized The hyper-converged system composes distributed video capture, tiering video storage and read/write splitting load balance In this loosely coupled system, all components are distributed on each node and communicate via message queues The components can harness unused resources for example CPU, memory and network of hyper-converged nodes This system exploits high extensibility and reliability of distributed system to satisfy large-scale video capture, storage and access

Feedback-Based Directly Ghost Imaging Method

Abstract: A directly ghost imaging method based on a feedback loop is proposed The image of an object is reconstructed directly by actual optical devices, rather than a post processing of a large number of data A voltage-controlled optical attenuator and a voltage reverse amplifier are added to the ghost imaging system The voltage-controlled optical attenuator is set in front of reference arm camera The signal of the object arm, which is detected by the bucket detector, passes through the voltage inverting amplifier and is connected to the control port of the optical attenuator Thus the reference speckles are controlled by the bucket signal, and a feedback loop is formed By adjusting the exposure time of the reference arm camera, these devices complete a simple fluctuations correlation operation, and the image of the object is reconstructed directly via the reference arm camera Results show that, by using cameras and detectors of common level, this new ghost imaging method can achieve the object image recovery within a time of less than one second

Free-Space Data Transmission with Pulse Seeded Supercontinuum Based Partially Coherent Beams in a Tunable Turbulent Channel

Abstract: We experimentally demonstrated a free space optical (FSO) transmission system with a pulse seedded wide-spectrum partially coherent beams (PCB) in a tunable turbulent channel With 4Gbit/s on-off keying (OOK) modulation, the FSO transmission characters of the wide-spectrum laser are analyzed The transmission link with the pulse seeded PCB shows the receive power fluctuation of 0278dBm, receive SNR of 1056dB, and the receive sensitivity of −348dBm, which is well beyond the performance of coherent beams (CB) and non-seeded PCB

A Deep Learning Approach to Detect Anomaly in Software-Defined Network

Abstract: Software-Defined Network (SDN), an innovating technique, endows flexibility and programmability to the network in contrast to the conventional one. Its idea of decoupling control and data plane brings convenience to manage networks like load balancing, traffic engineering, virtual firewall etc. Nevertheless, the separated plane structure results in novel threats, composed by the common in conventional network and the specialized in SDN. In this paper, we propose a deep learning approach to detect anomaly in SDN with OpenFlow by analyzing multiple metrics extracted from OpenFlow switch metadata. Evaluated by four trained deep learning models to classify the multivariate time series, we obtained an average accuracy of 83.8%.

A Novel Brain Tumor Segmentation Approach Based on Deep Convolutional Neural Network and Level Set

Abstract: In recent years deep convolutional Neural Network (DCNN) gets a big success in brain tumor segmentation. But there are artifacts in the border region of segmentation results using DCNNs. To solve this question, we propose a hybrid model including DCNNs and traditional segmentation methods. First, we use U-Net and ResU-Net network in coarse segmentation. In order to deepen the network levels and improve the network performance, we add residual module to U-Net and comprise the ResU-Net. Second, we use level set in fine segmentation of tumor boundary. We take the intersection of the coarse segmentation outputs of U-Net and ResU-Net as input of level set module. The aim of taking the intersection of U-Net and ResU-Net outputs is to get better initialization information for the level set algorithm and accelerate the evolution of level set functions. The proposed approach is validated on the BraTS 2018 challenge dataset. The metrics used to evaluate the segmentation results are: Dice, Specificity, Sensitivity, Hausdorff distances (HD). We compare our approach with U-Net, ResU-Net and some other methods. The experimental results indicate our approach is better than some other deep networks.

Implementing Multiple-Social-Network Meta-APIs to Support OSINT Programming

Abstract: Open source intelligence (OSINT) collects a set of approaches, methodologies, and tools, to make investigation about individuals on the basis of information publicly available over the Internet For this important task, online social networks (OSN) represent nowadays the main source of information, so that large attention should be devoted to any approach that improves the capabilities of investigating over OSNs The aim of this paper is to highlight that, despite the availability of tools and platforms for OSINT which also exploit OSNs, there is a gap in terms of software engineering perspective Indeed, there is no a programming paradigm oriented to the abstract concept of online social network, independent of the specific instance of this concept This resumes the multiple-social-network perspective in a software engineering key, and translates into the definition of a set of meta-APIs allowing the programmer to write generic and polymorphic code, naturally oriented to social-network investigation This paper, born through the experience of an industrial project, shows the above claim by choosing some relevant fragments of the proposed framework

Distributed Liquid Level Measurement with High Sensitivity and Large Sensing Range Based on ϕ-OTDR

Abstract: Liquid level sensor with large dynamic and high-resolution is essential for the application of industry monitoring In this work, a distributed optical fiber liquid level sensor is proposed and demonstrated based on phase-sensitive optical time domain reflectometry (φ-OTDR) In the basic of the thermal optic effect, the temperature change will induce the fluctuation of the effective refractive indexes of the fiber core, as well as vibration of the optical path of the light transmitting in the fiber Therefore, the φ-OTDR can detect the liquid level with a large sensing range by interrogating the phase information along the fiber due to the temperature difference between the liquid and air Further, the scattering enhanced optical fiber (SEOF) is used as the sensing fiber to improve signal to noise ratio (SNR) of the phase signal Moreover, a high sensitivity liquid level sensing head by wrapping the SEOF on a heat conductive cylinder is design and optimized to improve the sensing resolution Proposed liquid level sensor present a sensing resolution of 132 μm and a potential sensing range larger than 320 m

Smart Television Edge Server Deployment Scheme Based on CDN

Abstract: Cable TV resource allocation is a valuable research issue to current television industry However, the traditional one to multiple resource allocation model cannot meet the development requirements of China’s smart television In this paper, we propose a smart television edge server deployment scheme based on Content Delivery Network (CDN), including two models: the Distance and Population based method (DAP) and the Dijkstra’s shortest path based method (DSP) The DAP method is based on the effect of distance and population on edge server deployment The DSP model is based on the specific path of the Dijkstra algorithm The numerical experiments are illustrated through the data from Beijing, China and prove the effectiveness of our models

Optimization of Wide-Spectral Mode-Locking Fiber Laser Carriers in Free Space Optical Communication

Abstract: We experimentally demonstrate an optimization of wide-spectral mode-locking fiber laser carriers by the introduction of continuous light stable pulse scheme with fiber dispersion management. The bandwidth of the optimized supercontinuum and the fluctuation between pulses have been significantly improved, and the fluctuation between pulses has been reduced by nearly 60%. After 1km of simulated atmospheric transmission, it is found that the optimized eye diagram has obvious eye opening. The signal to noise ratio increased by 6.84dB. And this optimization scheme is easy to integrate into the communication light source. It can be an adaptive scheme for free space optical communication in the foreseeable future.

Investigation on Nyquist-WDM Channels with Optical Sinc-shaped Pulses Based on Electro-Optic Frequency Combs

Abstract: The practicability of Nyquist wavelength-division-multiplexed (WDM) channels with optical Sinc-shaped pulses is greatly limited by phase noise and power flatness of the WDM sources. We insert uniform and coherent lines around each line of flat electro-optic frequency combs to generate 5 Nyquist-WDM channels with high quality. The normalized root-mean-square error (NRMSE) of optical Sinc-shaped pulses for the 5 channels can be between 1.23%–1.96%. The optical signal-to-noise ratios (OSNRs) of optical Nyquist pulses have a 7 dB improvement by reducing optical linewidth of electro-optic combs. Further, the high OSNR of Nyquist-WDM channels can bring benefits to transmission performance in high spectral efficiency Nyquist-WDM communication system.

IoT Device Auto-Tagging Using Transformers

Abstract: The IoT platform can identify a device matching the data only when securing the information on a large amount of IoT devices in advance. However, in a situation where a lot of companies release a variety type of IoT, it is not easy to retain information about all IoT devices. That is, it is a difficult situation where the current IoT platform has difficulty in analyzing unstandardized tagging information. Therefore, this paper provides a technique of ‘Tagging’ that automatically identifies the device based on information collected from IoT devices or sensor devices. This technique was developed based on the natural language processing algorithm based on the attention mechanism among the machine learning models.

Research and Implementation of Dual-stack Web Service Architecture Based on Intelligent DNS and Reverse Proxy Technology

Abstract: In order to reduce the implementation cost and difficulty of providing WEB service based on IPv6 protocol, and to take full advantage of multi-export links, this paper designs a WEB service framework supporting IPv4/IPv6 by combining intelligent DNS and reverse proxy technology aiming at the environment of multi-export links in university network Based on open source software BIND and Nginx, this paper proposes and implements a WEB service scheme combining intelligent DNS and reverse proxy technology Practical application shows that the framework we proposed is capable of improving the speed of users accessing campus network WEB services in the access link network, reverse proxy effectively reduces the load of WEB service The framework is based on open source architecture, which can be implemented without the modification of existing networks and WEB applications On the basis of reducing the implementation cost, the advantage of multi export links can be exploited The scheme can be deployed in other universities or similar network environments only by modifying configuration files

Comparison of Ultra-Low-Loss G.652B Fiber and G.652D Fiber Sensing Links with Second-Order Raman forward Pumping

Abstract: How to extend the repeaterless transmission/sensing distance is the main demand for power grid as higher requirements are proposed for the optical transmission/sensing system. Although many technologies have been proposed to extend the working distance, most of previous works used G.652D fiber. In this paper, the performance differences between ultra-low-loss G.652B fiber and G.652D fiber sensing links with second-order Raman forward pumping are numerically compared, all with 150km length. The results show that the transmission distance of the signal light could be longer in ultra-low-loss G.652B fiber than G.652D fiber and better optical signal-to-noise ratio will be obtained in ultra-low-loss G.652B fiber since the amplified signal light power distribution is more uniform. It could conclude that ultra-low-loss G.652B fiber provides better C-band signal amplification and transmission.

Learning Based Media Access Control for UAV Ad-Hoc Networks

Abstract: With the rapid development of unmanned aerial vehicle (UAV) technology and the drastic growth of multi-UAV collaborative applications, research on UAV Ad-Hoc Network (UAVNET) has received extensive attention in recent years. Due to the decentralization nature and being prone to failure and high dynamics, the performance of traditional contention-based carrier sense multiple access (CSMA) protocols and variants is far from satisfaction for UAVNETs. In this paper, we propose an intelligent access mechanism based on actor-critic (AC) algorithm, called AC-CSMA, with aim of achieving low access collision rate and high throughput for UAVNETs in dynamic environments. The UAVs are modeled as decision-making agents without priori information about the network (e.g., the number of nodes, link conditions, and other agents’ strategies). Individual UAV agents learn their optimal strategy by using the historical sensory information including the number of collisions or successful transmissions. Numerical results show that the proposed AC-CSMA mechanism significantly outperforms traditional access mechanism in terms of collision rate and throughput for UAVNETs without compromising access fairness.