With the popularisation of convolutional neural networks (CNNs), many tools were developed to facilitate the development of models. One of the main tools developed is the Keras API, which is a high-level API for creating and training models. The Keras API acts as a high-level layer to facilitate the implementation of neural networks while other CNN libraries can run as backend. There is a demand for studies that compare the performance of a CNN implemented with the Keras API and the performance of CNNs implemented only with native methods to CNN libraries. To evaluate the impact of the Keras API on the performance of TensorFlow 1, TensorFlow 2 and CNTK libraries in relation to the execution time in GPUs, an experiment was conducted for measuring the execution time of the LeNet-5 model. It was implemented with the aid of the Keras API and implemented without it. Models implemented with Keras API methods showed longer execution times than models implemented only with native methods from CNTK libraries. The performance difference can be significant for some applications. The Keras API sacrifices library performance up to four times when compared to native methods.
A paradigm shift in machine learning (ML) application models has occurred in the preceding years due to privacy and deep learning aspirations. The recently created decentralised paradigm of ML is known as federated learning. Federated learning (FL) is a ML technique in which many dispersed nodes use their locally stored data to train a common prediction model. Better data privacy is possible because the training data is not routed to a central server. FL offloads the processed information to the server and does not need clients to share their personal information. However, FL is a new field that has yet to achieve mainstream acceptance and is still in the development phase. In this context, the purpose of our study is to provide a more comprehensive overview of the most important protocols, platforms, and real-world FL use cases, which will allow researchers to develop privacy-preserving solutions for businesses that require FL.
Chaos-based cryptosystems and their behaviour in cryptography attract many scientists' and researchers' attention in physics and computer science in recent decades. A new dynamic cryptosystem for video sequence based on the combination of chaotic Lorenz map and dual hash functions is proposed in this paper to improve the security level of the video applications. First, split up the video sequence into video frames and audio samples. Then, the initial conditions for both video and audio are generated using SHA-256 and MD5, respectively. Next, a higher dimensional Lorenz chaotic system is adopted to confuse and diffuse the audio samples and video frames components. Moreover, to improve the security level of the proposed cryptosystem against different types of cryptanalytic attacks, the concept of multi-key is employed. The security analysis is conducted and the results indicate that the presented cryptosystem satisfies the security requirements against various attacks and can be directly applied in the real life video applications.
Blockchain provides more security and decentralised systems. However, it is difficult for startups and small companies to build and deploy their blockchain networks. So, cloud computing providers make it easy for those to work with blockchain using their networks and cloud infrastructure through using blockchain as a service platform. There are many blockchain as a service providers and platforms. This survey provides an overview of 23 blockchain-as-a-service platforms and provides a comparison based on provider, blockchain protocol, blockchain type, cost, security, and support of smart contract.
Wireless sensor networks (WSN) detect and monitor various physical attributes from the environment, convert the observation into information and send it wirelessly to base station/sink. In most of applications of WSNs, localisation, that is, finding location of nodes sending information, is also required to be done for better inferencing. In this paper, we present a review of existing literature related to WSN node localisation problem. Firstly, an introduction on wireless sensor networks and its challenges are mentioned. Then, issues of localisation algorithms are discussed. The paper also discusses different localisation algorithms and their categorisation based on precision/hardware requirement, localisation algorithms can be categorised into two groups such as range-based and range-free algorithms. Various range-based and range-free algorithms with their merits and demerits are summarised and a list of research gaps is discussed.
This paper focuses on infinite impulse response (IIR) system identification which uses a recent nature inspired algorithm called antlion optimisation (ALO). The system identification problem is concerned with determining the viable parameters by minimising the cost function. Generally, gradient-based techniques are mostly used for IIR system identification. However, these traditional algorithms face the problem of getting trapped in local solution. So to get rid of this problem, a novel ALO algorithm is used for IIR system identification. The ALO is inspired by the preying process of antlions on the ants. The algorithm is free of the issues faced by the traditional techniques. The performance of ALO algorithm is measured using two measures mean square error (MSE) which is taken as cost function and the convergence profile. The results obtained using ALO are compared with those of the particle swarm optimisation (PSO) algorithm and cat swarm optimisation (CSO) algorithm. The obtained results confirmed that the algorithm surpasses the performance of the existing algorithms.
With the dawn of the 21st century and the growth of fast, always available and low power networks, wireless sensor networks are being implemented in diverse use-cases. Wireless sensor networks are being deployed to observe, explore and control the physical world. Wireless sensor networks are generally deployed in dynamic environments. Sensor networks utilise machine learning techniques to avoid the unnecessary redesign of a wireless sensor network deployment for adapting to changing requirements. Machine learning is also used to maximise the security, efficiency, lifetime, and resource utilisation in such networks. In this paper, we present an extensive literature survey of various machine learning applications that are used or are in research to address the operational and non-operational challenges in wireless sensor networks.
Advanced metering infrastructure (AMI) is an integral part of smart grid network that involves transmission of finest grain operational data of consumers' load profiles on wireless facilities. Unauthorised access to these data due to the vulnerabilities of wireless facilities may easily disrupt smart grid. Some of the existing schemes only focus on specific attack(s), leaving other attacks, and most of them cannot be used to secure energy profiles from different time of use without increasing communication overhead. In this paper, a provably lightweight security scheme to secure metering and information exchange and consumers' privacies irrespective of the attacking points and nature of the attacks is proposed for AMI. Our scheme ensures data security and cooperative aggregation among m number of consumers of the same service provider. The security analysis and performance evaluation of the scheme are also presented. The results illustrate that the approach secures metering at a low overhead.
The use of microservices is a new trend in software engineering, dividing an application into several services. This concept allows programmers to write each microservice code using the better language and framework they know. We perceive that research on microservices aims mainly at composability, portability, and interface, leaving uncovered in surveys relevant quality concerns. Therefore, this article reports a survey focused on providing classification and analysis of studies on evaluating and improving performance in microservice-based applications. Our contributions are threefold: 1) an in-depth analysis of state of the art on microservices through the lens of performance; 2) a novel taxonomy to reclassify the current microservice initiatives, looking at software and hardware aspects that interfere in the execution of the application; 3) an analysis of trends and open research opportunities in the joint combination of performance and scalability applied to microservices. The article supports developers and organisations in defining standards, strategies, and technologies to model and code microservices applications by presenting practical and theoretical issues.
Effective data management is very challenging to cloud providers, whose business model relies on maintaining an economic profit while satisfying the tenants' performance requirements. To address these challenges, many data replication strategies have been proposed. In this paper, we propose a new dynamic data replication strategy for cloud systems called RCPP 1 . In order to satisfy performance requirements, the proposed strategy exploits the valuable knowledge extracted from the tenants' past access history. Therefore, it uses the mathematical triadic concept analysis approach to determine correlated data to be replicated. Furthermore, the cloud provider's profit is taken into account. Hence, an economic model is proposed to estimate the revenues and expenditures of the provider. Experimental studies show the efficiency and effectiveness of RCPP compared to state-of-the-art strategies. RCPP is indeed proven able to reduce the total expenditures of the cloud provider significantly while achieving better performances.
Mobile ad-hoc network (MANET) is a collection of mobile terminals forming an infrastructure-less and quickly deployable network, in which nodes can communicate to each other via multiple hops. Mobility attribute is a notable one in MANET, as this leads to frequent topology changes, so this is the primary cause of link failure. Link failure time estimation has always remained an active area of research among researchers of the networking community. This paper explores various link failure time prediction techniques and proposes a novel least-squares polynomial regression-based statistical technique to estimate the link failure time in MANET. Each node in the network periodically broadcasts hello packets to register its presence with neighbouring nodes. A neighbour node receives these packets and uses its signal strength to estimate the link failure time with help of quadratic least-squares regression. The outlined technique is simulated using Network Simulator 2.35. The performance of the estimation accuracy of the suggested technique and existing interpolation-based technique has been examined for numerous mobility and scalability scenarios.
Replication protocol (RPCL) has been the major focused area for research in the replicated distributed real-time database system (RDRTDBS). The primary objective of RPCL is to improve the availability, scalability, and fault tolerance of the system. However, the most challenging task is to guarantee QoS for the data services in the RDRTDBS. The real-time transaction workload may not be balanced, and their access patterns may be time-varying and skewed. As a result, many RTTs miss their deadlines, or consistency constraints of the real-time data items may be violated. In the current paper, our objective is to propose an integrated version of a transaction manager that consists of the transaction sub-module and granule sub-module, thread to granule policy, and overload resolver (OLRE) to guarantee quality of service (QoS) during unpredictable workload. The experimental results show that our proposed algorithms reduce the overload condition and improve the processing of real-time transactions (RTTs) with maintaining mutual consistency.
The deployment of the WSN provides a non-uniform distribution in the surveyed area. This we offer low-density clusters and high-density clusters. Clusters that cover a limited number of nodes are quickly exhausted. This due to the load occupied by these nodes throughout the network operation. Many times these nodes will be elected as a CH, they consume more energy. Faced with this unbalanced repartition between clusters, we proposed a routing protocol 'coverage protocol to connected nodes' (CPCN) to improve the connectivity between nodes guarantee full coverage and ensure quality of service (QoS of the WSNs). Our proposed CPCN protocol is compared to other MCSL, TQCML-EM and HESL protocols with the TOSSIM simulator was an effective solution to extend network life by minimising power consumption.
Sensors are lightweight platforms, with limited computing power and energy reserves. However, our article is all about creating a secure routing protocol that provides reliable communications. The crypto-ECC contributions seek to accelerate the computation of scalar multiplications by using the paralleling technique which consists of distributing the calculation into several independent tasks that can be processed simultaneously by different nodes. This submission tries to design a secure multicast new routing protocol crypto-ECC that allows the generation and management of cryptographic keys and that takes into account the constraints of the wireless sensor networks and internet of things. Finally, the proposed solution will be evaluated using Telosb sensors, in comparison with earlier work. Performance evaluation using the TOSSIM simulator shows that crypto-ECC is scalable, offers secure links, and provides excellent key connectivity.
Speaker verification can be viewed as a process of verifying the person using his/her utterance. The major challenge to implement automatic speaker verification in security applications is spoofing attacks. Speaker verification systems can be spoofed using pre-recorded speech, synthetic and voice conversion speech. Hence, there is a need to develop spoof detection system in order to make voice biometrics viable for security applications. This paper proposes to explore time-frequency representations obtained using gammatone filterbank and constant Q transform for detecting presentation attack for automatic speaker verification. The experiments are carried out for ASV spoof 2017 database and the results are compared with state-of-art replay speech detection systems based on cepstral features.
Full reference video quality assessment based on optical flow is emerging. Human Visual System (HVS) based video quality assessment algorithms are playing an important role in effectively assessing the distortions in video sequences. There exist very few video quality assessment algorithms which consider spatio-temporal distortions effectively. To address the above issues, we present an enhanced optical flow based full reference video quality algorithm which considers the orientation feature of the optical flow while computing the temporal distortions as opposed to the use of feature, minimum eigenvalue as in the state of the art. Further, it presents an interquartile range based comparative weighted closeness (INT-CWC) measure which aimed to measure the comparative dispersion of video quality scores of any two video quality assessment algorithms with DMOS scores. Here INT-CWC measure is a novel attempt. The performance of proposed scheme is evaluated using the LIVE dataset and scheme is shown to be competitive with, and even out-perform, existing video quality assessment algorithms.
Psychological stress detection continues to remain a large problem among individuals. Identifying and combating stress before letting it take the face of some severe problems is of utmost importance. Traditional psychological stress detection techniques need professional devices and specialists to analyse the data, so it is very important that a method has to be introduced in which one can automatically detect the stress state of the user. In this work, we have made an effort to detect stress from the tweets of the users. We have collected different stressed and non-stressed related tweets from Twitter. Then, we have applied latent Dirichlet allocation (LDA), a popular machine learning algorithm, to detect stress among the individuals from their tweets and categorised the tweets into two classes stressed and non-stressed. We have also found experimentally that our LDA-based system performs better than the SVM-based system.
One of the most overwhelming and exhilarating advancements in the field of intelligent systems has been the introduction of visual question answering (VQA), a new and exciting problem that makes use of natural language processing and computer vision for measuring the ability of a system for image understanding and generating inferences beyond object recognition, segmentation and image captioning. One of the primary goals of research in visual-based artificial intelligence is to design systems that can understand and reply to questions about visual data. The first part of the study very vividly presents numerous diversified as well as diagnostic datasets testing different visual reasoning abilities, while the second part of the study details a plethora of the recently developed approaches for VQA. Finally, a qualitative comparison of the diversity present in the various approaches has been done that can serve as an important benchmark for analysing and comparing the different characteristics and applications of these algorithms and datasets that can prove to be helpful to someone new to the field of VQA and its community.
The extensive use of computational power from data centres causes huge energy consumption. A good virtual machine (VM) placement strategy would make better consolidation of VMs in a data centre (DC) that reduces energy consumption. However, it is hard to balance hosts in DCs due to the workload fluctuation by application and scaling of VMs. The optimal decision on VM placement and consolidation is an NP-hard problem and many researchers have proposed solutions to tackle this problem but they lack efficient exploitation of the mechanisms. Therefore, this paper proposes a hierarchical cluster-based approach with a meta-heuristic crow search algorithm (CSA) for the optimal selection of hosts to place the VMs and consolidate the maximum number of VMs on a minimum number of hosts. The work is simulated in CloudSim using real workload traces. Experimental results show that proposed work reduces energy consumption, SLA violations and VM migrations while ensuring better resource utilisation.
The imbalanced category of network traffic poses a challenge to the classification methods based on machine learning, because the unbalanced data structure affects the performance of machine learning algorithms. In this paper, we propose a multi-model coupling approach to address the imbalanced data problem in network traffic classification. We process the major class to some clusters by a clustering algorithm. Then, these clusters and the minor class are used to form the training dataset for training model respectively. During the test, the test dataset is input into the previously trained models respectively, and the classification results of respective models are coupled to obtain the final result. We tested our proposed method on two well-known network traffic datasets and the results showed that it could achieve better performance and less time consumption compared with recent proposed methods in the case where the ratio of minor to major classes is very small.
