Although authentication of users of digital voice-based systems has been addressed by much research and many commercially available products, there are very few that perform well in terms of both usability and security in the audio domain. In addition, the use of voice biometrics has been shown to have limitations and relatively poor performance when compared to other authentication methods. We propose using audio steganography as a method of placing authentication key material into sound, such that an authentication factor can be achieved within an audio channel to supplement other methods, thus providing a multi factor authentication opportunity that retains the usability associated with voice channels. In this research we outline the challenges and threats to audio and voice-based systems in the form of an original threat model focusing on audio and voice-based systems, we outline a novel architectural model that utilises audio steganography to mitigate the threats in various authentication scenarios and finally, we conduct experimentation into hiding authentication materials into an audible sound. The experimentation focused on creating and testing a new steganographic technique which is robust to noise, resilient to steganalysis and has sufficient capacity to hold cryptographic material such as a 2048 bit RSA key in a short audio music clip of just a few seconds achieving a signal to noise ratio of over 70 dB in some scenarios. The method developed was seen to be very robust using digital transmission which has applications beyond this research. With acoustic transmission, despite the progress demonstrated in this research some challenges remain to ensure the approach achieves its full potential in noisy real-world applications and therefore the future research direction required is outlined and discussed.

UASN (Underwater Acoustic Sensor Network) has intrinsic impediments, since it is utilized and utilizes acoustic signs to impart in the sea-going world. Examples include long delays in propagation, limited bandwidth, high transmitting energy costs, very high attenuation in the signal, expensive implementation and battery replacement etc. The UASN routing schemes must therefore take account of these features to achieve balance energy, prevent void hole and boost network life. One of the significant issue in routing is the presence of void node. A void node is a node that does not have any forwarder node. The presence of void may cause the bundle conveyance in the steering time which prompts information misfortune. The gap during steering influences the network execution regarding proliferation delay, vitality utilization and network lifetime, and so forth. So with the objective to remove the void node in the networks, this work presents an energy efficient optimal path routing for void avoidance in underwater acoustic sensor networks. This work uses the concept of gray wolf optimization algorithm to calculate the fitness function and that fitness function is used to select the best forwarder node in the networks. This work only consider the vertical directions which further reduces the end to end delay. The proposed work has been simulated on MATLAB and performances are evaluated in terms of broadcast copies of data, energy tax, and packet delivery ratio, number of dead nodes, network lifetime and delay.

Distributed Denial of Service (DDoS) is an ever-changing type of attack in cybersecurity, especially with the growing demand for cloud and web services raising a never-ending challenge in the lucrative business. DDoS attacks disrupt users' access to the targeted online services leading to significant business loss. This article presents a three-level architecture for detecting DDoS attacks at the application layer. The first level is responsible for selecting the best features of the samples and classifying the traffic into either benign or malicious, then the second level consists of a hard voting classifier to identify the type of the DDoS source: UDP, TCP, or Mixed-based. Finally, the last level aligns the attack to the appropriate DDoS type. This approach is validated using the CIC-DDoS2019 dataset, and the time, accuracy score, and precision are used as the model performance metrics. Compared to the existing machine learning (ML) approaches, the proposed architecture reveals substantial improvements in both binary and multiclass classification of application-layer DDoS attacks.

Many applications of mobile ad-hoc networks like conferencing, handling emergency situations, military operations require the multicast routing. Moreover, in such applications there is a demand for multimedia traffic such as audio/video calls or audio/video conferencing. For mobile ad-hoc environments, it is accepted that the on-demand reactive routing protocol AODV has become default. Moreover, to get the benefits of using a single protocol for both unicast and multicast routing, in this work, the multicast routing protocol MAODV (multicast extension of AODV) has been considered and its performance is observed for CBR, VoIP and video data traffics. Since to accommodate multimedia traffic, a routing protocol demands for stringent QoS requirements in terms of delay, jitter and packet losses; the performance of the protocol is measured in terms of QoS performance metrics such as average delay, average jitter and packet delivery ratio. Further, a modified version of MAODV (called M-MAODV) is taken and its performance is also evaluated for multimedia traffic. A fair comparison of MAODV and M-MAODV protocols is achieved through the use of same network conditions for the evaluation. From the results, the improved values of delay, jitter and packet delivery ratio have been observed for M-MAODV irrespective of node speeds and for all data traffic types.

In this work, adaptive learning of a monitored real-time stochastic phenomenon over an operational LTE broadband radio network interface is proposed using cascade forward neural network (CFNN) model. The optimal architecture of the model has been implemented computationally in the input and hidden units by means of incremental search process. Particularly, we have applied the proposed adaptive-based cascaded forward neural network model for realistic learning of practical signal data taken from an operational LTE cellular network. The performance of the adaptive learning model is compared with a benchmark feedforward neural network model (FFNN) using a number of measured stochastic SINR datasets obtained over a period of three months at two indoors and outdoors locations of the LTE network. The results showed that proposed CFNN model provided the best adaptive learning performance (0.9310 RMSE; 0.8669 MSE; 0.5210 MAE; 0.9311 R), compared to the benchmark FFNN model (1.0566 RMSE; 1.1164 MSE; 0.5568 MAE; 0.9131 R) in the first studied outdoor location. Similar robust performances were attained for the proposed CFNN model in other locations, thus indicating that it is superior to FFNN model for adaptive learning of real-time stochastic phenomenon.

The rapid increase in the number of mobile phones and IoT devices connected to the network reduces the bandwidth of the Internet communication channel, and as a result, delays occur in the delivery of data processed in remote clouds. Edge computing systems (cloudlet, fog computing, etc.) are used to eliminate resource shortages, energy consumption, and communication channel delays in mobile devices. Edge computing systems place processing devices (computers) close to users. Cloudlet-based mobile cloud computing is widely used to reduce delays in communication channels and energy consumption in mobile devices. Selection of the most suitable cloudlet allowing users to run applications fast in cloud is still a considerable problem. This paper proposes a strategy for the selection of high-performance cloudlets providing fast solutions, considering the complexity of application (file type). It offers a method for cloudlet selection out of large number of cloudlets with different technical capabilities providing faster processing of user application. The timing of user applications in cloudlets with different technical capabilities (operating frequency, number of cores, volume of RAM, etc.) also varies. The proposed method provides faster solution for the user application. User applications are grouped by type of application, and a set of cloudlets are clustered by the number of groups. Clustering is performed first by the parameters corresponding to the operating frequency of the cloudlets, then by the number of cores and the volume of RAM. The proposed method reduces energy consumption of mobile devices by providing faster processing of applications. Thus, the proposed strategy provides an energy consumption reduction on mobile devices, faster processing of results and decrease of network delays.

Mobile Ad-hoc Network (MANET) data transfer between nodes in a multi-hop way offers a wide variety of applications. The dynamic feature of ad hoc network mobile nodes is primarily influenced by safety issues, which limit data forwarding rate in multipath routing. As a supplementary method to improve safe data delivery in a MANET, this paper propose and analyse the cluster head (CH) selection and optimum multipath scheme. The CHs are chosen based on the possibility values of each node in MANET, which are considered from the residual energy of each node. During the present phase, the total remaining node energy is used to calculate the mean energy of the entire network. The most likely nodes are picked as the CH, which gathers packets from the cluster members through multi-hop communication. The fundamental idea is to partition a top-secret communication into several shares and then forward the shares via numerous routes to the destination. The Coral Reef Optimization method is used in this work to perform optimum multipath routing. The thorough simulation findings validate the feasibility and efficacy of the suggested strategy in comparison to Butterfly optimization algorithm (BA), Whale Optimization algorithm (WOA) and BAT algorithm techniques.