Federated learning (FL) is a decentralized approach for collaborative model training on edge devices. This distributed method of model training offers advantages in privacy, security, regulatory compliance, and cost efficiency. Our emphasis in this research lies in addressing statistical complexity in FL, especially when the data stored locally across devices is not identically and independently distributed (non-IID). We have observed an accuracy reduction of up to approximately 10%–30%, particularly in skewed scenarios where each edge device trains with only 1 class of data. This reduction is attributed to weight divergence, quantified using the Euclidean distance between device-level class distributions and the population distribution, resulting in a bias term$(\delta _{k})$. As a solution, we present a method to improve convergence in FL by creating a global subset of data on the server and dynamically distributing it across devices using a dynamic data queue-driven FL (DDFL). Next, we leverage Data Entropy metrics to observe the process during each training round and enable reasonable device selection for aggregation. Furthermore, we provide a convergence analysis of our proposed DDFL to justify their viability in practical FL scenarios, aiming for better device selection, a non-suboptimal global model, and faster convergence. We observe that our approach results in a substantial accuracy boost of approximately 5% for the MNIST dataset, around 18% for CIFAR-10, and 20% for CIFAR-100 with a 10% global subset of data, outperforming the state-of-the-art (SOTA) aggregation algorithms.

As the world calls it, data is the new oil. With vast installments of Industrial Internet of Things (IIoT) infrastructure, data is produced at a rate like never before. Similarly, artificial intelligence (AI) and machine learning (ML) solutions are getting integrated to numerous services, making them “smarter.” However, the data remain fragmented in individual organizational silos inhibiting data value extraction to its full potential. Digital marketplaces are emerging to allow data owners to monetize these data. Yet concerns, such as privacy, security, and unfair payment settlement deter adoption of such platforms. In addition, the state-of-the-art platforms are under the control of large multinational corporations with no transparency between the buyer and seller in terms of payment details, listing, data discovery, and storage. In this work, a novel decentralized platform of a digital data marketplace for IoT data has been proposed. The platform leverages a decentralized data streaming network to host IoT data in a reliable and fault-tolerant manner. The platform ensures fair trading, data storage, and delivery in a privacy-preserving manner and trust metric calculation for actors in the network. In order to study the feasibility of the proposed platform, an opensource library is developed using Hyperledger Fabric and data network layer built on VerneMQ, the library is deployed on a real-time Google cloud platform. The library is tested and results are analyzed for throughput, overheads, and scalability.

Privacy-preserving (PP) applications allow users to perform online daily actions without leaking sensitive information. The PP scalar product (PPSP) is one of the critical algorithms in many private applications. The state-of-the-art PPSP schemes use either computationally intensive homomorphic (public-key) encryption techniques, such as the Paillier encryption to achieve strong security (i.e., 128 b) or random masking technique to achieve high efficiency for low security. In this article, lattice structures have been exploited to develop an efficient PP system. The proposed scheme is not only efficient in computation as compared to the state-of-the-art but also provides a high degree of security against quantum attacks. Rigorous security and privacy analyses of the proposed scheme have been provided along with a concrete set of parameters to achieve 128-b and 256-b security. Performance analysis shows that the scheme is at least five orders faster than the Paillier schemes and at least twice as faster than the existing randomization technique at 128-b security. Also the proposed scheme requires six-time fewer data compared to the Paillier and randomization-based schemes for communications.

The proliferation of IoT in various technological realms has resulted in the massive spurt of unsecured data. The use of complex security mechanisms for securing these data is highly restricted owing to the low-power and low-resource nature of most of the IoT devices, especially at the Edge. In this article, we propose to use blockchains for extending security to such IoT implementations. We deploy a Ethereum blockchain consisting of both regular and constrained devices connecting to the blockchain through wired and wireless heterogeneous networks. We additionally implement a secure and encrypted networked clock mechanism to synchronize the non-real-time IoT Edge nodes within the blockchain. Further, we experimentally study the feasibility of such a deployment and the bottlenecks associated with it by running necessary cryptographic operations for blockchains in IoT devices. We study the effects of network latency, increase in constrained blockchain nodes, data size, Ether, and blockchain node mobility during transaction and mining of data within our deployed blockchain. This study serves as a guideline for designing secured solutions for IoT implementations under various operating conditions such as those encountered for static IoT nodes and mobile IoT devices.

In this paper, we provide an efficient and easy-to-implement symmetric searchable encryption scheme (SSE) for string search, which takes one round of communication, O(n) times of computations over n documents. Unlike previous schemes, we use hash-chaining instead of chain of encryption operations for index generation, which makes it suitable for lightweight applications. Unlike the previous SSE schemes for string search, with our scheme, server learns nothing about the frequency and the relative positions of the words being searched except what it can learn from the history. We are the first to propose probabilistic trapdoors in SSE for string search. We provide concrete proof of non-adaptive security of our scheme against honest-but-curious server based on the definitions of [12]. We also introduce a new notion of search pattern privacy, which gives a measure of security against the leakage from trapdoor. We have shown that our scheme is secure under search pattern indistinguishability definition. We show why SSE scheme for string search cannot attain adaptive indistinguishability criteria as mentioned in [12]. We also propose modifications of our scheme so that the scheme can be used against active adversaries at the cost of more rounds of communications and memory space. We validate our scheme against two different commercial datasets (see [1], [2]).

In this paper, we investigate energy harvesting capabilities of non-orthogonal multiple access (NOMA) scheme integrated with the conventional time division multiple access (TDMA) scheme, which is referred to as hybrid TDMA-NOMA system. In a such hybrid scheme, users are divided into a number of groups, with the total time allocated for transmission is shared between these groups through multiple time slots. In particular, a time slot is assigned to serve each group, whereas the users in the corresponding group are served based on power-domain NOMA technique. Furthermore, simultaneous wireless power and information transfer technique is utilized to simultaneously harvest energy and decode information at each user. Therefore, each user splits the received signal into two parts, namely, energy harvesting part and information decoding part. In particular, we jointly determine the power allocation and power splitting ratios for all users to minimize the transmit power under minimum rate and minimum energy harvesting requirements at each user. Furthermore, this joint design is a non-convex problem in nature. Hence, we employ successive interference cancellation to overcome these non- convexity issues and determine the design parameters (i.e., the power allocations and the power splitting ratios). In simulation results, we demonstrate the performance of the proposed hybrid TDMA-NOMA design and show that it outperforms the conventional TDMA scheme in terms of transmit power consumption.

This paper provides some preliminary results for a possible novel side channel attack on Android smart phones. This attack collects accelerometer readings when users type on soft keyboards. The work has the following two parts: 1) differentiate users based on sensor readings and 2) identify whether the user belongs to a particular nationality i.e., Chinese in this work. This work uses a novel signal processing technique along with random forest machine learning algorithm to extract unique features belong to Chinese nationalities. We collected more than 2000 keystrokes data from six users where three of them are Chinese nationals. Our model has correctly identified 86% of the sensor data to classify Chinese nationality. Since any apps installed on Android device can listen to the accelerometer sensor data, the side channel attack presented in this work demonstrates another potential privacy vulnerability which could be exploited by malicious apps for targeted activities such as advertisements.

This paper introduces an efficient algorithm to develop a privacy-preserving voice verification based on iVector and linear discriminant analysis techniques. This research considers a scenario in which users enrol their voice biometric to access different services (i.e., banking). Once enrolment is completed, users can verify themselves using their voice print instead of alphanumeric passwords. Since a voice print is unique for everyone, storing it with a third-party server raises several privacy concerns. To address this challenge, this paper proposes a novel technique based on randomization to carry out voice authentication, which allows the user to enrol and verify their voice in the randomized domain. To achieve this, the iVector-based voice verification technique has been redesigned to work on the randomized domain. The proposed algorithm is validated using a well-known speech dataset. The proposed algorithm neither compromises the authentication accuracy nor adds additional complexity due to the randomization operations.

This paper proposes an efficient algorithm to perform privacy-preserving (PP) facial expression classification (FEC) in the client-server model. The server holds a database and offers the classification service to the clients. The client uses the service to classify the facial expression (FaE) of subject. It should be noted that the client and server are mutually untrusted parties and they want to perform the classification without revealing their inputs to each other. In contrast to the existing works, which rely on computationally expensive cryptographic operations, this paper proposes a lightweight algorithm based on the randomization technique. The proposed algorithm is validated using the widely used JAFFE and MUG FaE databases. Experimental results demonstrate that the proposed algorithm does not degrade the performance compared to existing works. However, it preserves the privacy of inputs while improving the computational complexity by$120$times and communication complexity by$31$percent against the existing homomorphic cryptography based approach.

Smart devices capture users' activity such as unlock failures, application usage, location and proximity of devices in and around their surrounding environment. This activity information varies between users and can be used as digital fingerprints of the users' behaviour. Traditionally, users are authenticated to access restricted data using long term static attributes such as password and roles. In this paper, in order to allow secure and seamless data access in mobile environment, we combine both the user behaviour captured by the smart device and the static attributes to develop a novel access control technique. Security and performance analyses show that the proposed scheme substantially reduces the computational complexity while enhances the security compared to the conventional schemes.

Decentralized attribute-based encryption (ABE) is a variant of multi-authority based ABE whereby any attribute authority (AA) can independently join and leave the system without collaborating with the existing AAs. In this paper, we propose a user collusion avoidance scheme which preserves the user's privacy when they interact with multiple authorities to obtain decryption credentials. The proposed scheme mitigates the well-known user collusion security vulnerability found in previous schemes. We show that our scheme relies on the standard complexity assumption (decisional bilienar Deffie-Hellman assumption). This is contrast to previous schemes which relies on non-standard assumption (q-decisional Diffie-Hellman inversion).

The emerging cloud market introduces a multitude of cloud service providers, making it difficult for consumers to select providers who are likely to be a low risk from a security perspective. Recently, significant emphasis has arisen on the need to specify Service Level Agreements that address security concerns of consumers (referred to as SecSLAs) -- these are intended to clarify security support in addition to Quality of Service characteristics associated with services. It has been found that such SecSLAs are not consistent among providers, even though they offer services with similar functionality. However, measuring security service levels and the associated risk plays an important role when choosing a cloud provider. Data breaches have been identified as a high priority threat influencing the adoption of cloud computing. This paper proposes a general analysis framework which can compute risk associated with data breaches based on pre-agreed SecSLAs for different cloud providers. The framework exploits a tree based structure to identify possible attack scenarios that can lead to data breaches in the cloud and a means of assessing the use of potential mitigation strategies to reduce such breaches.

Recent trend towards cloud computing and outsourcing has led to the requirement for face recognition (FR) to be performed remotely by third-party servers. When outsourcing the FR, client's test image and classification result will be revealed to the servers. Within this context, we propose a novel privacy-preserving (PP) FR algorithm based on randomization. Existing PP FR algorithms are based on homomorphic encryption (HE) which requires higher computational power and communication bandwidth. Since we use randomization, the proposed algorithm outperforms the HE based algorithm in terms of computational and communication complexity. We validated our algorithm using popular ORL database. Experimental results demonstrate that accuracy of the proposed algorithm is the same as the accuracy of existing algorithms, while improving the computational efficiency by 120 times and communication complexity by 2.5 times against the existing HE based approach.

The authors propose a coordinated multi‐cell beamforming technique for signal‐to‐interference plus noise ratio (SINR) balancing under multiple base station power constraints. Instead of balancing SINR of all users in all cells to the same level, the authors’ propose a new approach to balance SINR of users in various cells to different maximum possible values. This has the ability to allow users in cells with relatively more transmit power or better channel condition to achieve a higher balanced SINR than that achieved by users in the worst‐case cells. This multi‐level SINR balancing problem is solved using SINR constraints based SINR balancing criterion and subgradient method. The simulation results support the optimality of the results through comparison with semi‐definite programming‐based optimisation.

Technology advancements in smart mobile devices empower mobile users by enhancing mobility, customizability and adaptability of computing environments. Mobile devices are now intelligent enough to capture dynamic attributes such as unlock failures, application usage, location and proximity of devices in and around its surrounding environment. Different users will have different set of values for these dynamic attributes. In traditional attribute based access control, users are authenticated to access restricted data using long term static attributes such as password, roles, and physical location. In this paper, in order to allow secure data access in mobile environment, we securely combine both the dynamic and static attributes and develop novel access control technique. Security and performance analyse show that the proposed scheme substantially reduces the computational complexity while enhances the security compare to the conventional schemes.

In cloud computing, measurable services such as packet loss and memory are quantized into different levels to provide different level of services to users. Initially, there will be a service level agreement (SLA) between users and service providers (SPs) and/or SPs and infrastructure providers (IPs). However, the most crucial service required by the users and SPs in cloud computing is security and privacy. Security parameters can be used to prevent attacks and to protect data and systems. In literature, there is no comprehensive solution which quantify all the security parameters associated with the cloud computing paradigm. In this paper, for the first time, we attempt to generalize and quantify the security parameters.

Emerging cloud computing infrastructure replaces traditional outsourcing techniques and provides flexible services to clients at different locations via Internet. This leads to the requirement for data classification to be performed by potentially untrusted servers in the cloud. Within this context, classifier built by the server can be utilized by clients in order to classify their own data samples over the cloud. In this paper, we study a privacy-preserving (PP) data classification technique where the server is unable to learn any knowledge about clients’ input data samples while the server side classifier is also kept secret from the clients during the classification process. More specifically, to the best of our knowledge, we propose the first known client-server data classification protocol using support vector machine. The proposed protocol performs PP classification for both two-class and multi-class problems. The protocol exploits properties of Pailler homomorphic encryption and secure two-party computation. At the core of our protocol lies an efficient, novel protocol for securely obtaining the sign of Pailler encrypted numbers.

A clinical decision support system forms a critical capability to link health observations with health knowledge to influence choices by clinicians for improved healthcare. Recent trends toward remote outsourcing can be exploited to provide efficient and accurate clinical decision support in healthcare. In this scenario, clinicians can use the health knowledge located in remote servers via the Internet to diagnose their patients. However, the fact that these servers are third party and therefore potentially not fully trusted raises possible privacy concerns. In this paper, we propose a novel privacy-preserving protocol for a clinical decision support system where the patients' data always remain in an encrypted form during the diagnosis process. Hence, the server involved in the diagnosis process is not able to learn any extra knowledge about the patient's data and results. Our experimental results on popular medical datasets from UCI-database demonstrate that the accuracy of the proposed protocol is up to 97.21% and the privacy of patient data is not compromised.

Surname (family name) analysis is used in geography to understand population origins, migration, identity, social norms and cultural customs. Some of these are supposedly evolved over generations. Surnames exhibit good statistical properties that can be used to extract information in names data set such as automatic detection of ethnic or community groups in names. An e-mail address, often contains surname as a substring. This containment may be full or partial. An e-mail address categorization based on semantics of surnames is the objective of this paper. This is achieved in two phases. First phase deals with surname representation and clustering. Here, a vector space model is proposed where latent semantic analysis is performed. Clustering is done using the method called average-linkage method. In the second phase, an email is categorized as belonging to one of the categories (discovered in first phase). For this, substring matching is required, which is done in an efficient way by using suffix tree data structure. We perform experimental evaluation for the 500 most frequently occurring surnames in India and United Kingdom. Also, we categorize the e-mail addresses that have these surnames as substrings.

We solve a mixed quality of services (QoS) requirement problem for a multiple-input-multiple-output (MIMO) network with multiple linear transmit covariance constraints. Specifically, we design linear transceivers to satisfy the data rate requirements for a set of users while the rates of the remaining users are balanced. In addition, the design will ensure a set of multiple linear transmit covariance constraints are satisfied. The coupled structure of the transmit filters makes the original problem difficult to solve in the broadcast channel (BC). Hence, we propose an iterative algorithm to solve this mixed QoS problem based on stream-wise mean square error (MSE) duality and alternating optimization framework where the optimization problem is switched between the virtual multiple access channel (MAC) and the BC by exploiting stream-wise MSE duality. The proposed iterative algorithm solves the rate balancing problem by modifying the target rates of the users. In each iteration, a quadratically constrained quadratic programming (QCQP) is solved to obtain the virtual MAC receiver filters by incorporating multiple linear transmit covariance constraints, where the downlink receiver filters are obtained by minimizing each layer MSE. The power allocation in the virtual MAC is determined by solving a geometric programming (GP) where the product of layer MSEs of each user is balanced with total transmit power constraint. Simulation results for an underlay MIMO cognitive radio network demonstrate the convergence of the proposed algorithm.

The authors propose two algorithms to solve sum mean‐square error (MSE) minimisation and mixed quality of service (QoS) requirement problems for a multiuser multiple‐input‐multiple‐output system with multiple linear transmit covariance constraints. These original problems in the downlink are complicated because of the coupled structure of the transmitter filters. To overcome this issue, MSE duality proposed in the literature is extended at different levels for a general linear transmit covariance constraint. Exploiting the general sum‐MSE duality and subgradient method, the sum‐MSE minimisation algorithm is proposed first for multiple linear transmit covariance constraints. Secondly, a novel algorithm is proposed to solve mixed QoS requirement problem, where multiple linear transmit covariance constraints are incorporated in the design of the receiver filters in the equivalent multiple access channel. This algorithm is developed based on stream‐wise MSE duality and alternating optimisation framework. Simulation results have been provided to validate the convergence of the proposed algorithms. In addition, the proposed sum‐MSE minimisation algorithm with per‐antenna power constraints outperforms the existing algorithm in terms of achieved sum‐MSE and power consumption at each transmit antenna.

Facial expression recognition forms a critical capability desired by human-interacting systems that aim to be responsive to variations in the human's emotional state. Recent trends toward cloud computing and outsourcing has led to the requirement for facial expression recognition to be performed remotely by potentially untrusted servers. This paper presents a system that addresses the challenge of performing facial expression recognition when the test image is in the encrypted domain. More specifically, to the best of our knowledge, this is the first known result that performs facial expression recognition in the encrypted domain. Such a system solves the problem of needing to trust servers since the test image for facial expression recognition can remain in encrypted form at all times without needing any decryption, even during the expression recognition process. Our experimental results on popular JAFFE and MUG facial expression databases demonstrate that recognition rate of up to 95.24 percent can be achieved even in the encrypted domain.

The authors propose a suboptimal algorithm for adaptive subcarrier, bit and power allocation for orthogonal frequency division multiple access-based spectrum-sharing networks. This problem in its original form is non-convex and may be solved using greedy algorithms or integer linear programming (ILP) techniques. However, the computational complexity of the latter techniques is quite high, while the suboptimal greedy algorithms are not very well suited for spectrum-sharing networks because of multiple constraints on the transmitted power, interference leakage and individual user data rate. Therefore the authors propose a novel recursion-based linear optimisation framework that provides a solution that is very close to the optimal one and that has the ability to perform adaptive subcarrier, bit and power allocation for multiple users in the presence of multiple individual user constraints. Owing to the convexity of the proposed algorithm at each recursion, its overall complexity is substantially lower than that of the ILP-based solution.

We provide a rate balancing technique with mixed Quality of Services (QoS) requirement for an underlay multiple-input-multiple-output (MIMO) cognitive radio network (CRN). Specifically, we have considered an optimization criterion such that a set of SUs are required to achieve a target data rate whilst the data rates for the remaining SUs are to be balanced. This problem with a mixed QoS requirement in the broadcast channel (BC) cannot be solved directly due to a coupled structure of the transmitted covariance matrices. Hence, we solve an equivalent multiple access channel (MAC) problem using BC-MAC duality and subgradient method. An iterative algorithm is proposed to determine the transmit covariance matrices. The convergence analysis and simulation results are provided to validate the proposed algorithm.

We consider an underlay cognitive radio network, in which the cognitive users (also referred to as secondary users (SUs)) are allowed to access the licensed spectrum simultaneously with the primary users (PUs). Specifically we solve a beamforming and power allocation problem in the downlink with mixed quality-of-service (QoS) requirements where a set of SUs are required to achieve a specific signal-to-interference and noise ratio (SINR) targets whilst the SINRs for the remaining SUs are balanced. This mixed QoS requirement problem is more complicated in the downlink because of the coupled structure of beamformers and power allocations. Hence, we solve an equivalent uplink problem based on the uplink-downlink duality and subgradient method. An iterative algorithm is proposed to determine the optimal beamformers and power allocation. Simulation results are provided to validate the optimality of the result and the convergence of the proposed algorithm.