978-1-5386-5367-8/18/$31.00 ©2018 IEEE
A Novel Secure Data Aggregation in IoT using Particle Swarm Optimization Algorithm
Neeraj Chandnani Research Scholar Devi Ahilya University, Indore(M.P.)-452001, India Research Centre: Military College of Telecommunication Engineering, Mhow (M.P.)-453441, India Email: chandnani.neeraj@gmail.com C N Khairnar Faculty of Communication Engineering, Military College of Telecommunication Engineering, Mhow (M.P.)-453441, India Email: cnkhairnar@gmail.com
Abstract-
Internet of Things (IoT) is a network paradigm inwhich data aggregation and data security plays a vital role. Data aggregation in IoT describes collection of information from different users and data security means encryption of collected data using cryptography method. The proposed work comprises of devices and gateway to perform data aggregation and data encryption. Data aggregation is performed using clustering in which data are clustered and securedby Particle SwarmOptimization (PSO) algorithm which finds the cluster head. After finding cluster head, nodes requests to join as cluster member. PSO computes fitness function using metrics i.e. energy, end-to- end delay, scoring factor, packet drops and successful packet transformation. After completion of clustering process, data encryption process is held in which, cluster head collects data from the cluster members and encrypts it using Elliptic Curve Cryptography (ECC) method. Finally, encrypted data are dispatched to gateway device.Experimental result shows, the proposed work on Secure Particle Swarm Optimization (S-PSO)prompts better performance in following metrics i.e.delay, throughput and energy consumption.
Keywords-
IoT, Data aggregation, Security, Particle Swarm Optimization, Elliptic Curve Cryptography.
I.
I
NTRODUCTION
The Internet of things (IoT) is the network of physical devices, vehicles, home appliances and other items embedded with electronics, software, sensors, actuators and connectivity which enables these things to connect and exchange data. The concept of IoT came into existence in the year 1999. However due to tremendous development in wireless devices and standards the concept of IoT has come to reality now. The number of IoT devices increased 31% year over year to 8.4 billion in the year 2017 and it is estimated that there will be 30 billion devices by 2020. There is need to look into IoT network and their characteristics. Emerging IoT network has lot of issues related to data aggregation, security, clustering of IoT devices and encryption. IoT is an emerging network standard in which our day-to-day objects are equipped with capabilities of identifying, sensing, processing and networking functionalities whichallow them to communicate between one another through internet. The current state of IoT is examined by identifying current trends and describing challenges that threaten IoT [1]. Data collection is significant challenge in IoT that is addressed by many authors in their studies. Data collection is performed by collecting information from devices at a certain place and certain time. The key point is to regulate network performance for data collection by considering application requirement [2]. Distribution algorithm is proposed to collect ubiquitous data. This algorithm transmits data collected by surrounding sensor nodes and prioritizes the transmission of most important and relevant data. The proposed algorithm can support smooth collection and coordination of different users described in reference [3].To collect information from physical and social layer learning-based method is proposed. In physical layer, Weiner process is used to find distance between two devices and connection probability is defined by Kolmogorov equation. In social layer, Bayesian non- parametric based learning is used to collect information from real world [4]. To avoid threats and attacks in data collection of IoT, one must authenticate the partner’s identity. In this,collector identifies the smart card owner without gateway node. Sensor node transferred data to collector through a secure channel. Collector also confirms collected dataare from valid sensor nodes [5].Ensuring the security of IoT and its application is important factor which is preserved by ultra-lightweight RFID authentication protocol. As per reference [6] a passive secrete disclosure attack against security threats. The success probability of attack is ‘1’ while complexity of attack is overhearing one session of protocol.Information generated from the cluster devices lacks in security of transmitting and storing data in IoT. This approach enables edge device to encrypt their data using Advanced Encryption Standard (AES) algorithm before transmitting into other devices. The key of AES is encrypted by using Rivest–Shamir–Adleman (RSA) crypto systems which is exchanged to authenticated person via e-mail [7].To secure data in IoT hybrid security model is proposed. Hybrid security model developed through 2-D discrete wavelet transform 1 level (2D-DWT-1L) or 2-D discrete wavelet transform 2 level (2D-DWT-2L) steganography technique. The proposed hybrid model is built using Advanced Encryption Standard (AES) algorithm and Rivest–Shamir–Adleman algorithms. The proposed model starts by encrypting the secret data; then it hides the result using 2D-DWT-1L or 2D-DWT-2L techniques [8]. Cluster resource indexing of IoT based on improved Ant Colony Optimization (ACO) algorithm use directed graph modelsto construct a distribution structure model of cluster resource indexing nodes of IoT. This algorithm initialize cluster center and target function of cluster resource indexing of IoT is
Authorized licensed use limited to: South Asian University. Downloaded on February 12,2024 at 17:23:19 UTC from IEEE Xplore. Restrictions apply.
978-1-5386-5367-8/18/$31.00 ©2018 IEEE
constructed [9]. Fuzzy clustering method is used for intrusion detection in IoT. Two algorithmssuppress fuzzy clustering algorithm and Principle Component Analysis algorithm are proposed. Using high frequency and low frequency data are classified into two specifically low risk and high risk data. Self-adjustment of the detection frequency is carried out according to the suppressed fuzzy clustering algorithm and the principal component analysis algorithm [10]. Efficient secure cryptography scheme designing for IoT is challenging task, the work in reference [11] proposes an authentication key agreement scheme to build a secure channel in the IoT. Thescheme has lightweight computation and provides security properties for key agreement. Data communication in IoT contains personal information and sensing information collected from the surrounding environment. Conventional cipher algorithm described in reference [12] is used for sharing and managing data in various circumstances. This method implements proxy re-encryption in order to manage data with fewer encryptions. It also provides data sharing function to supplement the insufficient capacity. Our proposed work shows novel data clustering and data securityprocess in IoT. In this work, two algorithms are proposed namely Bio-Inspired algorithm and encryption algorithm. Key points of proposed work are listed as follows:
•
In the proposed work two phases are performed specifically phase 1 and phase 2. In phase 1, clustering is performed to aggregate data securely. In phase 2, data security is accomplished to secure collected data during transmission.
•
At first, clustering is performed using PSO algorithm to secure data aggregation in IoT. It selects cluster head using fitness functions computations which is based on metrics such as energy, end to end delay, scoring factor, packet drops and successful packet transformation.
•
Cluster head collects data from its cluster member and encrypts it using ECCalgorithm to secure transmission and then it forwards it into gateway.
•
The proposed work, achieves better results compared to existing methods in terms of the following three metrics that are delay, throughput and energy consumption. II.
R
ELATED WORK
IoT devices increase the risk of security threatssuch as cyber attacks and viruses. To ensure, IoT trust solutions across various IoT node scalable trust management solution was proposed. In this approach, four algorithms are proposed;First algorithm takes outliers of trust values, second algorithm forms the intelligent trust based clusters, third algorithm migrates IoT node from one cluster to another cluster and Fourth algorithm examines current IoT cluster node states. This approach uses four algorithm which interns increases computational time [13]. To solve the various complex problems associated with IoT devices, this method proposes IoT device classification and clustering system. The classic EM (Expectation and Maximization) algorithm was used to cluster the IoT services. This algorithm divides 100 IoT devices into five clusters. Convergence speed of EM algorithm was very slow [14]. To improve cognitive ability of IoT, this method proposes device level multimodal data correlation mining model was designed. This model was designed based on Canonical Correlation Analysis (CCA) to transform data feature into a subspace and analyze the data correlation. The result of correlation analysis was used to classify a device in which Heterogeneous clustering model was performed. Finally, a device level clustering algorithm based on multimodal data correlation was proposedwhich combines the functions of multimodal data correlation analyze with device clustering. Heterogeneous clustering model selects low initial energy sensors to act as cluster head which interns reduces the cluster efficiency [15]. Sensors and servers in IoT communicate with each other one to one basis which increases the load, if number of sensors are increased. To overcome this problem, this method uses clustering approach for message queuing. This method proposes cluster based constrained Application Protocol for efficient gathering of sensing data. Cache memory was used in cluster head to improve the communication performance in IoT networks. Cache memory on cluster head takes more time to collect request from the user [16]. Cluster head in IoT consumes more energy, to overcome this problem anUnequal Clustering Algorithm Concerned with Time Delay (UCATD)was proposed. In this, K-means clustering algorithm was used to cluster the network. In transmission phase, data fusion mechanism was used to improve the energy utilization rate for cluster head. Optimized data fusion tree construction based algorithm was proposed to reduce the transmission delay. This method was only suitable for delay limited applications of IoT [17]. Collaboration among devices in IoT is difficult task. To overcome this problem, this proposal introduces two layer architecture frameworks for IoT which consist of sensing layer and IoT layer. These two layers are important for accomplishing IoT based applications. Clustering of device was performed at various levels and it helps in collaboration as well as network lifetime. This proposal introduces two clustering algorithms based on heuristic and graph respectively. To form a cluster, these clustering algorithms are not considered distance between two nodes which was important to form a cluster [18]. To overwhelm the problem of energy utilization in IoT services, this proposal introduces two foldmethods. At first, it analyzes the maximization of network lifetime by balancing the energy consumption among Cluster Heads (CHs) in which cluster radius of each level has significant role in maximization of network lifetime. Secondly, a novel Lifetime Maximizing optimal Clustering Algorithm (LiMCA) was used to maximize the network life time. LiMCA includes novel stochastic deployment scheme for Member Node (MN) and CHs. A training protocol was used to train CHs and MNs about their coarse grained location. This proposal allocates same amount of time slot for all MNs which may lead to packet loss during data transmission [19]. To secure the transmission data in IoT, this proposal introduces the Dynamic Bayesian Network combined with trusted protocol. Thismethod enhances the security information exchange during data transmission. Trusted Platform Module (TPM) was key part of the trusted computing platform. TPM manages many keys which are used to do the functions such as security storage, integrity measurement, storage and reporting. Combined Private Key (CPK) based
Authorized licensed use limited to: South Asian University. Downloaded on February 12,2024 at 17:23:19 UTC from IEEE Xplore. Restrictions apply.
978-1-5386-5367-8/18/$31.00 ©2018 IEEE
security algorithm was used to enhance the security of the protocol interaction. Dynamic Bayesian Network needs prior information of devices [20]. A terminal that runs untrusted in the IoT, may lead to severe issues. To avoid this problem, this proposal introduces the security enhanced attestation which can achieve shielded execution for measurement and attestations problems. The policybased measurement mechanism was proposed where sensitive data including secrete keys and policy details are veiled using enclave specific keys [21]. To improve the security in IoT environment, this proposal introduces the optimal lightweight cryptography scheme to enhance the security. This scheme compares the CPU usage and processing time. The lightweight cryptography method normally known as lightweight block cryptography which was characterized as that length of the key was proportional to the security. This lightweight block cryptography scheme has higher processing time and CPU usage [22]. The transmitted data through IoT devices secured using encrypted method. In this proposal, data are encrypted using instant encryption based security scheme was introduced. This model has three phases i.e registration phase, detection phase, implementation phase. In registration phase, Key Generation Center (KGC) generates private keys for all sensors. Second phase was for encryption phase which encrypts data using sign information which transmits data in short time. Implementation phase requires encrypted data and identity authentication of sender was performed very short time. Computations are more in these phases which increase computational complexity [23]. To preserve the privacy of data, this proposal introduces the new secure channel free certificate less searchable public key encryption with multiple keywords (SCF-MCLPEKS) scheme. The proposed scheme demonstrating, its security against chosen keyword attacks in random oracle model. It has two adversaries namely Type1 and Type2. A Type1 adversary has the ability to choose public key to replace the user’s public key and Type2 adversary was given the system’s master key. In this proposed scheme, computational steps are more which interns increases the computational time [24]. The proposed method increases throughput meanwhile decreases delay and energy consumption rates. III.
P
ROPOSED
W
ORK
The proposed work is illustrated briefly in the following sections:
A. Overview
To overcome problems discussed in previous work, a novel design is proposed in IoT. The proposed architecture is contains IoT devices and gateway which is used to support data aggregation. To secure data aggregation, our work is divides into two major processes such as, clustering and security. Clustering is performed using Bio-Inspired algorithm which is PSO. PSO is used to provide security, so that proposed algorithm is called Secure-PSO (S-PSO). In security, ECC algorithm is used which encrypts collected data given by cluster head and forward it to the gateway. Figure 1, shows overall architecture of the proposed work. Two phases are introduced in our work namely phase 1 and phase 2. In phase1, secure data aggregation is performed using clustering of nodes using PSO algorithm. In phase 2, data security is performed by ECC algorithm.
ECCECCECCInternetCluster 1Cluster 2Cluster nCluster Member(IoT device)Data AggregationSecure PSO algorithm for clustering
Cluster Head (IoT device)GatewayPhase 1Phase 2
Figure 1. Proposed Architecture
B. Data Aggregation
In phase 1, data aggregation is performed in whichS-PSO algorithm is proposed.Two processes are performed in clustering phase.First process in S-PSO algorithm is cluster head selection whereas second process is to form cluster. At first, S-PSO algorithm selects cluster head using fitness function computations. In fitness function computations, it considers metrics such as successful packet transmission, packet drops, energy, end-to-end delay and scoring factor. In these metrics, scoring factor describes trustworthiness of node, so that it can collects data from its entire cluster member. Scoring factor formulated using below formulae,
=
×
+
×
…. (1) Scoring factor is computed using forwarding ratio where,
represents cumulative count of correct forwarding data,
represents total count of data,
represents control forwarding ratio and
represents data forwarding ratio in time interval
.
represents trust value calculation on node
using node
.
and
are constants, which addition values are always equal to one (
). After finding scoring factor, fitness function is calculated. Fitness function is formulated as follows,
…. (2) Where
represents dropped packets,
represents total number of forwarded packets,
is remaining energy in node l,
represents initial energy,
is end to end delay,
is
Authorized licensed use limited to: South Asian University. Downloaded on February 12,2024 at 17:23:19 UTC from IEEE Xplore. Restrictions apply.
978-1-5386-5367-8/18/$31.00 ©2018 IEEE
maximum allowable delay. Using above formula, each node finds its fitness function and highest fitness function value node is elected as cluster head. Cluster head sends join request to its cluster member. By sending join request to each of its cluster member, it can form cluster.
C. Data Security
In phase 2, data security is performedin which security of transmitted data is achieved using ECC algorithm. Cluster head collects data from its cluster members. In ECC algorithm both transmitter and receiver must know the selected elliptic curve which is represented as H (A
P
). In this algorithm three stages are held namely key generation, encryption and decryption.
Key generation
Cluster head generates public key to transmit data safely. It selects aconstant G
(0 to n-1), where ‘n’ is considered as maximum limit which is prime number. Generated public key K is given below, K= G×T.…. (3) Where, T is point on the curve which is private key.
Encryption
After generating public key, cluster head encrypts the collected data. In encryption it selects a constant r
(0 to n-1). The given data is transformed into two cipher text J
1
and J
2
. J
1
= D×T.…. (4) J
2
=P+ (D×K).…. (5) The above equations represent cipher text J
1
and J
2
which are encrypted from the collected data D.
Decryption
After encryption, decryption is performed in gateway device. G is used for generating public key which is used in decryption. In this process original message is retrieved from the cipher text. C= J
2
- G×J
1
.…. (6) Figure 2 describes flow of the proposed work of this paper in which data are collected safely by cluster head and security is preserved using ECC algorithm which first generates public key and encrypts the collected data. IV.
P
ERFORMANCE
E
VALUATION
In this section we discuss about the performance achieved by the proposed work. This section is comprised of following sub-sections: (A) simulation environment and (B) comparative results.
A. Simulation Environment
The implementation of the proposed novel IoT architecture performed using Network Simulator 3 (NS3) which has been installed in Ubuntu Operating System. The simulation environment supports the proposed work characteristics. Table I, illustrates significant parameters that are considered for the simulation of IoT devices environment. Parameters are not limited up to these eight, it changes with relative change in IoT devices.
Figure 2.Flowchart for proposed work TABLE I.SIMULATION PARAMETERS
Parameters Value
Simulation Area 500×500 m Number of Nodes 50 Number of cluster 9 ECC encryption time 0.62 to 0.78ms ECC decryption time 0.44 to 0.5ms Key size 512bits Transmission Range 200m Simulation time 100s
B. Comparative Results
This section shows the effectiveness of the proposed novel IoT architecture which is compared with previous research works. For comparative analysis,three significant metrics namely delay, throughput and energy consumption have been considered.
Authorized licensed use limited to: South Asian University. Downloaded on February 12,2024 at 17:23:19 UTC from IEEE Xplore. Restrictions apply.