Privacy-aware secured discrete framework in wireless sensor network

ABSTRACT


INTRODUCTION
Presently, wireless sensory networks (WSN) are utilized at the forefront of communication systems that include environmental monitoring, smart home automation, and industries.This technology has a widely increasing potential concerning applications of real-time due to its small size, inexpensive and easy deployment [1], [2].The design of WSNs is dependent on various situations for their application.In applications of the industry as well as automation of the home, in which there is no constraint of energy, delivery packets are the most essential for designing a network.However, in an unsafe environment in which there is a battery that is not replaced or recharged which includes mining, which increased the lifetime of the network, is the main factor in designing the network.Although, the deployed area size has an important role in the design of the WSN.Considering smaller areas, the transmission of the packets is done directly from the sensory nodes to the base station or sinks nodes, whereas, in large areas, the transmission of packets happens via various intermediate nodes to sink nodes [3].Additionally, the system that is enabled using internet of things (IoT) has an ability for interconnection of various 'things' for effective communication as well as sharing of data for one network, there is a wide number of advantages that have attracted various technologists [4], [5].In the short term, IoT systems are an important part of various sectors that include healthcare, manufacturing, logistics as well as transportation, which enable crucial infrastructure of the IoT [6].The majority of the intelligent control nodes are linked to the internet in the IoT technology, such nodes are transmitters or sensors that can retrieve data from the other systems and process it without any intervention.Numerous IoT applications are adopted to improve system efficiency and quality in industrial management, transport, and medical; because of the increasing prevalence of the IoT [1]- [3].IoT technology focuses on multiple activities to achieve the objectives generated through smart services.Figure 1 shows the WSN and IoT integration working.Intelligent devices can interact with the real world through intelligent operations so that consumers may get the proper service whenever and wherever they need it.

Figure 1. WSN and IoT integration working
Figure 1 shows the integrated architecture of WSN-IoT; it comprises several modules.WSN is where several sensor networks are connected for efficient transmission through the gateway.The other two blocks are smart objects through which data are communicated to humans.Considering the increased complexity in architecture, several uses of heterogeneous devices and threats towards privacy are difficult to be identified, access, and mitigated.However, the large-scale increased complexity of IoT systems introduces a deluge of data.Confidential and sensitive data have been constantly shared among the networks, and privacy, as well as security, are huge concerns that are prevailing in the IoT's crucial infrastructure [7].Any of the attacks on cyber systems that are vulnerable could compromise the integrity and privacy of huge amounts of data that is sensitive.
The frequency and types of attacks have increased because of substantial technological advancements.Attackers frequently take advantage of the variability in the IoT to create trust challenges and modify behavior to mislead users about the dependability of the sensors and the services offered through them.Moreover, the variety and complexity of IoT systems as well as resource limitations have produced enormous issues in terms of reliability, confidentiality, and transparency, all of which are crucial for the successful deployment of IoT [5].As a result, the trust evaluation system is used to isolate unreliable objects and identify dishonest behavior.When taking any action, it also overcomes the possibility of ambiguity and minimizes the risks.This can enable IoT infrastructure operation in a controllable environment and prevent unpredictable events and service failures [4]- [6].Without a reliable security mechanism that prevents the creation of harmful models or at least lessens their influence, IoT systems are unlikely to be widely adopted [6], [7].As a result, IoT security uses systems for authentication and encryption.Strong authentication and encryption technologies help to address several IoT security concerns.The techniques for secure message transmission between nodes, such as authentication and encryption, serve as the first line of protection against outside threats [8], [9].These defenses can stop and identify external attacks, but they are unable to deal with insider threats and hostile network nodes.In practice, internal attackers can circumvent these defenses by gaining access to the common key and initiating several attacks against the IoT network.Because of this reason, it is essential to ensure trust to counter internal attacks in the IoT environment.Approaches towards the preservation of privacy have gathered attention over the years due to progress towards information technology that has threatened individual privacy.Although, considering the rise Privacy-aware secured discrete framework in wireless sensor network (Nandini Sonnappa) 77 towards the adoption of edge computing as well as fog paradigms for IoT crucial information has led to decreased latency, awareness of location, communication, and data sharing in real-time, and quality of service (QoS) [10], [11].The devices used for edge computing in IoT crucial information are prone to attacks of privacy [12].Considering this approach, many techniques for the preservation of privacy that include traditional as well as modern deep learning algorithms are proposed [13].The existing deep learning methodologies that are implemented based on fog/edge computing for crucial information solutions need expensive computation.
There are various kinds of attacks on IoT crucial information that include attacks for denial of service, Sybil attacks [8], [9].Access level-based cyber-attacks for IoT crucial networks are classified as passive and active attacks [10].Active attacks, which are also termed attacks that are security oriented, derange the communication of the network by avoiding the available security protection.On the contrary, passive attacks are termed as attacks that are privacy-oriented such as networks eavesdropping without any disturbance for gaining illicit access to confidential sensitive information.The widely growing IoT crucial information is now prone to various attacks by hackers as well as organized crime syndicates.The growth in the count of threats towards privacy that are targeted towards IoT crucial information results in motivation for the development of various solutions.Although, most of the security approaches that are proposed have the absence of applicability, which may be the result of the complexity of computation, expense, and other factors that are related.Motivated by the above issues, the contribution of the research is given as follows: i) this research takes privacy, security, and reliability into account for aggregation in WSN based IoT environment and develops a privacy and reliable aware discrete (PRD)-aggregation framework; ii) PRDaggregation framework is an integrated framework that utilizes the designed constraint, discrete, time, and event for preserving the nodes' privacy; furthermore, discrete based mechanism utilizes the threshold setting for nodes for detection at the packet level and node level; and iii) PRD-aggregation is evaluated by inducing the deceptive nodes in the network for detection of dishonest nodes for classification and misclassification of nodes identification.
This research is organized as follows: The first section starts with a background of IoT, sensor nodes, and issues of security and privacy of the nodes and the section concludes with research motivation and contribution.The second section presents the existing approach along with its shortcoming.The third section presents the mathematical modeling of the PR-aggregation model and the fourth section presents the performance evaluation.

RELATED WORK
The confidentiality issue of malicious node identification is solved in WSN, in [14] the algorithm is improvised for the neighbor weight trust detection (NWTD) mechanism.This algorithm in parallel updates the trust value to ensure the trust degree for the nodes and fix a minimum peak value in an acceptable range.However, this is capable of ensuring the segmentation of malicious nodes, the main aim of this approach the malicious node detection in WSN's states that in [15] a trusted method is developed based on Dempster-Shafer (D-S) evidence theory is developed by incorporating the indirect and direct trust of third-party nodes to assess the strength of the network.The validation of the data packet, in [16] and [17] based on evaluation have developed a trust model that is focused on the computation of trust degree to handle either direct or indirect levels of trust while taking into account the internal attacks that wireless sensor networks are susceptible.This method minimizes the amount of energy used by the network and generates the trusted benchmark to simplify decision-making, which is easier by frequently updating the degree of trust.To assure network security and dependability, it may further identify between malicious nodes and aging nodes.To reduce the uncertainty of the decisions made by the conventional trust approach, it analyses the signal success rate, node latency, accuracy, and fairness as trust metrics.A multi-attribute trust model was developed in [18] employing fuzzy processing to compute the overall trust value of each node; this demonstrated that the conclusion is accurate.A novel trust management system founded on the D-S evidence theory is discussed herein [19].
The D-S theory was used to analyze the spatiotemporal correlation of data collected by nearby sensor nodes to develop the trust model.Consequently, an evaluation is made to determine which nodes are malicious.In [20], the challenges are addressed by considering the single detection function and failure of the malicious node detection mechanism by introducing a novel malicious node detection model to counter the vicious libel behavior of high-reputation nodes within current WSNs.This model illustrated the Beta Distribution reputation distribution and the implicit reliability of third-party nodes.This guarantees the precise detection of malicious nodes; it additionally incorporates the trust values, which correspond to different attack types.In [20], a methodology of truth discovery for preserving privacy is proposed, although, the client overhead is high.In [21], a two-layer methodology is proposed for fulfilling the requirement of protecting the privacy of the user.In [22], a novel lightweight framework of truth discovery for preserving privacy is proposed that is used for the implementation of two cloud platforms that are non-colluding as well as adopting a homomorphic cryptosystem.In [23], a novel data poisoning disguise attack (DDPA) is proposed against private systems of crowd sensing that are empowered with the methodology of truth discovery.A stealth strategy is proposed that is, the malicious characteristics are disguised for avoiding the detection of methods for truth discovery.Along with this strategy of stealth, the limitation of maximizing the effectiveness of the attack is naturally avoided using optimization problems at the bi-level via structuring that is resolved by an alternate optimization algorithm.In [24], a novel semantic awareness for the preservation of privacy for the trajectory of online location sharing method called Semantic-awareness informationtheoretical privacy (SEITP), for the protection of privacy of data as well as semantic during semantic awareness of data is utilized can be preserved completely [25].

PROPOSED METHOD
The data that has been gathered from IoT devices must be required for analysis as well as utilization.However, the gathered data could also have information that is sensitive and personal to the user, this hinders the privacy protection of the users.The unwillingness of the users to contribute their information affects the usability of the data.Privacy and reliability while aggregation has been one of the wide research using the consensus-based approach where nodes are required to exchange and disclose their state information to its adjacent node.However, this directly involves the privacy and reliability of nodes.This research work aims to develop a PRD-aggregation mechanism to achieve the trade-off among privacy and aims to guarantee convergence.Moreover, secure data aggregation has been carried out in previous research work; this work utilizes a secured and efficient mechanism and further provides privacy and reliability for the user.Figure 2 presents the proposed PRD model that comprises various modules.At first, the system model is designed based on the connected graph; further total number of variables is selected along with the selection of adjacent nodes.Furthermore, considering the variable, nodes state is updated and constraints are checked and verified.

System modeling and problem definition
This section provides the mathematical foundation for modeling among the sensor nodes in the WSN, Let's consider any particular network Q(Q ≥ 2) that communicates with adjacent nodes; also, communication is established through the designed directed graph.In a designed graph, J g = (Y, H), where Y = [y 1 , y 2 , y 3 , … . .y q ] and q is the particular node set with G as the edges.Furthermore, we consider particular time step m ∈ C ≥ 0 with each node maintaining its particular state.Directed edge between y m and y l is denoted through (1).
The equation (1) validates that the sensor node   can receive information fromy l ; also, it is assumed that graph designs are connected such that there exists a direct path between the two distinctive nodes.Moreover, the sensor node set which can directly transmit to the node is known as an inner adjacent set of   and this can be represented as Q′ m = [y l ∈ Y| (y l , y m ) ∈ H]; also, sensor nodes that can receive direct information from node   is known as an outer adjacent set and denoted as Q′′ m .Furthermore, the behavior of sensor nodes in the network is discussed.Consider the particular strong connected graph as designed earlier, every The main aim of this research work is to design a particular strategy for a verifiable and reliable strategy for nodes y k ∈ Y s /Y f while communicating with the nodes.Moreover, the problem is designed to compute the B ′ as in (1).Moreover, communication takes only among the nodes, which are adjacent to the designed graph, and represents topology.A particular node in set  could be deceptive and tries to enter the network and violate privacy by trying to violate confidentiality, which means it tries to identify the particular stateb[0].Hence, Privacy can be defined as the node's and network's capability to hide the information, thus proposed model tends to preserve privacy by hiding their confidentiality.

Discretizing and consensus
Considering the characteristics of data sensing in the WSN-based IoT model, the proposed model is designed such that given time step m ∈ B ≥0 , every node holds the characteristics b In here t v is considered as yet another variable designed for output.Furthermore, it is assumed that each sensor node is aware of its other adjacent nodes and directly communicates with them; each node gets assigned with a particular order to each of the outgoing edges In ( 3) and ( 4), here

Designing constraints for privacy, reliability, and verifiability
In this section, we develop a constraint, which aims to preserve the node's information, and any malicious nodes, which aim to obtain information, get removed from the network.Moreover, the main intention of the defined algorithm is to compute the B′.The proposed model follows the event-based approach for security modeling i.e., whenever there is a violation, the event gets triggered.The existing privacy preservation approach utilizes the initial states of the node as b ′′′ [0] = b m [0] + x m .However, due to the imitation of preservation of privacy, the proposed model introduces the negative variable   such that it guarantees the computation of the consensus approach in a certain number of steps.Furthermore, each sensor node holds its privacy value parameter   [], steps   , counter variable, and transmission counter variable   .The absolute value of the initial value and the number of added variables   is required to be larger than the adjacent node   .Considering initialization, every node   selects the steps and variable parameter to satisfy the given below condition: Condition 1: Adding steps   in the variable of each node   required to be larger than or equal to   's degree such that the adjacent node receives at least one piece of information.
Constraint 4: Node   stops inducing variables such that states can be calculated as (9): The above constraints indicate that the initial variable is induced in the network hence it is negative and satisfies   ≥   ′′ .Furthermore, while data transmission constraint 1 and constraint 2 hold for each node after certain steps, and in case of violation it will fail to compute the average consensus.The proposed algorithm has a value transfer process in which each node has a connected digraph   = (, ), which performs executions according to a set of the event-triggered conditions.Each node here   ∈   to ensure privacy in these steps.− A counter   is set to zero and sets the total number of offset-added steps   such as   ≥   ′ and the set of (  + 1) with a positive offset   [  ] > 0, where   ∈ {0,1,2 … .,   }.The initial negative offset value   injects the initial state value Then it transmits to an out-neighbor ′′′  [ + 1] and   [ + 1] to an out-neighbor in pre-trained order.Here   holds the   [ + 1] and   [ + 1]., No message is received from any of-its neighbors, and with no transmission, the mass variable retains the same.

PRD-aggregation algorithm
This research work aims to develop a PRD-aggregation mechanism to achieve the trade-off among privacy and aims to guarantee convergence.Moreover, secure data aggregation has been carried out in previous research work.This work utilizes a secured and efficient mechanism and further provides privacy and reliability for the user.Input is taken as the connected graph   = (, ) with  = || edges along with the initial state of   [0] ∈  Step 1: Assigning a particular order    in a given set {0,1, …   ′′ } for each adjacent node   ∈

′′
Step 4: Setting up Step 9: If (1) and ( 2) hold then transmit the information about nodes for preserving privacy.
Step 10: Output as t m v [n] for each node   ∈ .In the above algorithm, the first   to zero and select the total number of variables adding with   steps such that   ≥   ′′ and with (  + 1) positive variable   [  ] > 0. Furthermore, considering   ∈   ′ following order   and transmit   ′ [0] and   ′ [0].Furthermore, while executing the algorithm, at every step , node   receive a set of requests for packet transmission from each adjacent node, conditions are checked for verification, and if holds then the data packets are transmitted and the node remains else nodes are discarded out of the network.

PERFORMANCE EVALUATION
While performing the aggregation it is very important, that reliable and truthful data be collected.PRD-aggregation aims to assure the privacy of sensor nodes as well the reliability of the data.Moreover, PRD-aggregation is evaluated considering the dishonest sensor nodes.It is evaluated on the system configuration that includes 2 TB of the hard disk loaded with 16 GB of random-access memory (RAM) along with 2 GB NVidia Cuda-enabled graphics.The proposed model here works on evaluating an incorrect identification of the node that results in unevenness in the network by accommodating various parameters like identification of the correct node, identification of the wrong node, and computing the throughput for 30, 40, and 50 nodes.Additionally, a comparative analysis is done between the proposed model with the existing model to ensure the model's security and efficiency and conclude that our proposed system performs better than the existing system.

Energy utilization
Figure 3 shows the utilization of energy over the deceptive nodes, the graph is plotted for energy utilization by the deceptive nodes for 30, 40, and 50 nodes.For 30 nodes there is a steep increase in the after 4.3 value, whereas for 40 nodes there is a steep rise after 4.5 value.For 50 nodes, there exists a steep rise after 4.9 value.An overall comparison is shown below for 30 40 and 50 nodes.

Correct identification
Identification of node and packet marks the evaluation and performance of the PRD-Aggregation model.This section evaluates the security framework based on packet level and node level.The correct nodes are identified wherein a comparison is made between the existing system and the proposed system by evaluating the correct identification of nodes with 30, 40, and 50 nodes.

Packet level
In this section, the packets are identified wherein a comparison is made between the existing system and the proposed system by evaluating the comparison with 30, 40, and 50 nodes.Figure 4 shows the comparison of the stated above; in the context of 30, 40, and 50 compromised nodes, the existing system detects 70, 76, and 77 sensor nodes respectively.In the context of 30, 40, and 50 compromised nodes, the proposed model identifies 89, 97, and 96 nodes respectively.

Node level
In this section, the correct nodes are identified wherein a comparison is made between the existing system and the proposed system by evaluating the correct identification of nodes with 30, 40, and 50 nodes.Figure 5 shows the comparison of the stated above; in the context of 30, 40, and 50 compromised nodes, the existing system detects 41.75 and 66 sensor nodes respectively.In the context of 30, 40, and 50 compromised nodes, the proposed model identifies 97.99 and 97 nodes respectively.

Wrong node identification
Figure 6 depicts the wrong identification of nodes for 30, 40, and 50 sensor nodes.In 30 nodes context, the existing model identifies 34 wrong nodes whereas the proposed model wrongly identifies 4 nodes.In 40 nodes, the existing model wrongly identifies 36 nodes whereas the proposed model wrongly

Throughput
Throughput is defined as the amount of work done in a specific amount of time; it displays the models' efficiency.This section evaluates the throughput parameter at the packet level and node level.Throughput_packet_level is identified wherein a comparison is made between the existing system and the proposed system.

Packet level
In this section, the throughput_packet_level is identified wherein a comparison is made between the existing system and the proposed system by evaluating the comparison with 30, 40, and 50 nodes.Figure 7 shows the comparison of the stated above; in the context of 30 compromised nodes, the existing system detects 0.266 packet_level and the proposed model identifies 0.3382 packet_level.Consequently, in the context of 40 sensor nodes, the existing system identifies 0.2128 packet_level whereas the proposed model identifies 0.2716 packet_level.For 50 sensor nodes, the existing system identifies 0.1155 packet_level whereas the proposed model identifies 0.144 packet_level.

Node level
Figure 8 shows the throughput comparison for 30, 40, and 50 sensor nodes.In the case of 30 compromised nodes, the throughput of the existing model is 0.123.For the proposed model, it is 1.09933.In the case of 40 compromised nodes, the throughput of the existing model is 0.3 and for the proposed model, it is 0.891.In the case of 50 compromised nodes, the throughput of the existing model is 0.33 and for the proposed model, it is 0.693.

Comparative analysis
PRD-aggregation framework is evaluated in the previous section and analyzed in this section.Comparative analysis is carried out by analyzing the performance enhancement over the low-energy adaptive clustering hierarchy (LEACH) protocol.Several parameters are mentioned for the comparative analysis.

Improvisation over correctly classified nodes
This section displays the comparative analysis and shows the percentage improvisation for the proposed model from the existing model.Table 1 shows the deceptive nodes improvisation table.The improvisation is carried out on correct node identification for 30 nodes is 8,159%.The improvisation for 50 nodes the improvisation is 2.758% for 50 nodes the improvisation is 3,803%.

Improvisation over wrong classified nodes
Table 2 shows the wrong node identification.For wrong node identification, the improvisation for 30 nodes is 1.25%.The improvisation for 40 nodes is 1.05263%.The improvisation for 50 nodes is 15.89%.Above mentioned table shows the improvisation.

Throughput improvisation
For throughput_packet_level, the improvisation for 30 nodes is 2.38%, and for 40 nodes, the improvisation is 2.427%.For 50 nodes, it is 2.196%.For throughput at the node level.Table 3 shows the throughput performance.The improvisation for 30 nodes is 15.95%, for 40 nodes, the improvisation is 9.924%, and for 50 nodes, it is 7.096%.

CONCLUSION
Privacy has been one of the major concerns in data transmission, and aggregation due to the adoption of a consensus-based mechanism for efficient operation; this research work develops a novel mechanism PRD-aggregation mechanism for ensuring the sensor nodes' privacy to enhance the confidentiality.The PRD-aggregation novelty lies in the optimization of the consensus approach through the proposed algorithm.PRD-aggregation framework is evaluated at node and packet level for security considering 30, 40, and 50 deceptive nodes.Moreover, the evaluation shows the significant improvisation over the existing leach protocol with classification, misclassification, and throughput.In terms of throughput, the PRD-aggregation framework achieves improvisation of 2-2.5 at the packet level whereas near or more than 10% at the node level.The future scope of research lies in the adoption of data integrity technologies such as blockchain due to the rise of deep learning-based attack models.

Figure 3 .
Figure 3. Energy utilization of deceptive nodes

Figure 7 .
Figure 7. Throughput packet_level comparison Figure 8. Throughput comparison 79node has a certain stage denoted as b l [0], thus nodes are required to compute the (2) that satisfies the verifiable and reliable criteria.
Privacy-aware secured discrete framework in wireless sensor network(Nandini Sonnappa) .Here   ∈   ′ , specifically to the link (y o , y m ) for node    depicted by    (where {   |  ∈   ′ } = {0,1, … . .,   ′ − 1}).The proposed model develops the consensus model to process and transmit the information for efficient communication and obtains   which is equivalent to the absolute average of the initial states of the node after a certain number of steps.Moreover, it is assumed that every sensor node in the network holds initial states b m [0] ∈ C ; furthermore, at every time step   ∈  retains the parameters   [] ∈  and   [] ∈  and state variables    [] ∈  and    [] ∈  and    [] =    []/   [].  [ + 1] =   [] + ∑   ∈  ′′ 1  []  [] Moreover, the accumulated variable is infused in computation through the node   such that the node state can be computed without any error.

Table 1 .
Deceptive nodes improvisation table

Table 2 .
Shows the wrong node identification

Table 3 .
Throughput performance Deceptive nodes Packet level (improvisation in percentage) Node level (Improvisation in percentage)