A condition-based distributed approach for secured privacy preservation of nodes in wireless sensor networks IoT

ABSTRACT


INTRODUCTION
To gather and share data, the internet of things (IoT) is a network or hub of machines, objects, and devices equipped with sensors and connectivity-enabled technologies.The technology intends to revolutionize human existence by boosting internet technology; consequently, its applications in a range of lifestyles are expected to increase substantially [1].The IoT is a new paradigm that enables the internet connectivity of multiple smart objects.Actuators and sensors may independently manage and transmit data to a system.Wireless sensor networks (WSNs) are an integral component of the IoT and are regularly used to collect data from local devices and send it to a central controller for further processing.IoT can benefit from WSNs, which can incorporate a range of processing, communication, caching, and sensing smart device components [2], [3].
WSNs are at the forefront of business, smart home, and environmental monitoring communication systems.The small size, low cost, and ease of deployment of this technology give real-time applications far more possibilities [1], [2].WSN application depends on several variables.Even if there is no energy constraint, the delivery packets are the most critical factor to consider when building a network for industrial − CDPP is evaluated considering the various compromised nodes to prove the CDPP model; further efficiency is proved by comparing with the existing low energy adaptive clustering hierarchy (LEACH) protocol for the detection of compromised and non-compromised nodes.This research is organized as follows: i) the first section starts with a background of WSN along with the importance of security along with privacy, further, the section moves forward with research motivation and contribution; ii) the second section presents the existing security technique to preserve node privacy along with its shortcomings; iii) CDPP and its mathematical modelling are presented in the third section of the research work; and iv) CDPP is evaluated in the fourth section of the research.

RELATED WORK
Security and privacy have been one of the early issues raised in the research area concerning WSN; apart from energy efficiency, it is considered the major area for the researcher to focus on and develop a lightweight security model.This section of the research reviews the existing privacy and security aware framework developed.The star and tree topologies are combined in this work by Naghibi and Barati [12] to offer a safe data aggregation structure.The network is physically divided into four equal parts.A predictable and consistent informational star structure is known to each component.Pirbhulal et al. [13] offer a securityand dual-resource-aware architecture for internet of medical things (IoMT)-based remote medical systems.Medical data is secured using a biometrics-keys generation approach to assure consistency in IoMT and lower the system's resource needs.WSNs can benefit from the data aggregation technique created by Hasheminejad and Barati [14] based on a tree topology [15].The plan aims to reduce energy consumption, increase network reliability, and prolong network life.This method yields a three-part binary tree as well as trustworthy data aggregation and verification.The routing dynamic data integrity (RDDI) approach enhances the data distribution and route-finding process [16].
Another option for securely sending data is to use a fuzzy hierarchical method.Efficient healthcare data aggregation (EHDA) technology makes data aggregation safe and portable [17].The collecting node receives compressed health data from the sensor nodes.The secure and portable sharing of medical data is made possible by the use of symmetric key-based data encryption.Compression of healthcare data also lowers the cost of storage and transmission.Wang et al. [18] provide a binary tree-aided model for fog-based approaches.Sending processed data to the edge node is accomplished using the current method.Many IoT and industrial IoT applications employ WSNs, and in this case, the system works better while using fewer processing resources from the edge servers.The data aggregation methods employed in this study's industrial IoT (IIoT) are well known in WSNs for their capacity to lower energy usage, according to Li et al. [19].Furthermore, these networks are exposed to a range of dangers due to their wireless connection.As a result, it is crucial to protect data while it is being collected.A novel technique for guaranteeing homogenous sensor devices in IoT-enabled WSNs is presented by Miao et al. [20], mobile-edge nodes are necessary for this technique.
While acknowledging the high level of consumer spending, Li et al. [21] addresses the problem of user privacy and security on two levels.The unique lightweight approach for privacy protection presented in Zheng et al. [22] is used to construct two non-colluding cloud platforms and develop a homomorphic cryptosystem.Li et al. [23] presents a novel strategy to disguise-based data poisoning attack (DDPA).It is determined to use a method in which the negative characteristics are masked to conceal the processes used to uncover the truth.In addition, the limitation of maximizing the attack's efficacy is automatically overcome by producing optimization issues at the bi-level, which are subsequently addressed by a separate optimization approach.Information-theoretical privacy (SEITP) is a unique semantic awareness for the protection of privacy in the progression of online location sharing, as outlined in [24], [25].The highest level of protection is offered for both data privacy and semantic awareness.

PROPOSED METHOD
Secure data aggregation has been one of the efficient approach for securing the WSN, especially when the consensus-based protocol is adopted; in previous work, efficient secure aggrigated data (ESAD) and integrated data model (IDM) has been developed which solely focuses on the securing the aggregation approach while being efficient.However, while securing the data, it is important to preserve the privacy of the node especially in consensus-based protocol.This research work adopts the secured data aggregation from the previous work and develop a CDPP algorithm that aims at hiding the vulnerable information to the compromised nodes.

Preliminary analysis and network design
The network considered here for (nodes ≥ 2), for communicating nodes with various neighbours, the communication topology is captured by a graph known as a communication digraph.A graph here is defined as H d = (X, Y), here X = {x 1 , x 2 , x 3 … … ., x n } with cardinality nodes = |K| ≥ 2 is the set of nodes and ∁⊆ X × X − {(x b , x b )|x b ∈ X} is the set of edges whose value is depicted by g = |∁|.An edge connected from the node x a to x b that is denoted as g ba ≜ (x a , x b ) ∈ Y, this captures the node x b that receives the information from the node x a .The given digraph H d = (X, Y) connected through the nodes x b , x a ∈ X, x b ̸ = x a , there exists a directed route from x a to x b .The subset nodes are responsible for directly transmitting information to the node x b is called the set of neighbours of x b depicted as β b − = {x a ∈ X|(x b , x a ) ∈ Y}, the subset of nodes directly receive information from node x b for a set of neighbours of

Designing sensor nodes operation
At each time step v ∈ S ≥0 , for each node x b ∈ Y retains the state variables . Here e h is a memoryless function for the states g h and p h .By assuming each node aware of its neighbors to directly transmit messages to all of them.This cannot necessarily receive messages from them.The distributed protocols here, each node x b allocates a unique order in the set The values of mass variables are updated as ( 1) and ( 2), here 1 ba [v] = 0 if no message is received at the node x b from the neighbor x a at iteration [v], the following cases are encountered.− Scenario 1: is satisfied, the node x b updates the state variables as (3),

Problem definition for securing the vulnerable information of nodes
A connected digraph H d = (X, Y), where |X| ≥ 3.Here each node x b ∈ X has an initial state of g b [0].The nodes here calculate the exchange of information.The information transfer takes place between the nodes aligned with H d , which represents the system topology.
Any node in this set X, which means that it tries to identify the initial states g [0] for all the subsets of nodes in the network.The set X is segmented in two different ways: i) a subset of nodes X O ⊆ X to ensure privacy, the node

Condition based distributed approach
The main objective in the system is to evaluate g whilst ensuring the nodes follow the protocol, the approach is focused on the event-triggered deterministic algorithm with some alterations.The main issue is the approach deployed that focuses on an offset of the mass variable for each node x b ∈ X O , to preserve the privacy of its initial state g b [0].In the existing system, the node x b for the initial state to In ( 5), the offset is added is A b for each node x b which is greater than or equal to the node x b outer degree such that each neighbor x a out-degree such that each of its neighbor x a ∈ β b + that receives a single value of d b [l b ] from node x b .
In ( 6) determines the accumulated offset integrated with the computation by the node x b is equal to zero and the exact average of the nodes' initial state is determined without any error.
In (7), the offset d b [l b ] is injected into the network for each node x b when the event condition is triggered that needs to be non-negative to hold for each node after a few steps.The average value of the initial state is evaluated.
In (8), the node x b stops the offset in the network so that the accurate average of the initial states is estimated without any error.

𝑑 𝑏 [𝑙
The above choice state that the offset d b for each node x b injects the network to be selected which is negative and satisfies d b ≤−β b + .Henceforth ensuring the operation of the proposed mechanism.The eventtriggered conditions do not hold on to the proposed protocol failing to evaluate the average of the initial state.The proposed algorithm has a value transfer process in which each node has a connected digraph H d = (X, Y), which performs executions according to a set of the event-triggered conditions.Each node here No message is received from any of -its neighbors, and with no transmission, the mass variable retains the same.Algorithm 1 presents the conditional approach for hiding the vulnerable information for securing the sensor nodes privacy. and

PERFORMANCE EVALUATION
When the data is aggregated, it is important to preserve the privacy of the nodes especially its initial information; thus to preserve the privacy of these nodes, this research develops CDPP mechanism which aims to preserve the model's node privacy.CDPP mechanism seeks to safeguard the privacy of sensor nodes and the integrity of the data.Additionally, CDPP is analyzed in consideration for classification and misclassification of sensor nodes.It is evaluated with a 2 TB hard drive, 16 GB of RAM, and 2 GB of NVidia CUDA-capable graphics.The model provided here analyses an inaccurate identification of a node that leads to network inequalities by including many parts, including the classification of the correct node, the misclassification of the node, and the computation of the throughput for 10, 15, 20, and 25 nodes.In addition, a comparison study between the proposed model and the existing model is undertaken to ensure the model's security and efficiency and to conclude that the proposed system outperforms the existing system.

Nodes classification
In this section, the classification of the sensor nodes is carried out wherein a comparison is made between the existing system and the proposed system by evaluating the correct identification of nodes with 10, 15, 20, and 25 nodes.Figure 1 shows the comparison of the stated above; in the context of 10 compromised nodes, the existing system detects 91 sensor nodes and the proposed model identifies 100 nodes.Consequently, in the context of 15 sensor nodes, the existing system identifies 91 nodes whereas the proposed model identifies 98 nodes.For 20 sensor nodes, the existing system identifies 87 nodes whereas the proposed model identifies 91 nodes for 25 nodes the existing system identifies 82 nodes whereas the proposed model identifies 97 nodes.

Nodes misclassification
Figure 2 depicts the misclassified nodes for 10, 15, 20, and 25 sensor nodes.In 10 nodes context, the existing model misclassifies 10 wrong nodes whereas the proposed model misclassifies 1 node.In 15 nodes, the existing model misclassifies 10 nodes whereas the proposed model misclassifies 3 nodes.In 20 nodes context, the existing model misclassifies 14 wrong nodes whereas the proposed model wrongly identifies 10 nodes for 25 nodes the existing system misclassifies 18 nodes and the proposed model misclassifies only 13 nodes.

Throughput
Throughput is defined as the amount of work done in a specific amount of time, it displays the models' efficiency; further depicted in Figure 3.In the case of 10 compromised nodes, the throughput of the existing model is 0.819 and for the proposed model, it is 0.989010989.In the case of 15 compromised nodes, the throughput of the existing model is 0.7735 and for the proposed model, it is 0.915384615.In the case of 20 compromised nodes, the throughput of the existing model is 0.696 and for the proposed model, it is 0.83781609 whereas for 25 nodes, the throughput of the existing model is 0.615 and for the proposed model, it is 0.887195122.

Comparative analysis
This section displays the comparative analysis and shows the percentage improvisation for the proposed model from the existing model.The improvisation is carried out for 10, 15, 20, and 25 sensor nodes the improvisation for 10 nodes is 20.75836252%, improvisation for 15 nodes the improvisation is 18.34319527% for 20 nodes the improvisation is 20.22724% whereas, the improvisation for 25 nodes is 44.25936942%.Table 1 shows the improvisation of the CDPP model over the existing mechanism.

CONCLUSION
Sensor node security has been an integral part of any security framework of WSN-IoT; however, due to the development of the lightweight protocol, nodes face the exposing of its information, which could lead to the compromising position for the data transmission and result in violation of security protocol.This research work aims at securing the privacy of sensor nodes deployed in the network; CDPP adopts the secure data aggregation from previous work discussed earlier and develops certain conditions to meet the criteria that can protect the nodes against compromised nodes.Thus, if the specified conditions are satisfied, transmission to neighboring nodes continues; otherwise, transmission terminates.In addition, the CDPP architecture is evaluated for misclassification of nodes for 10, 15, 20, and 25 nodes.After calculating throughput and comparing the CDPP model to the current aggregation method, it is found that the CDPP model comparatively works better with the improvisation of 20.75%, 18.34%, 20.22%, and 44.25% for compromised nodes 10, 15, 20, and 25 in a respective manner.In a recent development to the growth of attack models based on deep learning, the future of research will involve the adoption of data integrity solutions such as blockchain.
{0,1, … … ., α b + − 1} for each of the outgoing edges f len b here x len ∈ β b + , specifically to the link (x len , x b ) for node x b depicted by T len b (where {T len b |x b ∈ β b + } = {0,1, … . ., β b + − 1}).The pre-determined order for execution of the proposed algorithm in a way for allowing the node x b to transfer the messages to the neighbors in a round-robin fashion.Each node x b in the network has an initial state g b [0] ∈ S .At each step v for each node x b ∈ X retains the parameters g b [v] ∈ S and p b [v] ∈ S ≥0 and state variables g b h [v] ∈ S and p b h [v] ∈ S ≥0 and e b h it then transmits g b [v + 1], p b [v + 1], to an out-neighbor x len ∈ β b + and set the value g b [v + 1]=0 and p b [v + 1] = 0.
x b ∈ X O to retain its initial state g b [0] for other nodes remaining nodes in the set are X k = X\X O are indifferent to privacy and ii) a subset of nodes X c ⊆ Xto gather among them in identifying the Int J Reconfigurable & Embedded Syst ISSN: 2089-4864  A condition-based distributed approach for secured privacy preservation of nodes … (Bharat Kumara) 445 initial values of various nodes.The nodes herein X c are responsible for not caring about the privacy to share the initial state with different nodes in X c .
g b [0]=g b [0]+s b , here s b ∈ G.The offset is initially given as d b is a negative number shown as d b ∈S ≥0 , henceforth to lead the calculations to the initial average after a few finite steps.Each node x b ensures that privacy values like d b [v] ∈ S ≥0 to add the steps, A b ∈ S ≥0 , for the offset added to the counter a b ∈ S ≥0 and the transmission counter as l b ∈ S ≥0 .The value of the initial offset d b to select greater neighbors β b + to node x b .For initial purpose each node selects steps a b whereas the offsets as d b [v] ∈ S ≥0 and d b [l b ] ∈ S >0 for all a b ∈ {0,1,2 … ., a b }.

.
x b ∈ X O to ensure privacy in these steps.− A counter a b is set to zero and sets the total number of offset-added steps A b such as A b ≥ β b + and the set of (A b + 1) with a positive offset d b [a b ] > 0, here a b ∈ {0,1,2 … ., a b }.The initial negative offset value d b injects the initial state value g b [0] to d b = − ∑ d b [a b ] The node x b consists of four-out neighbors.− To select the v len ∈ β b + in the order H lenb to transmit p b [0] and g b [0] + d b + d b [0] to the out-neighbor.Then it sets the value to g b [0] = 0, p b [0] = 0, and a b = a b + 1. − The algorithm is executed, at each step v, node x b to receive a set of mass variables g a [v] = 0 and p a [v] = 0 for each-in neighbor x a ∈ β b − .The node x b updates the variables with g a [v] to check if the events-triggered condition holds.If true then d b [a b ] to g a [v + 1] and enhances the offset counter a b by one.It then sets the variables p b h [v + 1] and g b h [v + 1] irrespective of p b h [v + 1] and g b h [v + 1].Then it transmits to an out-neighbor p b [v + 1] and g b [v + 1] to an out-neighbor in pre-trained order.Here x b holds the p b [v + 1] and g b [v + 1].

Algorithm 1 .Step 3 .Step 5 .
ISSN: 2089-4864 Int J Reconfigurable & Embedded Syst, Vol. 13, No. 2, July 2024: 441-449 446 Condition based privacy preserving Input: A digraph, H d = (X, Y) with nodes = |x| and ∁= |Y| edges, each node x b ∈ X here has an initial state of g b [0] ∈ Step 1. Initialize each node x b ∈ X O Step 2. A unique ID is assigned H lenb in the set {0,1,….., α b + − 1} for each of its neighbors x len ∈ β b + This sets the counter value q b to 0 and priority index w b to q b Step 4. Set the counter a b to 0, selecting A b ∈ Z .>0, here A b ≥ β b + , and d b [v] ≥ 0 for v ∈ {0,1, … , A b } and d b [v ′ ] = 0 for v ′ > A b .To set the value A b = − ∑ d b [a b ] This sets the value g b [0] = g b [0] + d b , p b [0] = 1, p b h [0] = 1, g b h [0]=g b [0] Step 6.It then picks the neighbor v len ∈ β b + such that H lenb = w b and transmits p b [0] and g b [0] + d b [0] to this out-neighbor.It sets the value g b [0] = 0, p b [0], a b = a b + 1 Step 7. Set the value q b = q b + 1 and w b = q b mod α b + Iteration: For v = 0,1,2, … … ., each node x b ∈ Vp does the following This receives g a [v], p a [v] in at least one-neighbor x a ∈ β b − updates the value accordingly.If either of the conditions holds; This sets g a [v + 1] = d b [a b ] + g b [v + 1] and a b = a b + 1 This sets p b h this transmits p b [v + 1] and g b [v + 1] to out-neighbor δ γ ∈ β b + for which ϑ γb = w b to set the value g b [v + 1] = 0, g b [v + 1] = 0 and p b [v + 1] = 0; This sets the value q b = q b + 1 and w b = q b mod α b + else it stores g b [v + 1] and p b [v + 1] Output: e b h [v], for each x b ∈ X

Table 1 .
Improvisation of the CDPP model over the existing mechanism