Energy Efficient Approach with ANN based Intrusion Detection System for Securing Network and Route for Transmission in Wireless Sensor Network
 
Saziya Tabbassum1*, Chandra Kumar Jha2, Sneha Asopa3
1 Research Scholar, Department of Computer Science, Banasthali Vidyapith, Rajasthan, India
Email: saziyatabassum@gmail.com
2 Professor & Head, Department of Computer Science, Banasthali Vidyapith, Rajasthan, India
3 Assistant Professor, Department of Computer Science, Banasthali Vidyapith, Rajasthan, India
Abstract: The research focuses on hierarchical clustering-based intrusion detection using artificial neural networks (ANNs) for secure data transmission in wireless sensor networks (WSNs). WSNs consist of numerous tiny sensor nodes deployed to monitor environmental phenomena. These nodes face significant challenges, including restricted energy, memory space, communication range, and limited capacity for managing energy, storing, transmitting, and processing data. To address these limitations, a machine learning-based approach is proposed to detect intrusions and efficiently utilize energy by properly selecting cluster heads using a secure clustering protocol. The proposed method was implemented and tested using MATLAB software, employing the NSLKDD and UNSW-NB15 datasets for intrusion detection. The results demonstrated promising outcomes in detecting intruders and enhancing network efficiency, achieving a 92% packet delivery ratio (PDR) and 1.82 Mbps throughput. The study concludes that while WSNs are gaining popularity due to their simplicity, flexibility, and scalability, innovative solutions are necessary for efficient energy management and security. Future research should focus on advanced machine learning models, energy harvesting techniques, scalable protocols, real-time data processing, and integration with IoT platforms for broader applications and enhanced functionality.
Keywords: Wireless Sensor Networks, Intrusion Detection, Artificial Neural Networks, Energy Efficiency, Secure Clustering Protocols
INTRODUCTION
Wireless Sensor Networks (WSNs) are composed of a large number of sensor nodes deployed to monitor physical or environmental conditions, such as temperature, humidity, and motion, and to cooperatively pass their data through the network to a main location (Akyildiz et al., 2002). These networks are characterized by their ability to self-organize and self-heal, making them highly suitable for deployment in dynamic and remote environments where human intervention is impractical. WSNs are used in various applications ranging from military surveillance to environmental monitoring and smart agriculture.
Sensor nodes are the fundamental units of WSNs, each equipped with a microcontroller, transceiver, power source, and one or more sensors (Heinzelman et al., 2002). These nodes are responsible for sensing data, processing it, and communicating with other nodes or a central base station. The key components of a sensor node include the sensing unit, processing unit, transceiver unit, and power unit. Due to their limited power supply, typically from batteries, energy efficiency is a critical consideration in the design and operation of sensor nodes. WSNs have diverse applications across various domains some of them are illustrated in figure 1. In environmental monitoring, WSNs are used to track climate changes, detect natural disasters, and monitor wildlife habitats (Hart and Martinez, 2006). In healthcare, they facilitate remote patient monitoring, tracking of vital signs, and management of chronic diseases (Lo et al., 2005). Industrial applications include monitoring manufacturing processes, detecting equipment failures, and ensuring workplace safety (Yang, 2014). The flexibility and scalability of WSNs make them valuable for real-time data collection and analysis in these and other areas.
Figure 1: Applications of WSNs along with its Subcategories
The protocol stack of a WSN is designed to support the unique communication and energy efficiency requirements of sensor networks. It includes physical, data link, network, transport, and application layers, along with power, connection, and task management planes (Al-Karaki and Kamal, 2004). The physical layer deals with frequency selection, carrier frequency generation, signal detection, modulation, and data encryption. The data link layer is responsible for multiplexing data streams, data frame detection, medium access, and error control. The network layer handles routing, while the transport layer ensures reliable data transfer. The application layer provides the necessary software for specific applications such as sensor management protocols (SMP) and query processing. Designing WSNs involves addressing several challenges, including energy efficiency, scalability, fault tolerance, and latency. The limited battery life of sensor nodes necessitates the development of energy-efficient protocols and algorithms (Pantazis et al., 2013). Scalability is crucial to accommodate a large number of nodes and ensure robust communication. Fault tolerance is needed to maintain network functionality despite node failures. Routing protocols in WSNs are typically categorized into flat, hierarchical, and location-based protocols. Flat routing protocols treat all nodes as equals, hierarchical protocols organize nodes into clusters, and location-based protocols use the physical location of nodes to guide routing decisions (Chen and Zhang, 2005). Security is a paramount concern in WSNs due to their deployment in potentially hostile environments and the critical nature of the data they collect. Common security threats include eavesdropping, node capture, denial of service (DoS) attacks, and data tampering (Perrig and 2004). Ensuring the confidentiality, integrity, and availability of data is essential for the reliable operation of WSNs.
Various mechanisms have been developed to secure WSNs against these threats. Cryptographic techniques are used to protect data confidentiality and integrity. Key management schemes ensure secure communication between nodes (Eschenauer and Gligor, 2002). Intrusion detection systems (IDS) are employed to detect and respond to malicious activities (Onat and Miri, 2005). Additionally, secure routing protocols are designed to safeguard data transmission from attacks (Karlof and Wagner, 2003). These mechanisms collectively enhance the resilience of WSNs against security breaches.
REVIEW OF LITERATURE
Clustering is a crucial technique in wireless sensor networks (WSNs) aimed at enhancing the scalability and energy efficiency of the network. Research has extensively explored various clustering algorithms, such as LEACH (Low-Energy Adaptive Clustering Hierarchy), HEED (Hybrid Energy-Efficient Distributed clustering), and TEEN (Threshold-sensitive Energy Efficient Sensor Network protocol) (Heinzelman et al., 2001). These algorithms primarily focus on reducing energy consumption by minimizing the distance that data needs to travel, thereby prolonging the lifetime of the sensor nodes. Clustering helps in load balancing and improves the network's robustness by organizing the sensor nodes into clusters, each with a designated cluster head responsible for data aggregation and communication with the base station. Studies have shown that effective clustering can significantly impact network performance, particularly in terms of energy conservation and latency reduction (Abbasi and Younis, 2007).
Energy efficiency is a paramount concern in WSNs due to the limited power resources of sensor nodes. Various methods have been proposed to optimize energy consumption, such as duty cycling, data aggregation, and energy-efficient routing protocols. Duty cycling involves alternating nodes between active and sleep states to conserve energy (Kuliket al., 2002). Data aggregation techniques reduce the volume of data transmitted by combining multiple data packets into one, thus saving energy (Krishnamachar et al., 2002). Energy-efficient routing protocols, such as PEGASIS (Power-Efficient Gathering in Sensor Information Systems) and SPIN (Sensor Protocols for Information via Negotiation), focus on selecting optimal paths for data transmission to minimize energy consumption (Lindsey and Raghavendra, 2002; Kulik et al., 2002). These methods collectively enhance the longevity and efficiency of WSNs, making them more sustainable for long-term applications.
Security in WSNs is critical due to the susceptibility of these networks to various attacks, including eavesdropping, spoofing, and denial of service (DoS). Several security protocols have been developed to counter these threats, such as TinySec, SPINS, and SNEP (Karlof et al., 2002). TinySec is a lightweight security protocol designed for sensor networks, providing confidentiality, authentication, and integrity at the link layer (Karlof, Sastry, & Wagner, 2004). SPINS (Security Protocols for Sensor Networks) include two secure building blocks: SNEP (Secure Network Encryption Protocol) and μTESLA (Micro Timed, Efficient, Streaming, Loss-tolerant Authentication), which ensure data confidentiality, two-party data authentication, and broadcast authentication, respectively (Perrig et al., 2002). These protocols have been effective in mitigating various security threats, thus ensuring the secure operation of WSNs.
Secure data transmission is essential to protect sensitive information in WSNs. Protocols such as LEAP (Localized Encryption and Authentication Protocol) and ZigBee are widely used for secure data communication (Zhu et al., 2003). LEAP provides a set of key management protocols that support in-network processing and restrict the impact of node compromise to the immediate vicinity of the compromised node (Zhu et al., 2003). ZigBee is a specification for a suite of high-level communication protocols using low-power digital radios, widely used for secure and reliable data transmission in WSNs (Kinney, 2003). These protocols employ various encryption and authentication mechanisms to ensure the integrity and confidentiality of the transmitted data, making them resilient to attacks.
Intrusion Detection Systems (IDS) in WSNs are designed to identify and respond to unauthorized activities within the network. Various IDS have been proposed, including anomaly-based, signature-based, and hybrid systems (Sun et al., 2004 and Rajasegarar et al., 2006). Anomaly-based IDS detects intrusions by identifying deviations from normal behavior patterns, while signature-based IDS uses predefined patterns to recognize known attacks. Hybrid IDS combines both approaches to enhance detection accuracy and reduce false positives (Rajasegarar et al., 2006). These systems play a crucial role in maintaining the security and reliability of WSNs by providing real-time monitoring and alerting mechanisms.
The literature review highlights significant advancements in clustering techniques, energy efficiency methods, security protocols, secure data transmission, and intrusion detection systems in WSNs. However, several research gaps remain. For instance, there is a need for more efficient and scalable clustering algorithms that can adapt to dynamic network conditions. Additionally, while existing security protocols address many threats, new and evolving attack vectors necessitate continuous improvement and innovation in security mechanisms. Furthermore, integrating energy-efficient techniques with robust security measures without compromising network performance remains a critical challenge. Addressing these gaps will be essential for advancing the state-of-the-art in WSNs.
RESEARCH METHODOLOGY
The methodology integrates hierarchical clustering and artificial neural networks (ANNs) to enhance intrusion detection and secure data transmission in wireless sensor networks (WSNs). This approach aims to develop a robust, energy-efficient, and secure WSN.
Wireless sensor networks (WSNs) face significant challenges, including limited energy resources, vulnerability to various security attacks, and the need for effective intrusion detection mechanisms. Traditional intrusion detection systems (IDS) often struggle with high false positive rates and energy inefficiency, compromising the network's performance and longevity. There is a critical need for a solution that not only detects intrusions accurately but also manages energy consumption effectively to prolong the network's operational life. The primary research problem addressed in this study is the development of a hierarchical clustering-based IDS that leverages artificial neural networks to detect and mitigate security threats while ensuring energy efficiency and secure data transmission in WSNs.
The main objectives of this study are as follows:
1. Design a Hierarchical Clustering Algorithm: To create a clustering algorithm that enhances energy efficiency and extends the network's lifetime by optimizing the distribution of sensor nodes and balancing energy consumption.
2. Develop an ANN-Based Intrusion Detection System: To build an artificial neural network model capable of accurately identifying and responding to security threats within the WSN by analysing network traffic patterns and distinguishing between normal and malicious activities.
3. Ensure Secure Data Transmission: To integrate robust encryption and authentication protocols that secure data transmission between sensor nodes and the base station, maintaining data integrity and confidentiality.
4. Evaluate System Performance: To assess the proposed system's performance in terms of energy efficiency, detection accuracy, false positive rate, network lifetime, and data transmission delay, and to compare these results with existing techniques to demonstrate the system's effectiveness and advantages.
Variables under Study
This study involves several key variables critical to evaluating the performance and effectiveness of the proposed intrusion detection system in wireless sensor networks (WSNs). These variables include:
Dataset Description
  1. NSL-KDD Dataset: The NSL-KDD dataset is a refined version of the KDD Cup 1999 dataset, widely used for evaluating intrusion detection systems. It provides labeled data on various types of network intrusions and normal traffic, making it suitable for training and testing the ANN-based IDS. Every record of NSL-KDD dataset has 42 attributes among which 41 attributes are as given in table whereas the 42nd attribute of dataset has five different classes categorized as a normal and four attack class.
Table 1: The features of the NSL-KDD dataset
S. No.
Features
S. No.
Features
1
Duration
22
is_guest_login
2
protocol type
23
Count
3
Service
24
srv_count
4
Flag
25
serror _rate
5
src_bytes
26
srv _serror_rate
6
dst_bytes
27
rerror_rate
7
Land
28
srv_rerror _rate
8
wrong_fragment
29
same_srv _rate
9
Urgent
30
diff_srv_rate
10
Hot
31
srv_diff_host_rate
11
num_failed_login
32
dst_host_count
12
logged_in
33
dst_host_srv_count
13
num_compromised
34
dst_host_same_srv_rate
14
root_shell
35
dst_host_diff_srv_rate
15
su_attempted
36
dst_host_same_src_port_rate
16
num_root
37
dst_host_srv_diff_host_rate
17
num_file_creation
38
dst_host_serror_rate
18
num_shells
39
dst_host_srv_serror_rate
19
num_access_files
40
dst_host_rerror_rate
20
num_outbound_cmds
41
dst_host_srv_rerror_rate
21
is_host_login
 
 
 
UNSW-NB15 Dataset: The UNSW-NB15 dataset is another benchmark dataset for network intrusion detection. It contains a diverse range of attack scenarios and normal network traffic data, providing a comprehensive evaluation environment for the IDS. The dataset includes features such as source IP, destination IP, source port, destination port, and protocol type, which are essential for intrusion detection.
Table 2: Categorization of UNSW-NB 15 dataset features
S. No
Category Name
Description
1
Flow Feature
The identifier attributes between hosts such as client-to-serve or server to-client.
2
Basic Features
The attributes that characterize the connections of protocols.
3
Content Features
The attributes of TCP/IP and also contain some attributes of http services
4
Time Features
The attributes of time such as round trip time of TCP protocol start/end packet time arrival time between packets etc.
5
Additional Generated Features
General purpose features (from number 36 - 40): Own purpose features which to care for the protocols service.
Connection features (from number 41- 47): Built based on the chronological order of the last time feature.
6
Labeled Features
The label of the record.
 
Proposed Methodology
Figure 2: Proposed Methodology Workflow
The proposed methodology involves the following steps:
  1. Hierarchical Clustering Algorithm:
  1. Artificial Neural Network-Based Intrusion Detection System:
  1. Secure Data Transmission Protocols:
  1. Performance Evaluation:
FINDINGS AND ANALYSIS
Wireless sensor networks (WSNs) are becoming popular due to their simplicity, flexibility, and scalability, making them suitable for various applications. These networks consist of tiny sensors scattered in an area to monitor environmental phenomena. The sensors communicate with the Internet of Things (IoT) to screen and document environmental data. However, they have significant limitations, including restricted energy (battery), memory space, communication range, and limited capacity for managing energy, storage, transmission, and data processing. To address these limitations, a machine learning-based approach for detecting intrusions and efficiently utilizing energy has been proposed. This involves selecting cluster heads using a secure clustering protocol to optimize energy usage. The proposed method was implemented and tested using MATLAB software, employing the NSLKDD and UNSW-NB15 datasets for intrusion detection.
Table 3: Simulation Variables for Analysis
Simulation variables
Values
Number of sensor nodes
100
Simulator
Mat lab R2020b
Energy
10 joules
Time for
30sec
Channel type
Wireless
Dimension
500m×500m
 
Preparing the Network
Initially, an area of 500 × 500 m² is chosen for deploying the sensor nodes. Around 100 sensor nodes are randomly placed in this area without any specific pattern, and all sensor nodes are static. The figure 3 depicts the node deployment in the monitoring location.
Figure 3: Deployment of Nodes
Once the deployment of sensor nodes is complete, cluster formation begins. A threshold is defined using the consumed energy ratio, and an enhanced random number is generated. Based on these, cluster heads are chosen. These cluster heads then advertise to other sensors that they have been chosen. Depending on the signal strength from the cluster head node, other nodes decide to join the cluster head, forming six different clusters as illustrated in the figure 4.
Figure 4: Cluster Formation
Efficient Energy Utilization
The proposed method's energy efficiency is evaluated using various metrics, and the results are summarized in table 4 and illustrated in the figures below.