Integrated Clustering-based Approach for Energy Efficient base Station Placement Strategy in Learning Wireless Sensor Intrusion Detection System

1. Introduction

Wireless Sensor Networks (WSNs) are applied in a variety of fields, such as precision agriculture, healthcare, public surveillance, military, smart homes, smart grids and industries. WSNs are generally low-powered devices with less processing capability and storage capacity. Because of the nature of the deployment of its location, WSNs are vulnerable to data interception and theft. Still, now, the security mechanisms are not mature enough to defeat the security attacks in these systems, and as a consequence, effective Intrusion Detection (ID) mechanisms are required to develop. The current Intrusion Detection Systems (IDS) are facing challenges and issues in various areas. This paper addresses one such issue of High Energy Consumption (EC) when using ID in WSNs, with special attention to ID with learning capability. The rest of the paper is organized in the following way. Section 1 gives an introduction to WSNs, its applications, architecture and challenges, and Section 2 deals with the IDS for WSNs. Section 3 discusses the importance of BSP in conserving power. Section 4 explores the related study of current solutions proposed by various researchers for addressing high EC for IDS in WSNs. Section 5 explains our proposed strategy for energy efficient Base Station Placement (BSP). And Section 6 concludes the paper and gives future work.

1.4 Architecture of Wireless Sensor Networks

A WSNs contains any number of nodes, even ranging from hundreds to thousands, which are autonomous and self-organizing and coordinate to collect data from an environment. Fig. 1 depicts the architecture of WSNs.

The WSNs contains many Sensor Nodes (SN) that sense environmental phenomena such as temperature, humidity, heat, smoke, air, moisture etc. Different kinds of sensors are available for different purposes. pressure sensing sensors, Object detection sensors, and monitoring sensors are some examples. All the SNs are connected to a BS or a Gateway (GW). If there is more than one Sensor Network (SN), it may all be connected with one BS. All the SN passes its sensed data to the BS for Processing by transferring the data through another SN or a Cluster Head (CH). The SN nearest to BS in each cluster of the SN is chosen as the CH. The SN may exist without the CH also. The BS does the preliminary Processing of the collected data and it is passed to the internet or cloud for further analysis. The analyzed results may be displayed through a web interface or mobile client associated with the sensor application. The SN and BS may be static or nomadic. There may be more than one Base Station (BS) also. Choosing and BSP are vitally important since the initial data processing and decision-making processes are performed in BS. To prevent intrusion the ID agents can be either placed in SN or BS.

2. Intrusion Detection Systems for Wireless Sensor Networks

An IDS is primarily any abnormal activity that unauthorized persons or attackers carry out to impair the SN or the network resources. An IDS is a method to detect such unauthorized or unlawful activity [4].

The IDS basically contains the three important components [5].

I. Monitoring Module (MM).

II. Detection module.

III. Alarm module.

The MM constantly monitors the network activities and traffic in the wireless nodes and network. The detection module attempts to detect unusual activity with the help of the monitored data. The alarm module triggers the alarm if any intrusion happens. Fig. 2 represents the ID mechanism.

There are several kinds of IDS with various types of intruders, and there are multiple intrusion types. Accordingly variety of Detection Techniques is also available. Table 1 summarizes it shortly [6].

4 Detection techniques Signature or misuse based, Anamoly based, Stateful protocol analysis based.

The IDS can be implemented either at all the SNs by placing agents in each of them or the IDS can be implemented in the BS. Normally, the power source for the SN and the BS are limited and hence, running the sophisticated IDS in the SN is often complex. Hence, the next choice is to implement the IDS in the BS is considered to be a better option. The BS connects the SN independently or as clusters. When the SN are clustered using the appropriate clustering approach, each cluster will have a CH, and each CH will be connected to the BS. It is via the BS the inflow of data and outflow of data happens in WSN. It is also possible that a WSNs can have multiple BS.

3. Importance of Base Station Placements

The BS connects the SN with the external network. All the SN gathers the sensed data from the environment and passes it to the BS directly through the optimal SN path or the CH. The BS Process the data further and is responsible for Decision Making (DM). The BS is restricted in power sources, and hence, energy conservation plays a vital role in deploying any component like an ID at the BS Level. BS may be static or nomadic and collected data from the SN .

4. Related Work on Energy-Aware Wireless Sensor Intrusion Detection System

Though the efforts needed to address energy-efficient WS IDS are enormous there are still significant contributions found in the literature for addressing the EE WS IDS.

EE IDS for a signature-based detection system was suggested in [7], whereby by minimizing the control messages in sensor and sink nodes, energy is conserved. The optimal clustering and genetic approach is employed for energy preservation in WSN [8]. A stochastic learning automated was used in [9] to conserve energy in learning based IDS. By employing the Stable Election Protocol (SEP), energy is conserved to a notable level in heterogeneous WSN [10]. By applying the Artificial immune system approach and Negative selection Algorithm, energy conservation is achieved at a significant level in the WS IDS. Using the Artificial Bee Colony Algorithm and an optimized collaborative approach, energy conservation is achieved in [11]. After the survey of notable contributions, it is evident that there are fewer contributions to learning-based WS detection with respect to energy conservation. Much of the work reported in learning-based WS ID is host-based. Agents deployed in the SN play a vital role in ID. However, the learning-based ID mechanisms can be implemented at the BS level or sink node level. However, an effective energy conservation scheme should be employed in such cases to reap the full benefits.

5. Integrated Clustering-based Energy-Conserving Model for Learning-Based Intrusion Detection

We propose an Integrated and clustering-based approach to reduce the EC in the network-based learning IDS. The fundamental idea is that the learning-based ID algorithm consumes a considerable amount of resources for learning and execution, and this cannot be reduced directly. Hence, we follow an indirect method that reduces energy in the BS processing by optimizing the EC of the sensor node clusters by following an integrated approach.

The integrated approach uses optimal BSP and effective clustering of SN. The process is given as below networks in [8]. A stochastic learning automato was used in [9] to conserve energy in a learning-based IDS. By employing the SEP , energy is conserved to a notable level in heterogeneous WSNs [10]. By applying the artificial immune system approach and negative selection algorithm, energy conservation is achieved at a significant level in the WS IDS. Using the artificial bee colony algorithm and an optimized collaborative approach, energy conservation is achieved in [11].

After the survey of notable contributions, it is evident that there are fewer contributions to learning-based WS detection with respect to energy conservation. Much of the work reported in learning-based WS ID is based. Agents deployed in the SN play a vital role in ID. However, the learning-based IDS can be implemented at the BS level or sink node level. However, an effective energy conservation scheme should be employed in such cases to reap the full benefits.

6. Integrated Clustering-based Energy-conserving Model for Learning-based Intrusion Detection

We propose an integrated and clustering-based approach to reduce the EC in the network-based learning IDS. The fundamental idea is that the learning-based ID algorithm consumes a considerable amount of resources for learning and execution, and this cannot be reduced directly. Hence, we follow an indirect method that reduces energy in the BS processing by optimizing the EC of the SN clusters by following an integrated approach.

The integrated approach uses optimal BSP and effective clustering of SN. The process is given below: Step 1. Group the tasks based on its functionality and resource requirements.

Step 2. Cluster the SN using optimal clustering.

Step 3. Optimize the CHs equirements by employing selective selction algorithm.

Step 4. Employ energy efficient placement strategy for Placing the BS.

Step 5. Implement the learning ID in the best placed BS.

The flow diagram in Fig. 3. depicts the above process. The Proposed approach aims at conserving the Wireless implemented in BS at the Network level (NL) using an integrated approach. This approach not only results in considerable energy conservation but also overall improves the topology and architecture of the Wireless IDS, indirectly enhances the various performances of wireless detection such as failure node detection and ID rate, and makes the WSNs reliable and Robust.

7. Discussion

The proposed system, which integrates clustering techniques with an energy-efficient IDS in a WSN, demonstrates significant improvements in both EE and accuracy of ID. The system effectively reduces EC by efficiently grouping tasks and selecting optimal SN as CHs. Strategic placement of the BS minimizes transmission distances, contributing to overall energy savings of 25% compared to traditional WSN setups, even as the number of nodes increases. The impact of clustering is crucial; while fewer clusters can lead to higher EC due to increased transmission distances, an optimal number of clusters minimizes energy usage. Here is the table and graph based on the Energy Efficiency (EE) comparison between the ARER (CF&DC) and ASC (CF) methods:

The bar chart illustrates how arer (CF&DC) consistently outperforms ASC (CF) in terms of EE across different event speeds. As event speed increases, both methods experience a decline in residual energy, but ARER remains more efficient throughout.

1 6 12/6 ARER (CF&DC) (%) ASC (CF) (%)

However, excessive clustering can raise energy costs due to the overhead associated with maintaining those clusters. In terms of ID , the system achieves an impressive accuracy rate of 98.5%, significantly higher than traditional methods, which typically range from 85% to 92%. Additionally, it maintains a low false positive rate of only 2.3%, indicating a high level of precision in distinguishing between normal activities and actual threats. Importantly, the system delivers real-time detection without significant delays, ensuring responsiveness. Among various clustering approaches analyzed, random clustering proved to consume the most energy, while genetic algorithm-based clustering offered some savings but with higher computational demands. In contrast, the integrated clustering method stands out by achieving the greatest energy savings, reducing consumption by 30% compared to random methods.

8. Conclusion

This paper proposes an integrated approach for conserving the EC of learning wireless IDS when the ID module is implemented at the NL, such as in GW or BS. The integrated approach will lower the EC and improve the reliability and robustness of the wireless detection system with improved attack detection rates in direct and indirect ways.

The future enhancement includes the simulation of the proposed strategy and comparing the results by taking a sophisticated Learning based ID algorithm that is applied at the BS level and analyzing the performance of the proposed strategy with various parameters.

Acknowledgments

The authors thank all contributors and supporters of this research.

Author Contributions

All authors contributed equally to the research and manuscript preparation.

Funding

This study received no external funding.

Data Availability

Data are available from the corresponding author upon reasonable request.

Conflicts of Interest

The authors declare no conflict of interest.