A Systematic Literature Review on Outlier Detection in Wireless Sensor Networks

Outlier detection is an important task for intrusion detection and fault diagnosis in wireless sensor networks (WSNs). Outliers in sensed data may be caused due to compromised or malfunctioning sensor nodes. In this paper, we propose a centralized and a distributed approach based on the principal component analysis (PCA) for outlier detection in WSNs. In the distributed approach, we partition the network into multiple groups of sensor nodes. Each group has a group head and several member nodes. Every member node uses a fixed- width clustering algorithm and sends a description of its local sensed data to the group head. The group head then applies a distributed PCA to establish a global normal pattern and detect outliers. This pattern is periodi- cally updated using weighted coefficients. We compare the performance of the centralized and distributed approaches based on the real sensed data collected by 54 Mica2Dot sensors deployed in Intel Berkeley Re- search Lab. The experimental results show that the distributed approach reduces both communication over- head and energy consumption, while achieving comparable accuracy.

Data reliability in Wireless Sensor Networks (WSNs) has a substantial influence on their smooth functioning and resource limitations. In a WSN, the data aggregated from clustered sensor nodes are forwarded to the base station for analysis. Anomaly Detection (AD) focuses on detecting outlier data to ensure consistency during data aggregation. As WSNs have critical resource limitations concerning energy consumption and sensor node lifetime, AD is supposed to provide data integrity with minimum energy consumption, which has been an active research problem. Hence, researchers are striving for methods to improve the accuracy of data handled with a concern on the constraints of WSNs. This paper introduces a Feed-forward Autoencoder Neural Network (FANN) model to detect abnormal instances with improved accuracy and reduced energy consumption. The proposed model also acts as a False Positive Reducer intending to reduce false alarms. It has been compared with the other dominant unsupervised algorithms over robustness and other significant metrics with real-time datasets. Relatively, our proposed model yields an improved accuracy with fewer false alarms thereby supporting a sustainable WSN.

The rapid growth of the Internet and information technologies and also the diminution of the price of hardware components like wireless sensors leads to the fast growth in Wireless Sensor Network (WSN). The WSN is a bunch/group of sensors that are located across the area, in order to monitor and determine environmental conditions such as temperature, humidity, vibrations, wind, water levels, pollution levels. WSNs are susceptible to major attacks like Denial of Service (DOS), Wormhole attack, Sinkhole attack, etc. The main threat to the WSN is because of the broadcast nature of the nodes in the network. Therefore, the security of WSNs is the essential part that must be done. Hence, to overcome these problems or threats, we are trying to detect it using AI technology. With the expanding fields of Machine Learning and Deep Learning, we can apply various algorithms in order to classify different types of attacks. Once we detect the attack properly, we can prevent it accordingly. We are using WSN-DS. It has 4 classes of attacks which are Grayhole, Blackhole, TDMA(Scheduling), and Flooding which comes under the category of DOS attacks. In this chapter, we have analysed and compared the accuracies of 5 main machine learning classification algorithms. Also, we have analysed the 1 deep learning algorithm. The ANN (Artificial Neural Network) and 5 machine learning algorithms have been trained on the dataset. Furthermore, we have used K-fold cross-validation to get even more accurate predictions. After analysis of these algorithms, we came to the point that, the Machine Learning algorithms like Random Forest, Support Vector Machines and Deep Learning algorithm, namely, Artificial Neural Network can help us for detecting intrusions in the system or network. This will help researchers in designing their own machine learning model on top of our suggested model.

Wireless Sensor Networks (WSNs) are essential for real-time monitoring and control, but sensor faults can compromise data integrity and reduce the reliability of decisions. Supervised fault detection techniques typically require large amounts of labeled data, which are often unavailable in real-world scenarios. This paper proposes a data-efficient, unsupervised fault detection method that combines an LSTM-Attention Autoencoder with a dynamic thresholding mechanism. The model captures temporal patterns in normal sensor behavior and detects anomalies through context-aware reconstruction error analysis. Unlike conventional static thresholds, the dynamic approach adapts to changing sensor conditions, enhancing robustness in noisy or drifting environments. Experimental results on a synthetic dataset generated from real temperature sensor readings show that the proposed method achieves an F1-score of 98.90% and a recall of 99.24%, outperforming baseline methods including plain autoencoders, LSTM autoencoders, isolation forest, and one-class SVM. These findings demonstrate the model's effectiveness for real-time, label-free fault detection in resource-constrained WSNs.

How to sense and monitor the environment with high quality is an important research subject in the Internet of Things (IOT). This paper deals with the important issue of the balance between the quality of target detection and lifetime in wireless sensor networks. Two target-monitoring schemes are proposed. One scheme is Target Detection with Sensing Frequency K (TDSFK), which distributes the sensing time that currently is only on a portion of the sensing period into the entire sensing period. That is, the sensing frequency increases from 1 to K. The other scheme is Target Detection with Adjustable Sensing Frequency (TDASF), which adjusts the sensing frequency on those nodes that have residual energy. The simulation results show that the TDASF scheme can improve the network lifetime by more than 17.4% and can reduce the weighted detection delay by more than 101.6%.

To address the problem of unsupervised outlier detection in wireless sensor networks, we develop an algorithm that (1) is flexible with respect to the outlier definition, (2) works in-network with a communication load proportional to the outcome, and (3) reveals its outcome to all sensors. We examine the algorithm’s performance using simulation with real sensor data streams. Our results demonstrate that the algorithm is accurate and imposes a reasonable communication load and level of power consumption.

Since the nonstationary distribution of the detected objects is general in the real world, the accurate and efficient outlier detection for data analysis within wireless sensor network (WSN) is a challenge. Recently, with high classification precision and affordable complexity, one-class quarter-sphere support vector machine (QSSVM) has been introduced to deal with the online and adaptive outlier detection in WSN. Regarding the one-sided consideration of optimization or iterative updating algorithm for QSSVM model within current techniques, we have proposed comprehensive outlier detection methods in WSN based on the QSSVM algorithm. To reduce the complexity of optimization algorithm for QSSVM model in existing techniques, a fast optimization algorithm based on average Euclidean distance has been developed and employed to the comprehensive outlier detection method. Evaluated by real and synthetic WSN data sets, our methods have shown an excellent outlier detection performance, and they have been proved to meet the requirements of online adaptive outlier detection in the case of nonstationary detection tasks of WSN.

Event detection in wireless sensor networks is a sophisticated method for processing sampled data directly on the sensor nodes, thereby reducing the need for multihop communication with the base station of the network. In contrast to application-agnostic compression or aggregation techniques, event detection pushes application-level knowledge into the network. In-network event detection - especially the distributed form involving multiple sensor nodes - has thus an exceptional potential to increase energy efficiency, thus prolonging the lifetime of the network. In this article, we summarize recently proposed system architectures and algorithms employed for event detection in wireless sensor networks. On the example of the AVS-Extrem platform, we illustrate how energy-efficient event detection can be implemented through a combination of custom hardware design and distributed event detection algorithms. We then continue to present a brief evaluation of the detection accuracy and the energy consumption that is achievable by current systems.

The quality of dataset measured and collected by wireless sensor networks (WSN) is often affected by noise and error that are inherent to resource-constrained sensor nodes. The affected data points deviating from the normal pattern are termed as outlier(s). However, detected outlier can be a result of the occurrence of an actual event. Outlier detection techniques developed for WSNs perform binary labelling of the data points and does not indicate the context stating whether the outlier is the result of an actual event or the noise/error. This paper proposes a contextual outlier detection framework specifically designed for WSNs named as in-network contextual outlier detection on edge (INCODE). The proposed framework also estimates the degree of outlierness associated with the detected outlier(s) to provide better insight into the measured data point. Algorithms used in INCODE are designed around the edge computing concept to minimize the communication and computational complexities to make it suitable for resource-constrained WSN. The results suggest an impressive 98% accuracy in identifying the context of the outlier(s). The low communication and computational complexity suggest INCODE’s suitability for resource constrained WSNs.

Wireless sensor networks are gaining more and more attention these days. They gave us the chance of collecting data from noisy environment. So it becomes possible to obtain precise and continuous monitoring of different phenomenons. However wireless Sensor Network (WSN) is affected by many anomalies that occur due to software or hardware problems. So various protocols are developed in order to detect and localize faults then distinguish the faulty node from the right one. In this paper we are concentrated on a specific type of faults in WSN which is the outlier. We are focus on the classification of data (outlier and normal) using three different methods of machine learning then we compare between them. These methods are validated using real data obtained from motes deployed in an actual living lab.

Due to the major operating restrictions, ensuring security is the fundamental problem of Wireless Sensor Networks (WSNs). Because of their inadequate security mechanisms, WSNs are indeed a simple point for malware (worms, viruses, malicious code, etc.). According to the epidemic nature of worm propagation, it is critical to develop a worm defense mechanism in the network. This concept aims to establish novel malware detection in WSN that consists of several phases: “(i) Preprocessing, (ii) feature extraction, as well as (iii) detection”. At first, the input data is subjected for preprocessing phase. Then, the feature extraction takes place, in which principal component analysis (PCA), improved linear discriminant analysis (LDA), and autoencoder-based characteristics are retrieved. Moreover, the retrieved characteristics are subjected to the detection phase. The detection is performed employing combined shallow learning and DL. Further, the shallow learning includes decision tree (DT), logistic regression (LR), and Naive Bayes (NB); the deep learning (DL) includes deep neural network (DNN), convolutional neural network (CNN), and recurrent neural network (RNN). Here, the DT output is given to the DNN, LR output is subjected to CNN, and the NB output is given to the RNN, respectively. Eventually, the DNN, CNN, and RNN outputs are averaged to generate a successful outcome. The combination can be thought of as an Ensemble classifier. The weight of the RNN is optimally tuned through the Self Improved Shark Smell Optimization with Opposition Learning (SISSOOL) model to improve detection precision and accuracy. Lastly, the outcomes of the suggested approach are computed in terms of different measures.

Widely used computer systems have always become a target for different type of attacks. With the specifications of easiness in deploying and setting up low cost and low power consumption systems, Wireless Sensor Networks (WSN) have also become a specific target for attackers. The WSNs, which consist of sensor nodes that sends information they perceive and forward to a base station/sink, cannot be directly protected by the traditional security precautions because of low resources of nodes, such as memory and computing capability. Hence, it is tried to find out different methods for intrusion detection in WSNs. This paper surveys frameworks for intrusion detection in WSNs by using a swarm optimization algorithm as ant colony approach.

Rare events detection is one of the main applications in Wireless Sensor Networks (WSN) and is currently a central concern of a vast literature. Compressed Sensing (CS) theory has been proved to be quite adapted to this objective. Although this is not the first work on applying CS to sparse events detection in WSN, it is the first to highly justify the validity of the targets detection and counting problem formulation. In order to enhance the CS recovery capacity in WSN, this work considers an approach based on a coherence reduction of the sensing matrix premised on the transmitted power control (PC). Simulation results prove that, under the constraint of equal power consumption, the detection and counting performance is improved when the proposed power control scheme is employed compared to the case without PC.

Outlier detection is a key component for the proper functioning of a wireless sensor network. It helps in identifying malicious events taking place in the field and faulty sensors. This paper presents a distributed outlier detection technique that separates outlying cases from normal data locally at sensors and sends only the summarized information of data to sink nodes to minimize the communication overhead in the resource-constraint wireless sensor network. The proposed solution employs variance of normal data as a measure to identify outliers. The performance of the proposed method has been evaluated on real-world datasets collected by WSN deployment projects. Experimental results prove the effectiveness of the proposed method in terms of higher outlier detection rate, lower false alarm rate and lower communication overhead as compared to the simulated centralized approach.KeywordsData varianceOutlier detectionWireless sensor networks

Outlier detection (OD) constitutes an important issue for many research areas namely data mining, medicines, and sensor networks. It is helpful mainly in identifying intrusion, fraud, errors, defects, noise and so on. In fact, outlier measurements are essential improvements to quality of information, as they are not conforming to expected normal behaviour. Due to the importance of sensed measurements is collected via wireless sensor networks, a novel OD process dubbed density‐based spatial clustering of applications with noise (DBSCAN)‐OD has been developed based on the algorithm DBSCAN, as a background for OD. With respect to the classic DBSCAN approach, two processes have been jointly combined, the first of computing parameters, while the second concerns class identification in spatial temporal databases. Through both of these modules, one is able to consider real‐time application cases as centralised in the base station for the purpose of separating outliers from normal sensors. For the sake of evaluating the authors proposed solution, a diversity of synthetic databases has been applied as generated from real measurements of Intel Berkeley lab. The reached simulation findings indicate well that their devised method can prove to help effectively in detecting outliers with an accuracy rate of 99%.