In 2019, Meng et al. [18] suggested a spoofing generator based on a software-defined receiver (SDR) with vector tracking. In 2019, Semanjski et al. [19] developed an intelligent technique to identify possibly manipulated GNSS signals by tracking the cross-correlation of several GNSS measurements. In 2022, Shang et al. [20] developed a brand-new time synchronization attacks (TSA) mitigation algorithm based on a single antenna. Initially, the parameters of the real and fake signals were determined using a multi-correlator estimator. In 2021, Singh et al. [21] explored a cutting-edge method for spoofing attack detection and mitigation. To implement the suggested algorithm in the Android application, the OpenStreetMap dataset are considered. In 2020, Schmidt et al. [22] suggested an approach for identifying and classifying GNSS spoofing in single-antenna receivers. In order to reduce false alerts, the researcher applied a threshold. In 2022, Pardhasaradhi et al. [23] suggested the extended Kalman filter (EKF) architecture, which is utilized by the GNSS receivers to estimate their PVT. In 2022, Blais et al. [24] put forth a unique framework for GNSS multipath prediction. The process was initiated by creating datasets that are processed by the CNNs. Grayscale 2D images were created by mapping correlations between the local replica grid and the received signal. In 2023, Nayfeh et al. [25] gave a developed model for the identification and categorization of unmanned aerial vehicle (UAV) GPS spoofing.

Due to the broad application, GNSS has presently grown and requires the minimum cost. The occurrence of GNSS is used in phased measurement unit (PMU), emergency service, power grid, and financial transactions. The spreading of GNSS receivers in smart cars and phones allows the general to have access to our localization. Moreover, the position quality determined by the GNSS equipment minimizes based on the received signals. This form of degradation is formed to find the signal generation system. SDR [18] has the capacity to track the original code accurately in the urban environment. It easily changes the spoofing into carrier or code but requires a mature vector tracking framework. Support vector machine (SVM) [19] has the maximum capacity in helping the GNSS signal spoofing detection and is also performed under huge datasets. However, it has slight variation in the validation result. The TSA mitigation process [20] requires only a single antenna and is cost-effective. However, it has the chance to form false time information while transmitting signals. Genetic algorithm [21] has the capacity to escape from local optimal problems, but it is computationally more expensive. Least absolute shrinkage and selection operator (LASSO) [22] supports for enhancing the automatic peak discrimination and generates robust outcomes. However, it requires an optimization algorithm to solve the computational requirement. The generalized likelihood ratio test (GLRT) [23] helps to identify the spoofing attack and attains the optimal rate of overall analysis. However, it requires recalculation for accurate results. CNN [24] has high multipath detection power and takes less time for parameter estimation. However, it requires data features in prior. Machine learning (ML) [25] gives a solution for detecting GPS spoofing and supports in identifying real-time detection. However, it takes more time and resources. Therefore, to solve the difficulties, this paper proposed the ML technique for detecting spoofing in the GNSS framework.

The intellectual transference system and the locality-aided services are successfully operated and implemented in the market by the PVT application. The PVT effectively attains the information via the GNSS. However, the GNSS is highly susceptible to several types of attacks. The spoofing is the most common attack in the GNSS system. In this process, the authentic GNSS signals are alleviated by the spoofer and convey the wrong information to the victim. In order to make the safe navigation process, GNSS spoofing detection is necessary. The conventional GNSS spoofing detection approaches require too much cost for the spoofing detection process. Moreover, the existing approaches for detecting GNSS spoofing are not suited for large-scale deployment. The GNSS spoofing attack can occur at any time and place within a limited period of time, which cannot be monitored by the existing approaches. The existing model does not identify the spoofed signal that may provide wrong data to civil and military applications. The conventional techniques for spoofing attack detection do not identify the presence of a spoofer along the quality of the signal cannot be monitored by the traditional techniques. Further, the existing approaches utilize the autocorrelation function to identify the spoofed signal, which may raise the time required for the GNSS spoofing. The planned and the inadvertent interference in the GNSS are not identified by the traditional approaches. The power level fluctuation in the GNSS signal is not handled by the traditional techniques. Moreover, the hardware requirement for the GNSS spoofing is quite high for the traditional approaches. To counter the challenges of the traditional approaches, the hybrid ML approach is developed for the GNSS spoofing.

A hybrid GNSS spoofing detection and mitigation framework is developed to find the spoofing signal in the GNSS and mitigate it to offer PVT services for civil and military services. Moreover, the developed model is applied in the GNSS receiver to find and eliminate the false signal induced by the spoofer to provide proper navigation in the country [26]. Besides, the suggested ML system identifies the attack modes of the spoofer and its corresponding properties with fewer hardware requirements [27]. In order to begin the GNSS spoofing detection and mitigation process, the needed data are taken from online sources. Further, the collected data is subjected to the feature extraction procedure. The feature of the data is retrieved using two approaches like t-SNE and the PCA. The feature extraction enhances the spoofing detection efficiency of the ML methods. The high-dimensional data is converted into the lower-dimensional data with the help of the t-SNE. Moreover, the most significant features in the data are represented using the t-SNE structure. Similarly, the PCA approaches extract the principal components that are called as features and it is available in more understandable form. The features attained from the t-SNE and PCA are integrated and passed to the HABMLP. The HABMLP is formed by integrating the two machine learning structures BL and MLP. Both of the hybrid structure works together and produce higher accuracy in the GNSS spoofing detection and mitigation process. Both BL and MLP examine the features obtained from the t-SNE and PCA to get accurate results. Here, the BL model adopts the Bayes rule [28] to get the detection result and similarly, the MLP structure analyses the features using its complex layer. During this process, the suggested EOOA optimizes the hidden and epoch count of the BL as well as the steps per epoch of MLP to obtain the accurate prediction result. Once the HABMLP detects the GNSS spoofing, the mitigation process is executed by the same HABMLP, here the alarm is produced as a warning sign once the spoofer can tamper the GNSS signal. To this end, the experimental analysis is performed on the output of the developed hybrid ML-based structure to confirm its effectiveness. The structural representation of the HABMLP-based GNSS spoofing recognition and mitigation framework is illustrated in Fig. 1.

The dataset utilized in the GNSS spoofing attack detection and mitigation approach is elaborated in the below section. Here, the collected dataset is used to validate the GNSS spoofing framework performance. The consideration of these datasets splits into two equal halves and it represents datasets 1 and 2 according to the collected data. This dataset considers publicly available data. The detailed description of this dataset is mentioned below.

Dataset 1 & Dataset 2: Galileo open service navigation message authentication (FGI OSNMA) datasets are taken from the link of https://etsin.fairdata.fi/dataset/09dc5c1b-933d-4efd-aa66-be2c07fab3b3/data with access: 2024-04-09. Galileo E1 and GPS L1 C/A signals in raw GNSS in-phase and quadrature (I/Q) data are presented in this dataset. Stereo-dual-band GNSS front-end is used to gather I/Q data samples. The open sky in this dataset helps to examine the performance of OSNMA at no spoofing condition. The position of this dataset is calculated by 60.182°N, 24.828°E geodetic grade receiver with 47.248 m altitude. The collected data is regenerated with LabSat wide-band 3 and captured through the GNSS front-end. This dataset is highly important to validate the OSNMA-based spoofing detection effectiveness in a real-world scenario. The actual data for datasets 1 and 2 is shown in Table 1.

The parameters utilized in the hybrid ML-based GNSS spoofing detection and mitigation framework are elucidated in Table 2.

The gathered data from the online resources is represented as $ G_m^r $ and $ m = 1,2,\cdots, M $, here, the total quantity of the gathered data from the online resources is represented as $ M $.

The EOOA is a new optimization algorithm, which is modeled from the baseline osprey optimization algorithm (OOA) [29] by upgrading the random parameter. It is suggested to maximize the efficacy of the hybridized expert systems in the GNSS spoofing recognition and mitigation process. The parameters of the BL and MLP are optimized by the EOOA to enhance the precision and accuracy of the developed model. The explored EOOA optimizes the steps per epoch of the MLP as well as the number of epochs and epoch count of the BL to achieve effective performance in the GNSS spoofing detection and mitigation process. The OOA is the meta-heuristic algorithm and it is intended to consider the simulation activities of the OOA [30]. The factual world optimization trouble is solved by the OOA. At the time of the searching process, the exploitation and the exploration phases can be handled by the OOA. The convergence and the precision rate of the OOA are very poor. Additionally, the control engineering concerns are not solved by the OOA [31], and the time consumption of the OOA while maintaining the balance between the exploitation and the exportation phase is too high. Considering the above hitches, the new EOOA is developed. The accuracy and precision of the GNSS spoofing and mitigation model are enhanced using the EOOA. The worst, and best fitness and the number of population $ O $ play the most crucial role for upgrading the random number $ {s_{j,k}} $. The mathematical term for calculating the random number $ {s_{j,k}} $ is provided as follows:where the random number is indicated as $ {s_{j,k}} $. The best and worst fitness is denoted as $ b_t $, and $ w_t $, the number of population is elucidated as $ O $. The developed EOOA solves the control engineering concerns and convergence as well as the precision rate is also enhanced by the new EOOA. Further, the accuracy and the precision rate of the GNSS spoofing are enhanced by the new EOOA.

The pseudocode of the EOOA is stated in Algorithm 1.

The size of the entire data is condensed by the powerful technique called as t-SNE. The t-SNE is used to analyze the gathered data’s structure. If the linear boundary of the collected data is too complex, the t-SNE is adopted which analyzes the entire data point and retrieves the most important feature. The collected data $ G_m^r $ is given into the t-SNE [32] for obtaining the relevant features from the data. The t-SNE is used to take out the feature from the collected data. The t-SNE has a high multi-dimensional scaling property. The t-SNE is applied between the data points of the gathered data to project the Euclidean distance among those points. The coordination between the data points $ {y_k} $ and $ {y_j} $ is identified with the aid of conditional probability $ {Q_{k|j}} $ and it is represented as follows:

In the original feature space, this probability is arithmetically signified as indicated as follows:

Thus, the dimension of the data is represented as $ o $, the vicinity of the data points is identified with the help of the perplexity constant in the t-SNE and it is expressed as follows:where $ I\left( {{Q_j}} \right) $ is the Shannon entropy.

The pairwise distance between the data points can be used to adjust the variance $ \wp _j^{} $. Similarly, the low dimensional probability $ {r_{jk}} $ can be expressed as follows:

The input data $ {y_j} $ can be converted into lower dimensional data $ {n_j} $ with the aid of t-SNE. The data can be effectively visualized with the help of the t-SNE. The features obtained from the collected data $ G_m^r $ using t-SNE are denoted as $ d_m^{{\mathrm{tsne}}} $.

The principal component analysis (PCA) [33] approach reduces the dimension of the data and converts it into a lower dimension by projecting the data in a geometric way. The features obtained from the PCA are very easy to learn so it gives accurate results. The relevant features from the gathered data are hauled out using the PCA [33]. From the set of data, the dominant features also called as principal components are extracted using the technique named as PCA. The data are projected on the ortho-usual subspace by applying the linear transform. The collected data can be minimized by the PCA and very useful information for the spoofing process is available in the retrieved features [34]. The collected data is converted into a column vector $ {\nu _o} $ in order to attain the Eigen data. Based on the utilization of the variables, the length of the column vector is varied. The matrix array $ \nu $ with the dimension of $ N \times O $ is developed for the input observation $ N $ and it is given as follows:

The average data $ \kappa $ is represented as follows:

The average data and the training data are subtracted to get the difference data and it is represented as follows:

Later, the orthogonal Eigenvector and covariance $ D $ of the PCA is identified and it is stated as follows:

The computation process can be lowered by considering the Eigenvector represented as $ {w_j} $.

The Eigenvector of the covariance $ D $ can be identified bywhere the Eigen data is represented as $ {v_j} $. Based on the large Eigenvalue, the required Eigen data $ N $ could be selected. For any data, the principal component is defined as follows:

Based on the Eigenvector, the value $ {x_l} $ can be mapped with the axis. The values are signified as the feature of the data that can be employed in the spoofing detection process. The features from the PCA are represented as $ d_m^{{\mathrm{pca}}} $.

Both of these feature like $ d_m^{{\mathrm{tsne}}} $ and $ d_m^{{\mathrm{pca}}} $ are fused and it is passed to the HABMLP for spoofing detection and mitigation process. Here, the fused feature is termed as $ d_m^{{\mathrm{com}}} $.

The BL model performs the detection process effectively on a limited amount of data. The fused feature $ d_m^{{\mathrm{com}}} $ is passed to the BL to achieve the first detection score. The learning process of the BL is very simple as compared to another model. The BL [35] is used to find the spoofing in GNSS. The feature obtained from the t-SNE and PCA is given to the BL for getting the first detection result. The BL works on the basis of belief revision. The previous belief can be updated if any new data is available in the network. In Bayesian learning, the learning issues are considered as Bayesian probabilistic interference. For the casual model, the hypothesis space $ I $ and the data $ e $ can be developed by the learner [36]. Later, the degree of belief can be expressed by computing the subsequent probability distribution and it is represented as $ Q(i|e) $. Based on the availability of the data and likelihood $ Q(e|i) $, the prior probabilities $ Q(i) $ can be identified. According to the Bayes rule, the multiplication of the likelihood and the prior is directly proportional to the subsequent probabilities and it is expressed as follows:

The term $ i $ indicates the hypothesis and the alternative hypothesis is signified by the term $ {i^{'}} $. The detection result from the BL is denoted as $ S_m^{bl} $. The pictorial delineation of the BL is presented in Fig. 2.

The MLP analyzes the vectors of the feature to get an accurate detection result. In MLP, the value of false alarms is lower. This structure is used for GNNS spoofing detection. The fused feature $ d_m^{{\mathrm{com}}} $ is given to the MLP to get the second detection score. The MLP [37] is one type of neural network and it is used to employ the spoofing detection process. The complex neuron structure of the brain is the main inspiration behind the development of MLP. It can be mathematically expressed as follows:where the variable $ g $ presents the commencement function, the input layer of the $ {j\text{th}} $ neuron is presented as $ {y_j} $, the $ {l\text{th}} $ perceptron output is represented as $ {Z_l} $ for the $ {j\text{th}} $ row and the $ {l\text{th}} $ column of the neuron; the weight matrix is denoted as $ {{\boldsymbol{x}}_{l\,j}} $, and for the layer $ l $, the bias value is denoted as $ {c_l} $.

The loss function can be used to lower the prediction error of the MLP. The weights of individual layers can be iteratively enhanced to lower the occurrences of detection errors. The detection error is calculated at each iteration during the training process [38]. The weight and the bias are adjusted to back-propagate the error in the MLP. The loss function of the MLP can be identified as follows:

Thus, for the data $ E $, the $ {j\text{th}} $ flow of label is represented as $ {z_j} $, and the $ {m\text{th}} $ predicted prospect of spoofing is represented as $ {\hat z_m} $. The detection result obtained from the MLP is denoted as $ S_m^{{\mathrm{MLP}}} $. The pictorial visualization of the MLP is indicated in Fig. 3.

The HABMLP is developed for the GNSS spoofing detection and mitigation process. In HABMLP, BL and MLP are hybridized to get the accurate detection result. The hybridization process is mainly used to adapt the strength of both BL and MLP in the GNSS spoofing detection and mitigation process. The BL works on the basis of the Bayes rule to get an accurate detection result. The learning capacity of the BL is very well and it can function effectively for a limited amount of data. Likewise, the MLP is a complex brain-like structure so it deeply analyzes the vectors of the features to get the accurate detection result. The detection accuracy of hybrid ML is high. The feature from the PCA and t-SNE is fused and it is given to the BL. The BL identifies the GNSS spoofing by analyzing the nature of the signal. During the GNSS spoofing, the probability values are used to train the BL. Based on the probability of existence, the properties of the GNSS signal is identified. Likewise, in the MLP, the signal space is mapped with the input layer of the MLP. Further, the hidden layer of the MLP processes the features and conveys it into the output layer. Here, the MLP detects the GNSS spoofing by collecting the fresh GNSS signal. The output layer compares the properties of the new GNSS signal and existing information to find the presence of spoofing. The detection result from the BL and MLP is averaged to get the final detection result. At the time of the GNSS spoofing detection, the EOOA is applied to fine-tune the attributes to improve the accuracy and precision value.

Once the GNSS spoofing is detected, the GNSS mitigation is carried out, here, the alarm is created in terms of a warning signal when the spoofer is entered into the GNSS and it refreshes the GNSS receiver to its initial state. The developed HABMLP-based spoofing detection cum mitigation model is used to maintain secure navigation and it can be used to provide correct PVT to civil and military applications. The objective function of the HABMLP-based spoofing detection and mitigation model is given as follows:where the purpose function is represented as $ d $, the term $ K_c^{bl} $ in $ [5,255] $ represents the count of BL’s hidden neuron, the steps per epoch of MLP is denoted as $ N_d^{{\mathrm{mlp}}} $ in $ [100,500] $ and the epoch count of BL is expressed as $ H{}_n^{bl} $ in $ [5,50] $. The precision and accuracy rates are determined fromandwhere the true and false positives are delineated as $ t $ and $ y $ correspondingly. The negative and true are expressed as $ h $ and $ g $. The structural view of the HABMLP network for GNSS spoofing recognition and mitigation is signified in Fig. 4.

The developed HABMLP-based GNSS spoofing attack recognition and alleviation approach is implemented using Python software. In this analysis process, chromosome length, maximum number of iterations, and population number are taken as 50, 3, and 10. In addition, the evaluation processes consider the existing algorithms like rat swarm optimizer (RSO) [39], dwarf mongoose optimization (DMO) [40], coronavirus herd immunity optimizer (CHIO) [41], OOA [29] and the traditional approaches such as K-nearest neighbour (KNN) [42,43], BL [35,44] and MLP [37,45] for verifying the efficacy of the suggested HABMLP-based GNSS spoofing attack detection and mitigation approach. The hardware requirements of the developed suggested HABMLP-based GNSS spoofing attack recognition framework are represented in Table 3.

The simulation indices in the HABMLP-based GNSS spoofing attack detection and mitigation approach are listed as follows:

The ROC curve examination of the EOOA-HABMLP-based GNSS spoofing recognition and alleviation model for dataset 1 and dataset 2 is shown in Fig. 5(a) and Fig. 5(b). The overall efficacy of the EOOA-HABMLP model in the spoofing detection and mitigation process is identified by the ROC investigation. For the second dataset, the suggested EOOA-HABMLP-based GNSS spoofing discovery and mitigation model attains the true positive rate of 8.16%, 8.1%, 6.12%, and 5.91% enhanced than the KNN, SVM, BL, and MLP at the false positive rate 0.4. Therefore, the performance of the EOOA-HABMLP-based method is better than the other baseline approaches.

Fig. 6(a) and Fig. 6(b) shows the confusion matrix analysis of the HABMLP-based GNSS spoofing attack detection and mitigation approach for dataset 1 and dataset 2. The developed EOOA-HABMLP method performance is summarized through the confusion matrix in terms of matrix form. From the confusion matrix analysis, the overall accuracy of the HABMLP-based GNSS spoofing attack detection and mitigation approach is acquired as 94.65% for the first dataset and 94.33% for the second dataset.

Fig. 7(a) and Fig. 7(b) represents the convergence assessment of the explored EOOA-HABMLP-based model for dataset 1 and dataset 2. In the second dataset, the suggested model’s cost function is lower than the RSO-HABMLP, DMO-HABMLP, CHIO-HABMLP, and OOA-HABMLP with 87.5%, 12.5%, 75% and 11.25% at the 20th iteration. Thus, the cost consumption of the EOOA-HABMLP-based method is lower than the existing approaches.

The algorithmic assessment of the implemented EOOA-HABMLP-based GNSS spoofing detection and mitigation model for the dataset 1 in terms of accuracy, F1-score, FDR, FNR, FPR, NPV, FPR, Precision, Recall, and Specificity is illustrated in Fig. 8(a)−Fig. 8(j). Here, the population count varies in the X axis for accurately determining the efficacy variation of the implemented method with the existing mechanisms. The algorithmic assessment of the implemented model for the dataset 2 in terms of accuracy, F1-score, FDR, FNR, FPR, NPV, FPR, Precision, Recall, and Specificity is illustrated in Fig. 9(a)−Fig. 9(j). In the second dataset, the accuracy of the EOOA-HABMLP-based GNSS spoofing detection and mitigation model is higher than the RSO-HABMLP, DMO-HABMLP, CHIO-HABMLP, and OOA-HABMLP with 5.69%, 5.18%, 5.69% and 3.14% at the 6th population value for the 1st dataset. Likewise, the type II indices like FNR and FDR of the explored EOOA-HABMLP-based GNSS spoofing detection and mitigation model are lower than other traditional frameworks. Therefore, the EOOA-HABMLP-based GNSS spoofing detection and mitigation model performance is better than the traditional algorithms.

The developed EOOA-HABMLP-based GNSS spoofing detection and mitigation method is compared with the baseline detection method for the dataset 1 in terms of accuracy, F1-score, FDR, FNR, FPR, NPV, FPR, Precision, Recall, and Specificity is illustrated in Fig. 10(a)−Fig. 10(j). In addition, the developed method is compared with the baseline detection method for the dataset 2 in terms of accuracy, F1-score, FDR, FNR, FPR, NPV, FPR, Precision, Recall, and Specificity is illustrated in Fig. 11(a)−Fig. 11(j). At the second dataset, the developed EOOA-HABMLP-based GNSS spoofing detection and mitigation model accomplished the MCC value of 21.73% enhanced than KNN, 13.04% enhanced than SVM, 11.9% enhanced than BL and 8.69% at tanh activation function. On the contrary, the other indices like accuracy, F1 score, and NPV of the EOOA-HABMLP-based GNSS spoofing detection and mitigation model are higher than the existing methods. Therefore, it proves the efficacy of the EOOA-HABMLP in the GNSS spoofing detection and mitigation process.

The statistical analysis of the designed EOOA-HABMLP-based GNSS spoofing detection and mitigation model is elucidated in Table 4. As in (18), the values should be low for the developed technique than the traditional models. In the second dataset, the best measure of the designed EOOA-HABMLP-based GNSS spoofing detection and mitigation model is 7.01%, 7.54%, 8.36%, and 8.56% boosted than the RSO-HABMLP, DMO-HABMLP, CHIO-HABMLP, OOA-HABMLP. Therefore, the performance of the EOOA-HABMLP-based GNSS spoofing recognition and mitigation method is enhanced compared to other traditional models.

The numerical validation of the developed EOOA-HABMLP-based GNSS spoofing detection and mitigation model is listed in Table 5 and Table 6. For the first and the second datasets, the accuracy of the suggested EOOA-HABMLP-based GNSS spoofing detection and mitigation model is 94.29% and 96.27% respectively, both of these values are greater than the conventional model. Likewise, the specificity of the developed EOOA-HABMLP-based GNSS spoofing detection and mitigation method for the second dataset is 86.26%, which is greater than the KNN, SVM, BL, and MLP. Therefore, the performance of the designed EOOA-HABMLP-based GNSS spoofing detection and mitigation model is enhanced compared with other existing methods.

Table 7 shows the comparative study of the implemented EOOA-HABMLP technique with other state-of-the-art models. In traditional techniques, timely detection is not efficient so, it affects the system performance in GNSS spoofing. However, the classical techniques are not efficient in validating larger sizes of data. Considering dataset 1, the existing SDR method shows less accuracy rate as 92.90 which can minimize overall performance in GNSS spoofing. The developed EOOA-HABMLP model shows 94.29 higher accuracy rates than the traditional models. This significant analysis in the developed model helps to manage large-size data and reduce the maximum duration of the detection process. In dataset 2, the performance validation of implemented EOOA-HABMLP is maximized by 7% SDR, 7.5% GLRT, 8.3% CNN, and 8.5% ML in terms of precision analysis. The implemented method achieves a high precision rate compared to traditional methods thus it produces better detection without any interfaces. It effectively increases the performance in the GNNS spoofing detection.

Position, velocity, and time (PVT) analysis of the developed EOOA-HABMLP method is shown in Fig. 12. PVT analysis plays a crucial role in GNSS spoofing detection performance. This analysis is used to provide valuable insights from the raw observable data. Here, the results are validated by east, north, and up (ENU) coordinate. Here, the X-axis is validated based on time. In the Y-axis, velocity, altitude, longitude, and latitude are evaluated in different ranges to show the true latitude and spoofed latitude.

In Fig. 7, shows the convergence analysis of the developed model. This analysis helps to maximize the decision-making performance and faster response in a network system. Also, it can reduce the overall processing time to provide a better solution to minimize cost function analysis. Fig. 8 represents the algorithmic examination of the proposed model. Here, several performance metrics are used to analyze the algorithmic examination process. This process helps to reduce the computational time to detect malicious activities in GNSS. Therefore, it can easily find the hackers before damaging the network to provide better security for data. Fig. 9 compares the performance of the existing and developed method detection techniques using measures like accuracy, FNR, FDR, recall, precision, specificity, NPV, F1-score, and MCC. From Fig. 9 and the second dataset, the developed EOOA-HABMLP-based GNSS spoofing detection and mitigation model accomplished the MCC value of 21.73% enhanced than the KNN, 13.04% enhanced than the SVM, 11.9% enhanced than the BL and 8.69% at tanh activation function. In FNR, the developed model produces low error rates when compared to existing approaches. This low error rate helps to improve the detection performance in the GNSS system and provides better security in the data transmission process. In Table 4, several traditional methods and the implemented model are analyzed using statistical analysis with the help of best, worst, median, mean, and standard deviation. This performance helps to handle large-size data and reduce overfitting issues. Numerical examination among traditional and developed models is shown in Table 5. Here, the precision rate is higher in the developed model to effectively provide security in the network and take less time for the detection process. This may help to improve accurate prediction while the GNSS signals are weak or obstructed. In Table 7, the comparison analysis of the implemented method and state-of-the-art method with the help of several measures is validated among two datasets. Based on this validation, the implemented method facilitates to handle errors in the system.