Ensembled Machine Learning Algorithm Based Intrusion Detection with Blockchain Technology

Jothy Chandrikha Raveendran¹, Judith John Edwin¹ and Anju Ajith Jaya²
1. Computer science and Engineering Noorul Islam Centre for Higher Education, Kumaracoil, Thuckalay, Kanyakumari, Tamil Nadu, India
2. Computer science and Engineering Vimal Jyothi Engineering College, Chemperi, Kannur, Kerala, India
Correspondence to: Jothy Chandrikha Raveendran , crjothiesh@gmail.com

DOI: https://doi.org/10.70389/PJS.100137

Additional information

• Ethical approval: N/a
• Consent: N/a
• Funding: No industry funding
• Conflicts of interest: N/a
• Author contribution: Jothy Chandrikha Raveendran, Judith John Edwin and Anju Ajith Jaya – Conceptualization, Writing – original draft, review and editing
• Guarantor: Jothy Chandrikha Raveendran
• Provenance and peer-review:
  Unsolicited and externally peer-reviewed
• Data availability statement: N/a

Keywords: Blockchain-integrated intrusion detection, Ensemble machine learning, Random forest–gradient boost stacking, Tamper-proof security logging, Decentralized ledger scalability.

Peer Review
Received: 16 August 2025
Last revised: 25 September 2025
Accepted: 29 September 2025
Version accepted: 3
Published: 17 October 2025

Plain Language Summary Infographic

Introduction

Recent report on digital population says among the world’s total population 66.2% are using internet. Internet users have grown by 1.8% over the past year, with 97 million new users coming online for the first time during 2023.¹ As the usage of internet among the population increases there is a tremendous rise in cyber-attacks. Due to the intelligent act of intruders, we are in need of intelligent systems to detect those attacks once they arise. Most of the intrusion detection systems uses labelled set of data for predicting which means it produces an alert based on the trained set of data which in turn will fail to respond to the real time attacks and so there is a need for automated IDS which can be able to alert or prevent the upcoming intruders so as to safeguard the network security.² Machine learning based intrusion detection system will able to address this kind of real time attacks and by integrating the blockchain technology along with this intrusion detection model will double the chance of securing the intrusion attacks by its immutable feature.³

Background of Intrusion Detection System

An intrusion detection system (IDS) analyses network traffic to identify such malicious traffic. Intrusion detection systems are broadly classified as misuse based and anomaly detection whereas the former method commonly called signature-based IDS works based on the known patterns and the latter method identifies any deviations in the normal traffic.^2,4 In some cases, the anomaly detection gives wrong decisions and so it should be properly trained in such a manner to detect the exact intruders. Anomaly-based, signature-based and hybrid-based are the three existing common ways of intrusion detections. The anomaly-based detection of intrusion is also called behaviour-based detection as it detects the intrusion based on the behaviour of the user, network and the host system which the generates an alert to the user regarding the threat. The signature-based intrusion detection system also referred to as knowledge-based IDS which rely on the database with previous attack history. The hybrid-based intrusion detection works by integrating the performance of both the anomaly based and signature-based IDS so that to enhance the performance.^4,5

Machine Learning in Intrusion Detection

Machine Learning (ML) being a subset of Artificial Intelligence (AI) is securing popular traits among recent times it will act as the catalyst to enhance the performance of the intrusion detection techniques so that training IDS using machine learning algorithm will be capable of pinpointing the exact intruders from other users efficiently.^2,3 Since, machine learning has three broad categories such as supervised, unsupervised and reinforcement algorithm, choosing the right algorithm will help to train the model more efficiently.

Blockchain in Intrusion Detection

Blockchain being a decentralized immutable distributed ledger, able to record the network activities securely, each and every action of the IDS will be stored in this distributed ledger and can be accessed by any authorized users.^6,7 Conventional intrusion detection systems have well documented shortcomings, such as high false positive rates, poor flexibility to accommodate changing threats, and scalability issues in distributed systems. Although ensemble ML methods can enhance detection rates by aggregating multiple models, ML-based IDS alone still have serious vulnerabilities. System access attackers can manipulate or erase intrusion logs, undermining forensic analysis. In addition to detection, conventional systems tend not to have adequate measures of device authentication, exposing IoT and cloud networks to spoofing.

The integration of BC technology bridges these shortcomings with immutable record storage of intrusion logs, protection against tampering and support for credible post-attack investigation. In addition to immutability, BC provides decentralized consensus, where multiple nodes can authenticate ML-generated alerts, thus minimize false positives and increase trust in detection results.⁸ Implementing smart contracts on the blockchain to automate responses to detected anomalies, such as isolating systems or alerting security teams will make a considerable enhancement in the overall performance of the Intrusion detection model. Blockchain also enables real-time verification of anomaly alerts, ensuring that only verified alerts trigger automated responses.

Section II gives a detailed description on the existing works done based on the proposed method. Section III describes the methodology of the proposed model along with the architectural diagram and describes the features behind the methodology. Section IV gives the summary of the various evaluation metrics involved in the diagnosis of the performance of the proposed model in predicting the real time data. Section V gives an inference of the results obtained by the proposed model along with the visualization of the curves. how blockchain can be used along with the machine learning intrusion detection system also this section enumerates the existing works based on this approach. Finally, the paper concludes by stating the key findings and also it gives an inference about the future directions.

Discussions on Related Work

Aliyu Ahmed Abubakar et al.⁹ proposed a new methodology with improved accuracy of about 92.6% in which they integrated blockchain along artificial intelligence with the intrusion detection system. The model was tested with MIT Lincoln Labs and DARPA 99 datasets. The model used here uses the blockchain network and also it exhibits good scalability when compared to other models. The model exhibits lower risk of false positives. Swapna Siddamsetti and Muktevi Srivenkatesh¹⁰ suggested a frame work by combining intrusion detection and blockchain towards the purpose of increasing the data privacy named Machine learning blockchain framework (MBF). The research proposes both security over IoT networks and also blockchain privacy. The models uses XGboost algorithm and the dataset used was N-BaloT. The model when tested is able to give 98% accuracy with the above said dataset. Eman Ashraf et al.¹¹ proposed an edge – cloud intrusion detection mechanism by integrating blockchain and FIDChain (Federated Intrusion Detection Chain) for health care systems along with artificial neural networks and eXtreme Gradient Boost algorithm.⁹ The model provides 99.99% accuracy with BoT – IoT datasets. The model also provides balanced results of intrusion detection.

Khonde, S.R., Ulagamuthalvi, V5 proposed an innovative approach to share signatures over distributed environment of IDS. This approach utilizes two detection methods one is the signature-based and the other is the anomaly based in order to improve the security of network.¹² The approach splits the approach into three different phases in the first phase ANN algorithm with CIC-IDS dataset is used, whereas in the second phase anomaly-based detection here machine learning and deep learning algorithms were used to do the behaviour analysis and finally in the third phase the blockchain framework for signature extraction and distribution was done and the results shows there is a significant improvement in accuracy with 94.86% while using KDD 99 dataset. Table 1 discusses the methodologies used and also the limitations in the existing models.

Table 1: Comparative analysis of bc integrated ml approach for ids.
Authors	ML Method	Blockchain Role	Accuracy	Limitations
Aliyu et al.6	Unspecified AI	Log storage, scalability	92.6%	Generic ML; no attack-specific tuning
Swapna et al.10	XGBoost	Data privacy	98%	No consensus for ML validation
Ashraf et al.11	XGBoost and ANN	Federated learning	99.99%	Assumes trusted edge nodes
Khonde et al.15	ANN and DL	Signature distribution	94.86%	Static signatures; poor zero-day detection

Proposed Blockchain Based Ml for Ids

Blockchain being a decentralized ledger places its root in various fields because of its immutable nature its growth towards various sectors has increased. AI being the current trending technology is incorporated with many existing technologies so that to enhance the performance of those by analysis huge volumes of data and also able to process real time data.¹³ Hence both these technologies come into the frame of solving issues regarding the intrusions in a network. Table 2 shows some of the considerations while designing an IDS system with Machine learning and blockchain. Our study utilizes the Ethereum Transaction Dataset with features, including gas usage, transaction value, and temporal patterns. The dataset covers diverse attack types such as phishing, re-entrancy attacks and the missing values were imputed using mean substitution by SimpleImputer, followed by StandardScaler standardization to ensure consistent feature scales across models.

Table 2: Important features for ml based ids with blockchain.
Sl. No	Feature	Description	Considerations
1.	Data Privacy	Confidential and secure	Federated learning
2.	Immutability	Preventing tampering	Blockchain with cryptographic hashing
3.	Scalability	Huge volumes of data	Efficient data processing and storage
4.	Latency	Time delay to intrusions	Consensus algorithms
5.	Inter-operability	Supports different platforms	Adaptable frameworks
6.	Cost	Financial implications	Cost-effective model

The proposed model comprises the function of both the Random Forest and Gradient Boost models to predict whether the encountering data is a malicious data or not. Random Forest is a versatile machine learning algorithm that builds multiple decision trees during training and merges their outputs for improved accuracy and robustness. To ensure variation among the trees, a random subset of the data and features is used to generate each tree.³ The individual tree forecasts are combined to get the final prediction, typically by average for regression or majority vote for classification. This ensemble method effectively manages missing values and big datasets while lowering overfitting and improving generalization. With its excellent performance and interpretability, Random Forest is frequently used for tasks including feature selection, regression, and classification.¹⁴

Gradient Boosting is a powerful machine learning algorithm used for both classification and regression tasks. It sequentially constructs an ensemble of weak learners, usually decision trees. With an emphasis on the most difficult-to-predict cases, each new tree is taught to fix the mistakes made by the ones that came before it. All of the individual weak learners’ outputs are combined to create the final model, which produces a reliable prediction.^12,15 By iteratively adding trees that reduce the residual errors, gradient boosting improves a loss function. Although it may achieve high accuracy and is very successful for complicated datasets, careful hyperparameter adjustment may be necessary to prevent overfitting and guarantee optimal performance.

Based on the acquired knowledge from the recent studies the integration of machine learning algorithms with blockchain technology will considerably enhance the performance of Intrusion Detection Systems (IDS).¹⁶ Our proposed system integrates the blockchain technology along with the IDS model. Initially the data gathered from the real time network is reposited in a local database which can be tampered, so that to make the acquired data safe and secure it should be placed in a tamper-proof immutable ledger for which the blockchain technology is used here. Blockchain stores the data in, multiple locations as it is a decentralised distributed ledger which utilises cryptographic and consensus techniques to all the data stored in it. Figure 1 shows the flow diagram for the proposed model which is trained to form an efficient Intrusion Detection System.

The network data is the base for the Machine learning based intrusion detection system model with the integration of Blockchain technology to securely storing the collected data as well as the predicted events so that no external inference will affect the data which is kept in data blocks created by the blockchain.¹⁷ Initially the process begins with continuous monitoring and collection of network traffic. The real time traffic data is fed to preprocessing where the real time data needs to be cleaned so as to make the huge volumes of raw data occupies less space in the blockchain ledger and also it has more unwanted data so that to format the raw data preprocessing is done as the second step. The pre-processed data needs to be stored which should not be tampered so we need to store the data as a decentralized manner, hence blockchain with distributed ledger technology stores the pre-processed data.

Once pre-processed data is stored one can retrieve it at any time for further processing. After this step feature engineering is involved to extract relevant features by eliminating feature which are less important in predicting the anomalies. The extracted features form the input to the next block which is the training model. The model training and prediction goes by two phases. In the first phase the featured data will be used to train the Random Forest and Gradient Boost separately and the prediction outputs are obtained. In the second phase a meta-model which is framed by the Random Forest model is trained and then it is evaluated with the test data to obtain the prediction. The meta model classifies the malicious threats more effectively in the second phase of prediction.

Ensembled machine learning model forms the integral part of the entire prediction system. Two algorithms Random Forest and Gradient Descent Algorithms are combined together with the help of stacked generalization or stacking.¹⁴ Our stacking ensemble uses two phases one is the Base Models which comprises RF and XGBoost generate probabilistic predictions via 5-fold cross-validation to prevent data leakage. In the Meta-Model the Random Forest combines base predictions, learning optimal weights through supervised training. Figure 1 illustrates the workflow. For Gradient Boosting (XGBoost), we implemented Bayesian optimization to tune the learning rate, tree depth, subsampling rate, and number of estimators with early stopping. Using the stacked generalization, both the models Random Forest and Gradient Boost models are separately trained and after training them separately the results of prediction from the two models are fed to a meta- model which is another Random Forest algorithm in our proposed architecture design and the meta model is trained to get the exact prediction of the overall model.^18,19

The ML-based IDS model, trained using historical data, identifies patterns of normal and malicious activities, detecting intrusions effectively. When an intrusion is detected, the system generates alerts and logs detailing the suspicious activity. These detection results are securely stored and verified on the blockchain, ensuring data integrity, transparency, and immutability. Each detection event is recorded as a transaction on the blockchain, providing a tamper-proof audit trail. Use of permissioned consortium network for the blockchain layer helps to securely log any anomalies spotted by the intrusion detection model. Practical Byzantine Fault Tolerance (PBFT) consensus protocol keep validation quick and reliable among trusted nodes. Smart contracts are set up to automatically record anomalies along with details like the type of attack, confidence score, and timestamp, while also sending real-time alerts to system administrators. By only logging the anomalous transactions, we significantly cut down on storage needs compared to logging all network traffic. This way, the blockchain strikes a nice balance between handling a high number of transactions per second and maintaining security.

This design ensures that we have immutability, transparency, and solid proof of any detected attacks, which in turn boosts both the reliability of detection and the ability to conduct forensic investigations. The final step involves the admin dashboard and response system, which allows administrators to monitor alerts, view logs, and respond to detected intrusions in real-time.²⁰ This integrated approach leverages the scalability and accuracy of machine learning while utilizing blockchain’s decentralized and immutable nature to enhance security.^10,21 The combination ensures that IDS not only detects threats efficiently but also maintains a secure and transparent record of all detected intrusions. This dual-layered system addresses traditional IDS challenges, offering a comprehensive and future-proof solution to network security. Unlike the current studies on IDS with blockchain, like DeepFed-IDS and GAN-based IDS, our work introduce a stacked ensemble model that combines Random Forest and XGBoost with a Gradient Boost meta-learner.

The model was tested using two benchmark datasets, NSL-KDD and CIC-IDS-2017, and finally we integrate it into a blockchain framework where we assess latency, throughput, and storage costs. The proposed system encounters several risks, including model poisoning, where malicious actors might try to compromise the IDS training data, and smart-contract exploitation, where attackers could take advantage of the anomaly logging logic. To counter these threats, we implement strong data validation, ensemble-based detection methods, and smart contracts that have been formally verified. The blockchain layer was set up on a private Ethereum network which uses Proof-of-Authority/ (PoA), which ensures low latency. The blockchain layer creates a trade-off between security and efficiency. PoA consensus and on-chain storage provide tamper-proof validation, but they also add some latency and storage costs. This shows the need for lighter consensus methods and mixed storage solutions to balance scalability and security in real-world use. Moreover, the use of a permissioned PBFT blockchain with authenticated nodes helps prevent tampering and provides resilience against Sybil or collusion attacks. The pseudocode for the proposed methodology with the input of publicly available datasets to produce final prediction alert with blockchain integration is listed as

1. Data processing
2. Base learner trained with the Random Forest
3. Meta learner which is the stacking design trained with Gradient Boost
4. Blockchain integration for tamper proof record and alert sharing
5. Finally model evaluation to validate the model

Key Metrics in Intrusion Detection

While designing any system model it is essential to consider some factors which directly show the performance of the model designed. Intruders can attack a system in many ways and it is the responsibility of the network organizer to come up with preventive measures. The measuring metrics are True Positive (TP), True Negative (TN), False Positive (FP) and False Negative (FN). Some of the common factors involved in the performance metrics are listed as follows:

1. Accuracy: It measures the overall correctness of the model. It is the ratio of correctly predicted instances (both true positives and true negatives) to the total instances.

2. Attack Detection Rate (ADR): Ability of the model to identify actual attacks was measured using ADR. It is the ratio of true positives to the sum of true positives and false positives.

3. Precision: It measures the accuracy of the positive predictions. It is the ratio of true positives to the sum of true positives and false positives.

where,

• TP – correctly identified attack instances
• TN – correctly identified normal instances
• FP – normal instances incorrectly identified as attacks
• FN – attack instances incorrectly identified as normal

Comparative Evaluation of IDS Models

We compared the proposed model with the existing models and found that our proposed model is well suited for the Intrusion detection providing a better accuracy when compared to the existing ones. Our experimental evaluation shows that the Random Forest baseline hit an impressive 98.67% accuracy, while XGBoost alone soared to 99.40%. The stacked meta-model we proposed, which combines Random Forest and XGBoost with a Gradient Boost meta-learner, achieved a solid 99.16% accuracy, along with balanced precision, recall, and an F1-score of 99%. When we compare this to LightGBM (97.3%), DeepFed-IDS (98.5%), and CNN-LSTM (98.7%), our model consistently outperformed them by 0.3–1.7%. These results really highlight the strength of our ensemble approach, especially when it comes to tackling minority attack classes. Also, the integration of blockchain adds a layer of transparency and ensures a tamper-proof IDS deployment. Table 3 makes a comparison of the proposed ensembled model performance with the LightGBM, DeepFed-IDS, CNN-LSTM IDS to understand how the proposed model works.

Table 3: Comparison of proposed model performance with existing models.
Model	Accuracy	Precision	Recall	F1-Score	Model Response
LightGBM IDS	97.3	96.8	96.5	96.6	Fast, efficient baseline for tabular IDS tasks.
DeepFed-IDS	98.5	98.3	98.0	98.1	Federated deep learning for privacy-preserving IDS.
CNN-LSTM IDS	98.7	98.5	98.2	98,3	Captures temporal and spatial features.
Stacked Meta-Model	99.16	99.0	99.0	99.0	Combines RF and XGBoost with meta-learner; blockchain-integrated IDS.

Statistical Significance Analysis

Using McNemar’s test and Paired t-test statistical significance analysis has been conducted to justify how the new stacked meta-model performs well against the baseline IDS solutions. Table 4 shows the statistical significance analysis of the proposed model with the existing models and hence the stacked model demonstrates superior robustness across minority attack classes. Table 4 illustrates how significant the proposed model when compared to the existing intrusion detection models

Table 4: Statistical significance analysis.
Comparison	p-Value	Significance (=0.05)	Findings
Proposed with LightGBM	P < 0.001	Significant	Proposed model outperforms
Proposed with DeepFed-IDS	p = 0.002	Significant	Remarkable improvement over DeepFed-IDS
Proposed with CNN-LSTM	P = 0.04	Significant	Achieved valid gain
Proposed with XGBoost	P = 0.18	Non Significant	Comparatively equal performance

Result and Discussions

For the NSL-KDD dataset, we took care of missing values, removed any constant columns, and applied a quantile transformation. We divided the data into three parts: 70% for training, 15% for validation, and 15% for testing. We put Random Forest, Gradient Boosting, and a stacked ensemble with Gradient Boosting as the meta-learner to the test. To fine-tune the hyperparameters, we used Bayesian optimization, and all our experiments were conducted in a Python 3.10 environment (using scikit-learn and XGBoost) on a machine equipped with an Intel i7 CPU and 16 GB of RAM. The proposed ensemble model, combining Random Forest and Gradient Boosting with a meta model, achieved an impressive accuracy of 99.2%. While our model achieves 99.2% accuracy on the test set. XGBoost regularization and feature selection removing 30 low-importance features. The meta model in the proposed work uses the Random Forest Machine Learning model which uses the ensembled predicted output, raw network data and then it analyses with the malicious behaviour data and the results obtained will be more accurate than what we obtain initially through the individual models at the first phase of prediction. Figure 2, shows the confusion matrix obtained by the model. The confusion matrix reveals most errors occur as false positives in normal traffic (5%), while critical false negatives are limited to 0.8% of attacks.

Figure 3, describes the Receiver Operating Characteristic curve which diagnosis the ability of the model in classifying the anomalies, it provides the curve plotted between the TPR against the FPR to show the model performance. The multi-class ROC curve obtained as the result of XGBoost model shows a significant enhancement in the performance of the model when considering the different classes of the intrusion detection system. Well-performing classes include Class 0, Class 1, and Class 2. An AUC of 1.0 indicates very excellent discrimination also Class 7 gives an AUC of 0.99, and Class 5 scores 0.90. Each class represents a particular type of network traffic or intrusion behaviour where Class 0 represent normal traffic, while others may represent attack types. The results show that the model is good for most classes but needs improvement in detecting some attack categories. The proposed system gives a throughput of over 85 transactions per second while the storage overhead was approximately 12 MB per 10,000 alerts. Our ensembled model with optimized blockchain integration achieves scalable intrusion detection through parallelized based model inference and smart meta-model validation.

Conclusion and Future Scope

This paper highlights the potential of current intrusion detection models that integrate both machine learning algorithms and blockchain technology. It examines the performance of these models by comparing them with various alternatives and offers a comprehensive overview of existing IDS research. our research brings three significant contributions to the table. First, we introduce a stacked ensemble model that merges Random Forest and XGBoost, enhanced by a Gradient Boost meta-learner. Second, we seamlessly integrate this model into a blockchain framework and thoroughly assess its latency, throughput, and storage overhead. The study systematically compares different systems proposed by researchers to identify an efficient IDS model capable of preventing unauthorized access to secure networks. The proposed model is efficient in achieving the efficiency along with scalability.

Therefore, it is recommended to develop a new model that addresses scalability issues, ensuring that the model is more adaptable and interoperable. In order to increase accuracy even more, the future scope will investigate other ensemble strategies as bagging and boosting using various base models. Performance on increasingly complicated datasets can be improved by incorporating deep learning models into the ensemble, such as recurrent neural networks (RNNs) and convolutional neural networks (CNNs). Furthermore, addressing unbalanced datasets and enhancing model generalization can be facilitated by implementing sophisticated feature engineering and data augmentation approaches. To create effective and scalable intrusion detection systems for dynamic situations, real-time data streams and edge computing might also be investigated. A promising direction is the integration of federated learning with blockchain to facilitate secure, privacy-preserving sharing of threat intelligence models across organizational networks while maintaining data confidentiality and auditability.

References

1. DataReportal. Digital 2024 Deep Dive: The State of Internet Adoption. [Internet]. 2024. Available from: https://datareportal.com/reports/digital-2024-deep-dive-the-state-of-internet-adoption
2. Prachi HM, Sharma P. Intrusion detection using machine learning and feature selection. Int J Comput Netw Inf Secur. 2019;11(4):43–52. https://doi.org/10.5815/ijcnis.2019.04.05
3. Upadhyay D, Manero J, Zaman M, Sampalli S. Gradient boosting feature selection with machine learning classifiers for intrusion detection on power grids. IEEE Trans Netw Serv Manag. 2020;18(1):1104–16. https://doi.org/10.1109/TNSM.2020.3037330
4. Morfino V, Rampone S. Towards near-real-time intrusion detection for IoT devices using supervised learning and Apache Spark. Electronics. 2020;9(3):444. https://doi.org/10.3390/electronics9030444
5. Khonde SR, Ulagamuthalvi V. Hybrid intrusion detection system using blockchain framework. EURASIP J Wireless Commun Netw. 2022;2022(1):58. https://doi.org/10.1186/s13638-022-02122-4
6. Abubakar AA, Liu J, Gilliard E. An efficient blockchain-based approach to improve the accuracy of intrusion detection systems. Electron Lett. 2023;59(18):e12888. https://doi.org/10.1049/ell2.12888
7. Kumar R, Kumar P, Tripathi R, Gupta GP, Garg S, Hassan MM. A distributed intrusion detection system to detect DDoS attacks in blockchain-enabled IoT network. J Parallel Distrib Comput. 2022;164:55–68. https://doi.org/10.1016/j.jpdc.2022.05.002
8. Mishra S. Blockchain and machine learning-based hybrid IDS to protect smart networks and preserve privacy. Electronics. 2023;12(16):3524. https://doi.org/10.3390/electronics12163524
9. Saveetha D, Maragatham G, Ponnusamy V, Zdravković N. An integrated federated machine learning and blockchain framework with optimal miner selection for reliable DDoS attack detection. IEEE Access. 2024;12:127903–15. https://doi.org/10.1109/ACCESS.2024.3420059
10. Siddamsetti S, Srivenkatesh M. Implementation of Blockchain with machine learning intrusion detection system for defending IoT botnet and cloud networks. Ingén Syst Inform. 2022;27(6). https://doi.org/10.18280/isi.270613
11. Ashraf E, Areed NF, Salem H, Abdelhay EH, Farouk A. FIDChain: Federated intrusion detection system for blockchain-enabled IoT healthcare applications. Healthcare. 2022;10(6):1110. https://doi.org/10.3390/healthcare10061110
12. Douiba M, Benkirane S, Guezzaz A, Azrour M. An improved anomaly detection model for IoT security using decision tree and gradient boosting. J Supercomput. 2023;79(3):3392–411. https://doi.org/10.1007/s11227-022-05046-z
13. Markovic T, Leon M, Buffoni D, Punnekkat S. Random forest based on federated learning for intrusion detection. In: IFIP International Conference on Artificial Intelligence Applications and Innovations. Cham: Springer; 2022. p. 132–44. https://doi.org/10.1007/978-3-031-08333-4_11
14. Dietterich TG. Ensemble methods in machine learning. In: International Workshop on Multiple Classifier Systems. Berlin: Springer; 2000. p. 1–15. https://doi.org/10.1007/3-540-45014-9_1
15. Firmansyah R, Utami E, Pramono E. Evaluation of naive Bayes, random forest and stochastic gradient boosting algorithm on DDoS attack detection. In: Proceeding International Conference on Information Science and Technology Innovation. 2022;1(1):92–7. https://doi.org/10.31764/istial.innovation.v1i1.10213
16. Saranya T, Sridevi S, Deisy C, Chung TD, Khan MA. Performance analysis of machine learning algorithms in intrusion detection system: a review. Procedia Comput Sci. 2020;171:1251–60. https://doi.org/10.1016/j.procs.2020.04.134
17. Alsharif NA, Mishra S, Alshehri M. IDS in IoT using machine learning and blockchain. Eng Technol Appl Sci Res. 2023;13(4):11197-203. https://doi.org/10.48084/etasr.6076
18. Zhang J, Zulkernine M, Haque A. Random-forests-based network intrusion detection systems. IEEE Trans Syst Man Cybern C Appl Rev. 2008;38(5):649–59. https://doi.org/10.1109/TSMCC.2008.923876
19. Hasan MA, Nasser M, Pal B, Ahmad S. Support vector machine and random forest modeling for intrusion detection system (IDS). J Intell Learn Syst Appl. 2014;6(1):45–52. https://doi.org/10.4236/jilsa.2014.61005
20. Alasmari S. Optimized blockchain-based deep learning model for cloud intrusion detection. Int J Adv Comput Sci Appl. 2024;15(9). https://doi.org/10.14569/IJACSA.2024.0150924
21. Sunanda N, Shailaja K, Kandukuri P, Rao VS, Godla SR. Enhancing IoT network security: ML and blockchain for intrusion detection. Int J Adv Comput Sci Appl. 2024;15(4). https://doi.org/10.14569/IJACSA.2024.0150427

Cite this article as:
 Jothy CR, Judith JE and Anju AJ. Ensembled Machine Learning Algorithm Based Intrusion Detection with Blockchain Technology. Premier Journal of Science 2025;15:100137

Export to EndNote Export to Mendeley