Title Author Journal Year Summarized Abstract Objectives Methods Results Conclusion Research Gap

Analysis of Tehseen University 2023 1.The paper analyzes IoT The objectives of the study are to The study uses a review of existing The results of the study show that The conclusions of the study are that 1.Lack of standardized
IoT Security Mazhar; of security challenges and explore the use of machine literature and research to explore the use machine learning and deep learning machine learning and deep learning security measures for
Challenges Dhani Bux Johannesbur solutions using AI. learning and deep learning of machine learning and deep learning techniques can be effective in improving techniques are essential for improving N.B. IoT in healthcare.
and Its Talpur; g 2.IoT enhances various techniques to improve IoT techniques in IoT security. IoT security, and that clustering IoT security, and that further research 2.Insufficient
Solutions Tamara Al sectors but faces significant security, to identify the challenges algorithms can be used to resolve IoT is needed to address the challenges exploration of deep
Using Shloul et al. cyber threats. and limitations of IoT security, The research method involved a security issues. and limitations of IoT security. learning methods for
Artificial 3.Traditional security and to discuss the potential literature review of IoT security studies, The study concludes that machine IoT security.
Intelligence methods are ineffective solutions and future work in this using search operators to find relevant The study found that IoT devices are learning and deep learning solutions 3.Need for addressing
against new vulnerabilities. area. information on topics such as IoT, vulnerable to network threats, including are essential for effective IoT security, unique IoT node
4.AI, particularly machine machine learning, deep learning, threats, spoofing and denial of service, and that and that further research is needed to security characteristics.
learning, is essential for IoT cyberattacks, and vulnerabilities. machine learning and deep learning develop effective strategies for 4.Limited research on
security. solutions can be used to detect and continuously improving IoT security. integrating AI with IoT
5.The study examines attack prevent these threats. security solutions
patterns in unstructured data.
6.Future research directions
and challenges in IoT security
are discussed.

Using Hosam F. El- Scientific 2024 1.The paper addresses IoT The main objective of the study is The study uses seven ML algorithms to The proposed ML-based security model The proposed ML-based security 1.Limited datasets may
machine Sofany; Reports system security using to develop a novel ML-based identify the most accurate classifiers for achieves a high accuracy rate of 99.9% model is a viable option for improving not cover all IoT attack
learning Samir Abou machine learning. security model that can improve their AI-based reaction agent’s compared to previous research for the security of IoT systems, types.
algorithms El-Seoud; 2.It presents a novel ML- the security of IoT systems by implementation phase, which can identify improving IoT systems’ security. outperforming existing ML-based 2.Need for optimizing
to enhance Omar H. based security model for IoT. autonomously managing the attack activities and patterns in networks models in terms of accuracy and execution time for real-
IoT system Karam et al. 3.The model autonomously growing number of security connected to the IoT. The proposed ML-based security model execution time. world applications.
security manages increasing security issues associated with the IoT The study employed various machine achieved high accuracy, precision, and 3.Integration with other
issues in IoT. domain. learning algorithms, including Random detection rates, and outperformed The development of this novel ML- security solutions is
4.Seven ML algorithms were Forest, Naive Bayes, Decision Tree, previous solutions in terms of accuracy, based security model is a significant necessary for
utilized for attack detection. Neural Networks, XGBoost, AdaBoost, precision, detection rate, CPE, and time contribution to the literature on ML robustness.
5.The proposed approach and Ensemble RF with backpropagation complexity. security models and IoT security, and 4.Lack of explainability
achieved 99.9% accuracy NN, to detect and mitigate IoT security further work and improvements will in AI-based IDS
and perfect AUC score. threats. The proposed ML-based model is found continue to advance the field. decision-making
6.It outperforms previous The research applied the proposed IDS to have a good accuracy rate of 99.9% process.
models in speed and model to a dataset that included more compared to previous research, and can
accuracy. than 23 types of attacks, and utilized the automatically address rising concerns
NSL-KDD dataset to evaluate the about high security in the IoT domain.
intrusion detection mechanism in a real-
world smart building environment.
Enhancing Ashish Future 2023 1.The paper focuses on IoT The objective of this research is The methods used in this research The results of this research show that The study concludes that the 1.The paper does not
IoT Device Koirala; Internet device security enhancement. to improve the security of IoT include generating normal and attack the machine learning models are able to employment of machine learning address real-time
Security Rabindra 2.It addresses confidentiality devices by investigating the network traffic using Wireshark, identify the normal behaving network applications, especially for classifying attack detection
through Bista; João and security issues in IoT. likelihood of network attacks converting packet-based data into a flow- and attack network with optimal complex and imbalanced datasets, methods.
Network C. Ferreira 3.Network attack data is using ordinary device network based format using the Argus tool, and accuracy. The ensemble approach utilizing suitable data preprocessing 2.Limited exploration of
Attack Data analyzed using machine data and attack network data implementing machine learning provided significant results, and the techniques, is effective in improving the diverse IoT device
Analysis learning algorithms. acquired from similar statistics. algorithms to categorize normal and gradient boost classifier showed high security of IoT devices. types and
Using 4.Botnet attacks were attack traffic. accuracy. environments.
Machine conducted on smart The study concludes that machine 3.Lack of
Learning healthcare IoT devices. The study involves capturing network The results show that the random forest learning models can effectively classify comprehensive
Algorithms 5.Data was analyzed using packets in real-time using packet classifier provided the optimal result of normal and attack network traffic with evaluation of machine
statistical measures for analyzer tools and network flow tracers, 99.9872% to classify the attack and high accuracy, and the generated learning algorithms.
feature extraction. and then preprocessing the data using normal network traffic in binary dataset has the potential to serve as a 4.Insufficient focus on
6.Machine learning techniques such as imblearn, classification valuable resource for further long-term security
categorized normal and normalization, correlation, and entropy investigations in the area of IoT security implications of IoT
attack traffic effectively. selection devices.
7.The BoT-IoT dataset was 5.No discussion on
used for cross-evaluation. user privacy concerns
8.The study emphasizes during data collection
advanced strategies for
attack detection

Security Prafulla E Research 2024 1.The paper discusses IoT The objective of the study is to The study uses the CICIDS 2017 The results show that Hybrid algorithms The study concludes that Hybrid 1.Lack of
Measures at Ajmire , security using machine design a framework for detecting benchmark dataset and implements achieve better accuracy and efficiency algorithms are effective in detecting comprehensive
Network Pragati learning techniques. DDoS attacks in IoT networks machine learning models such as Linear in detecting DDoS attacks compared to DDoS attacks in IoT networks and can evaluation of hybrid
Layer of IoT Vishal 2.IoT devices face unique using machine learning Regression, Decision Tree, Random individual machine learning models. The be used to improve the security of IoT algorithms.
Using Thawani security and privacy techniques and to evaluate the Forest, and Hybrid algorithms using study also presents the performance devices and networks. 2.Limited exploration of
Machine challenges. performance of different machine Python Jupyter notebook. metrics of each model, including real-time
Learning 3.Machine learning learning models. accuracy, precision, recall, F1 score, The study concludes that machine implementation
Techniques addresses various security The study uses machine learning detection rate, and false alarm rate. learning techniques, specifically hybrid challenges.
issues in IoT. techniques, including decision trees, algorithms, are effective in detecting 3.Insufficient analysis
4.The study uses CICIDS random forest, and hybrid algorithms, to The study finds that the hybrid algorithm, DDoS attacks on IoT devices, of varying IoT device
2017 dataset for analysis. detect DDoS attacks on IoT devices. The specifically Hybrid Algorithm 2 (DT+RF), specifically Raspberry Pi. The study capabilities.
5.Three ML models are study also uses a dataset, specifically is the strongest ML classifier with an also emphasizes the importance of 4.Need for broader
evaluated: Linear CICIDS 2017, to evaluate the accuracy of 99.98%, precision of using lightweight machine learning dataset diversity for
Regression, Random Forest, performance of the algorithms. 99.97%, and a good false alarm rate of approaches to safeguard IoT training.
Decision Tree. 0.0002. The study also finds that the deployments. 5.Limited focus on long-
6.Decision trees provide a algorithm is suitable for real-time term security
balance of speed and deployment on Raspberry Pi. effectiveness.
accuracy.
7.Hybrid algorithms aim for
improved accuracy in results
IoT Security Chaw Su Journal of 2024 1.The paper focuses on IoT The objective is to develop a The proposed system consists of four The experiment results indicate that the The proposed system contributes to 1.The paper does not
Using Htwe; Zin Computing security using machine machine learning-based phases: feature extraction, feature detection results with selected features the development of a real-time attack address unknown
Machine Thu Thu Theories learning methods. approach for attack detection in selection, feature encoding, and are improved, especially in terms of the detection capability in an IoT attack detection
Learning Myint; Yee and 2.IoT devices face security IoT network traffic data, using machine-learning-based attack detection. processing time and detection rate in environment, with faster processing methods.
Methods Mon Thant Applications weaknesses and malware feature extraction, selection, and The feature extraction mechanism uses Naïve Bayes. The CART algorithm and better detection accuracy by 2.Limited focus on
with attacks. classification algorithms to Algorithm 1, and the feature selection outperformed the result of Naïve Bayes. reducing the number of features. feature selection impact
Features 3.Proposed framework improve detection accuracy and process applies the Pearson correlation The proposed system is effective in on performance
Correlation improves attack detection efficiency. method. The CART algorithm is used for The results show that the proposed detecting attacks in IoT network traffic metrics.
with feature correlation. attack detection. system achieves high accuracy with data, with CART achieving perfect 3.Lack of exploration
4.Achieves nearly 100% both all and selected features using accuracy with selected features. on alternative machine
detection accuracy with The methods used include feature CART and Naïve Bayes, with CART Feature selection improves learning algorithms.
selected features. extraction using CICFlowMeter and achieving perfect accuracy with selected computational efficiency and detection 4.Insufficient analysis
5.CART algorithm is Zeek, feature selection using Pearson features. The results also show that accuracy with CART, but may lead to a of real-time attack
preferred for processing time correlation, and classification using feature selection improves slight decrease in some performance detection challenges
and accuracy CART and Naïve Bayes. computational efficiency with CART. metrics with Naïve Bayes.
The study concludes that feature
The study uses the Pearson correlation The results of the study show that CART selection is advantageous for
feature selection method to reduce can detect attacks with 100% accuracy, improving computational efficiency, and
irrelevant features and applies the CART even when reducing many features. CART is a better choice for
algorithm for building the attack detection Naive Bayes performance also computational efficiency based on the
model. The performance of CART is achieves up to 99.93% with the selected provided data.
compared with Naive Bayes. features.