[논문]Genetic Algorithm based hyperparameter tuned CNN for identifying IoT intrusions

Alexander. R; Pradeep Mohan Kumar. K

doi:10.3837/tiis.2024.03.013

[국내논문] Genetic Algorithm based hyperparameter tuned CNN for identifying IoT intrusions

KSII Transactions on internet and information systems : TIIS, v.18 no.3, 2024년, pp.755 - 778

Alexander. R (Department of Computing Technologies, SRM Institute of Science and Technology) , Pradeep Mohan Kumar. K (Department of Computing Technologies, SRM Institute of Science and Technology)

Abstract ▼ AI-Helper

In recent years, the number of devices being connected to the internet has grown enormously, as has the intrusive behavior in the network. Thus, it is important for intrusion detection systems to report all intrusive behavior. Using deep learning and machine learning algorithms, intrusion detection systems are able to perform well in identifying attacks. However, the concern with these deep learning algorithms is their inability to identify a suitable network based on traffic volume, which requires manual changing of hyperparameters, which consumes a lot of time and effort. So, to address this, this paper offers a solution using the extended compact genetic algorithm for the automatic tuning of the hyperparameters. The novelty in this work comes in the form of modeling the problem of identifying attacks as a multi-objective optimization problem and the usage of linkage learning for solving the optimization problem. The solution is obtained using the feature map-based Convolutional Neural Network that gets encoded into genes, and using the extended compact genetic algorithm the model is optimized for the detection accuracy and latency. The CIC-IDS-2017 and 2018 datasets are used to verify the hypothesis, and the most recent analysis yielded a substantial F1 score of 99.23%. Response time, CPU, and memory consumption evaluations are done to demonstrate the suitability of this model in a fog environment.

주제어

표/그림 (25)

그림 Fig. 1. General Architecture of CNN.
그림 Fig. 2. Structure of the Genes in the Chromosome
표 Table 1. Gene Encoded values [36]
그림 Fig. 3. System Model of eCGA based CNN
그림 Fig. 4. Incorporation of the proposed hyperparameter tuned model into the Fog Network
그림 Fig. 5. Class Distribution plot for the CIC-IDS 2017 Dataset showing the number of instances for each class.
그림 Fig. 6. Distribution percentage of the major class labels Benign and Attack
그림 Fig. 7. Number of attacks in each of the attack class
그림 Fig. 8. Group of Attacks that are holding the highest number of attack instances
그림 Fig. 9. Group of Attacks that are holding the smallest number of attack instances
그림 Fig. 10. Class Distribution plot for the CIC-IDS 2018 Dataset
표 Table 2. Multi Class Label count values and the percentage of distribution
그림 Fig. 11. Feature Map extracted features
표 Table 3. Strong Correlated features requiring feature map reduction
그림 Fig. 12. Distribution of Individuals in the initial and final generation CIC-IDS-2017 and 2018
그림 Fig. 13. Distribution of Individuals in the initial and final generation CIC-IDS-2018
그림 Fig. 14. Distribution plot for the error generated without genetic algorithm and with genetic algorithm
그림 Fig. 15. Confusion Matrix of the CIC-IDS-2017 dataset
그림 Fig. 16. Confusion Matrix of the CIC-IDS-2018 dataset
표 Table 4. Quantitative values for extreme cases
표 Table 5. Quantitative values for best performed case in CIC IDS2017
표 Table 6. Quantitative values for best performed case in CIC IDS2018
표 Table 7. State-of-the-art comparative analysis with detection accuracy for CIC-IDS-2017
그림 Fig. 17. Latency analysis for the Fog environment in terms of the QoS parameters
그림 Fig. 18. Latency analysis for the Cloud environment in terms of the QoS parameters

AI 본문요약
AI-Helper

제안 방법

The probabilistic model genetic algorithm's overall chromosome structure thus consists of 13 genes and the concept of evolving the CNN model to the chromosome is borrowed from this work [36]. No significant changes have been made because the goal of our research was to adapt the evolutionary algorithm to deal with large-scale, multidimensional data in the IoT environment by using the Bayesian approach. The gene structure thus resembles the one described in [36], with the addition of the feature map CNN being the only variation made at this level.
The concept is examined using the CIC-IDS 2017 [21] and CIC-DDoS 2019 datasets, and the proposed methodology is compared to other cutting-edge techniques for higher accuracy and shorter latency.
The method proposed in this study better automates the hyperparameter selection by utilizing the CNN feature map and an enhanced genetic compact algorithm. For executing the categorization of IoT-IDS network traffic, the genetic algorithm is used to achieve the goal of obtaining a superior neural network with automatic adjustments.
As illustrated in the figure, traffic generated by various IoT devices and received in the fog network is handled by multiple subcomponents to determine the traffic's location. The proposed methodology is specifically deployed inside the detection module, where the regular IDS runs and takes care of the classification of the traffic.
This study expands on DeepNEAT's concept [19], in which the neural arrangement is transformed into a chromosome of the Extended Compact Genetic Algorithm(eCGA), in order to eliminate the issue of manually setting up the hyperparameter values and to provide a superior latency IDS
Though there is a wide variety of intrusion datasets are available for the public, to carry out this work specific two variations of the IoT-IDS dataset CIC-IDS2017 and CIC-IDS2018 dataset. The Canadian Institute for Cybersecurity dataset 2017 and 2018 is chosen because of its closeness towards the real IoT traffic data.

이론/모형

An algorithmic approach for performing hyperparameter tuning is proposed in [50], where the arithmetic optimization algorithm is enhanced to improve the candidate solution's performance. The AutoML technique is used along with the ensembled strategy for the ideal selection of hyperparameters for the supervised setup using the voting mechanism proposed in [51]. Though these approaches propose a strategy for hyperparameter tuning, their major limitation is their inability to handle varied forms and volumes of data, which makes them require human assistance at some stage.
The best characteristics are chosen for the classification task using a method called multi-objective particle swarm optimization with a Levy flight randomization component (MOPSO-Lévy), which is suggested in [35]
So to address the concerns of having low latency with a higher degree of parallelism, a novel approach using the Bayesian genetic technique is proposed to address the problem of manual ideal hyperparameter setting. The idea comes from the probabilistic model-building genetic algorithm (PMBGA) that operates based on the probability distribution recorded by the Bayesian network. This study expands on DeepNEAT's concept [19], in which the neural arrangement is transformed into a chromosome of the Extended Compact Genetic Algorithm(eCGA), in order to eliminate the issue of manually setting up the hyperparameter values and to provide a superior latency IDS.

성능/효과

The convolutional neural network (CNN) is made up of three basic components, as illustrated in Fig. 1, an input layer, a hidden layer that contains the convolution and pooling layers, and the classification layer, which is often a fully connected layer.
Fig. 1, shows the general architecture of CNN.
From Fig. 14, it is clear that the error rate is significantly lower at the final generation of parameters. The grey line indicating the final generation parameters is lower than that of the blue line that is showing the initial generation parameters and the similar results are observed for the final generation error rate as well.
In terms of data preparation, only the most elementary pre-processing removing null values and non-numeric characters is carried out for the two datasets. 2867 records from the CIC IDS 2017 dataset were found to be useless and removed. Also, according to [42], there were 15 classes present that were reduced to 7 classes.
Fig. 4, shows the way the fog network uses the trained model for detecting intrusive behavior. As illustrated in the figure, traffic generated by various IoT devices and received in the fog network is handled by multiple subcomponents to determine the traffic's location.
A generic architecture is also described because the proposed method's latency reduction needs to be confirmed in the fog network.
Two types of representation—local and global—are identified because of the master-slave connection and are learned utilizing the local gated recurrent unit (Local GRU) and multi-head attention mechanism. According to the findings, this strategy is ideally suited to manage the high volume of traffic in IoT-IDS.
The number of threats aimed at Internet of Things (IoT) devices has significantly increased, according to the SonicWall report [4]. According to the research, intrusions and other encrypted threats have increased worldwide by 132% while IoT malware attacks have increased by 77%. Additionally, it is predicted that this number will rise in the ensuing years.
It was necessary to create appropriate metrics like scalability and throughput to confirm the performance of the IDS in a very big network because measurements like accuracy, recall, and f1 score were insufficient to determine the performance of an IDS in such a huge network. Additionally, the majority of machine learning algorithms contributed to the overfitting issue, leading to the employment of more intricate modelling and deep learning techniques in the development of effective intrusion detection systems.
After the input has been flattened, it is supplied into the feed-forward network, which is made up of several neurons and does backpropagation. After going through several epochs, the learning significantly rises.
Additionally, creating a model with more characteristics can cause the network to be overfit. As a result, feature selection is essential for solving accurate classification problems, and feature map exploitation-based CNN is used to do this.
To create a feature map, the kernel is multiplexed in accordance with the stride value. As a result, the dimension of the feature map produced by the convolution process is decreased. The number of convolution layers in a CNN is also flexible, as is the kernel size.
As shown in Table 4, the final generation error rate and the number of parameters is significantly lower that shows with the proposed approach the detection performance and the complexity are greatly increased. The average number of parameters at both the initial and final generation shows a decreased error rate when the number of parameters is significantly reduced.
5 billion devices will be connected using these technologies [3]. Better functionality will undoubtedly result from an increase in the number, but it could also pave the way for security-related assaults. The number of threats aimed at Internet of Things (IoT) devices has significantly increased, according to the SonicWall report [4].
The best performers within the population are selected to provide quantitative values for detection accuracy. By the results, we were able to show that the suggested procedure is effective at identifying the optimal choice. In terms of result analysis, the error rate is lower following population analysis, the latency is confirmed by contrasting the cloud and fog settings, and in terms of response time, it is clear that the fog requires less time than the cloud.
The development of artificial intelligence has made it possible to create a robust IoT-IDS, which is what is required today. For an IoT-IDS to be effective, machine learning and deep learning are essential. Thus, we first examine the different methods that have been applied thus far to increase detection accuracy in this literature, and then we examine the different ways that have been utilized thus far for parameter adjustment.
The method proposed in this study better automates the hyperparameter selection by utilizing the CNN feature map and an enhanced genetic compact algorithm. For executing the categorization of IoT-IDS network traffic, the genetic algorithm is used to achieve the goal of obtaining a superior neural network with automatic adjustments. IoT traffic is categorized using the feature map CNN, which is subsequently encoded as a chromosome for hyperparameter tweaking.
The existing machine and deep learning approaches concentrate mainly on improving the detection accuracy of the intrusion detection system, and the accuracy holds significantly good for many of the models. However, the accuracy obtained cannot remain the same because the traffic volume and type are varied. This generates instability in training, hence affecting accuracy.
In terms of restricted interaction, which lowers the computation's memory demand, parameter sharing, and changes in the input directly affecting the output, this configuration of CNN's convolutional layer makes it superior.
By the results, we were able to show that the suggested procedure is effective at identifying the optimal choice. In terms of result analysis, the error rate is lower following population analysis, the latency is confirmed by contrasting the cloud and fog settings, and in terms of response time, it is clear that the fog requires less time than the cloud.
This population study is done to highlight the subset of present generational solutions. It is crucial to comprehend how evenly the individuals were dispersed throughout the first generation and in the last generation in order to say that the proposed methodology succeeded successfully. The fact that the individuals should be evenly dispersed suggests there hasn't been any premature convergence The CIC-IDS-2017 and 2018 beginning and final generation populations are plotted for the multiclass problem taking this into account.
A memory remembering technique LSTM for handling the IoT traffic data is suggested in [34]. It is not a new concept to use evolutionary algorithms in intrusion detection systems, and there is a substantial body of research addressing algorithms like Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), Artificial Bee Colony Optimization (ABC), Firefly Algorithm (FA), Bat Algorithm (BA), and Flower Pollination Algorithm (FPA). However, the majority of work using evolutionary algorithms aims to collect a minimal set of features to improve the model's classification accuracy.
The parameters used to gauge the overall performance of the IDS have been the main source of complaints for most machine learning techniques. It was necessary to create appropriate metrics like scalability and throughput to confirm the performance of the IDS in a very big network because measurements like accuracy, recall, and f1 score were insufficient to determine the performance of an IDS in such a huge network. Additionally, the majority of machine learning algorithms contributed to the overfitting issue, leading to the employment of more intricate modelling and deep learning techniques in the development of effective intrusion detection systems.
Since each device implements a security system differently, using strong encryption techniques typically requires a lot of CPU power, which most IoT network devices don't have
Two essential components of the IoT-IDS are increased compute power and accurate detection. Since the values of the hyperparameters have a significant impact on the computation resource and detection accuracy, this needs to be automatically configured to optimal so that, when it is deployed in a fog environment, it can create a lower latency. This is the primary driver behind the CNN-based feature-map exploitation conversion to a chromosome.
So, this paper suggests an Extended Compact Genetic Algorithm deep neural network to develop the IoT-IDS, similar to the strategy used in [36], where the DNN hyperparameters are updated to a chromosome that takes the Bayesian framework for optimization into account. The measures that identify the linkage learning process are the model's complexity and performance.
The CPU and memory consumption for the fog node when these nodes attempted to do this ranged from 10% to 20%, demonstrating the fog nodes' ability to support this methodology, which will result in lower latency
As shown in Table 4, the final generation error rate and the number of parameters is significantly lower that shows with the proposed approach the detection performance and the complexity are greatly increased. The average number of parameters at both the initial and final generation shows a decreased error rate when the number of parameters is significantly reduced. The point that needs to be noted here is the effect of the number of parameters on the detection accuracy.
The extended genetic compact algorithm (eCGA), which is combined with this multi-objective function conversion, is then used to optimize the neural network. The choice of eCGA is made in this work because the encoding of eCGA is using the Bayesian network, and if the traditional eCGA, which uses probabilistic polling for subset generation, is modified to be deterministic, the computation resources will be reduced and it can process faster.
15, 16 and Table 5, 6 for CIC IDS 2017 and 2018 data. The detection accuracy outperformed in both the dataset with a value above 99%.
The existing machine and deep learning approaches concentrate mainly on improving the detection accuracy of the intrusion detection system, and the accuracy holds significantly good for many of the models. However, the accuracy obtained cannot remain the same because the traffic volume and type are varied.
The fact that the individuals should be evenly dispersed suggests there hasn't been any premature convergence The CIC-IDS-2017 and 2018 beginning and final generation populations are plotted for the multiclass problem taking this into account
The number of parameters and error rate have increased since the distribution is noticeably greater. The extreme values for both datasets are displayed in Table 4, for a quantitative examination.
The pretrained eCGA-based model must be able to execute in the fog node for the claim of lower latency to be valid, and the amount of time it takes to react to attacks must also be confirmed. These fog nodes are simulated using the iFogSim simulation environment, and the pretrained models are made to operate in those nodes, in order to test this.
Despite the automated hyperparameter adjustment capabilities of the proposed model, this system is evaluated on balanced data due to sample availability. These model parameters tuning findings must be confirmed on a significantly unbalanced dataset, though. Adversarial synthetic data production is required to include the data's imbalance.
Since each device implements a security system differently, using strong encryption techniques typically requires a lot of CPU power, which most IoT network devices don't have. This necessitates a particular kind of IDS that can operate in spite of heterogeneity and also recognize and respond to an intrusion attempt like a human.
Using the Extended Compact Genetic Algorithm, a novel evolutionary deep neural network is suggested for IoT-IDS in the fog environment (eCGA). This will determine which DNN model, in terms of accuracy and latency, should be used in the fog node.
If the right parameters are not selected, it will have an impact on factors like training duration, computational expense, structure, and prediction accuracy. Thus, the auto-hyperparameter setting is crucial for the intrusion detection data so that the intrusion detection system can act more like a human. To circumvent this problem, the intrusion detection system, if equipped with automatic hyperparameter tuning, can set an ideal parameter for any type of traffic and volume.
For an IoT-IDS to be effective, machine learning and deep learning are essential. Thus, we first examine the different methods that have been applied thus far to increase detection accuracy in this literature, and then we examine the different ways that have been utilized thus far for parameter adjustment.
Thus, with the plotting of the population at the initial and final stages of population denotes a substantial reduction in the error rate and this clearly showed the relationship between the number of parameters and the error. With the reduction in error, the detection accuracy is improved and that is compared to the state-of-the-art for its superiority and the latency analysis exhibits the suitability of this approach at the fog node.
Two essential components of the IoT-IDS are increased compute power and accurate detection. Since the values of the hyperparameters have a significant impact on the computation resource and detection accuracy, this needs to be automatically configured to optimal so that, when it is deployed in a fog environment, it can create a lower latency.
Two types of representation—local and global—are identified because of the master-slave connection and are learned utilizing the local gated recurrent unit (Local GRU) and multi-head attention mechanism
When examining the CNN feature map's structure, it becomes evident that selecting the appropriate amount of convolution layers, pooling layers, and fully linked layers was challenging
The Extended Compact Genetic Algorithm's goal is to minimize the classification error rate (eCGA) with the decrease in the number of parameters. With fewer relevant parameters, the accuracy of the model is increased.
The measures that identify the linkage learning process are the model's complexity and performance. With the linkage learning process, a higher number of parameters results in a better detection value, which is desirable; nevertheless, a lesser number of parameters in DNN has an effect on performance. The objective is to optimize these metrics so that the parameters and detection rate of the ideal DNN setup are improved.
Thus, with the plotting of the population at the initial and final stages of population denotes a substantial reduction in the error rate and this clearly showed the relationship between the number of parameters and the error. With the reduction in error, the detection accuracy is improved and that is compared to the state-of-the-art for its superiority and the latency analysis exhibits the suitability of this approach at the fog node.

후속연구

The main goal of this research is to provide an enhanced-performance IoT-IDS with a high detection rate and minimal delay consumption. The state of the art in studies relevant to this study is discussed in this section, along with the present state of knowledge in those fields.
Adversarial synthetic data production is required to include the data's imbalance. The next goal of this study is to verify the integration of Jacobian saliency approaches with SMOTE and GAN networks to produce adversarial samples instantaneously and create such a massive amount of diverse data.
With the linkage learning process, a higher number of parameters results in a better detection value, which is desirable; nevertheless, a lesser number of parameters in DNN has an effect on performance. The objective is to optimize these metrics so that the parameters and detection rate of the ideal DNN setup are improved. A special fusion of genetic and probabilistic methods, probabilistic model-building genetic algorithms are used for this optimization.
Table 2, includes the same information in tabular style along with the number of each traffic source and its percentage. These simulated datasets were chosen because of their authenticity; even though the dataset demonstrates class imbalance, it can be useful to do a meta-analysis using them. Also, they are more similar to actual IoT traffic statistics.

참고문헌 (51)

Khan. R, Khan. S.U, Zaheer. R, and Khan. S, "Future internet: the internet of things architecture, possible applications and key challenges," in Proc. of 2012 10th international conference on frontiers of information technology, pp. 257-260, Dec. 2012.
Perera. C and Vasilakos. A.V, "A knowledge-based resource discovery for Internet of Things," Knowledge-Based Systems, 109, pp. 122-136, 2016.
Al-Fuqaha. A, Guizani. M, Mohammadi. M, Aledhari. M, and Ayyash. M, "Internet of things: A survey on enabling technologies, protocols, and applications," IEEE communications surveys & tutorials, 17(4), pp. 2347-2376, 2015.

상세보기
Mid-Year Update: SonicWall Cyber Threat Report, 2022. [Online]. Available: https://www.sonicwall.com/resources/white-papers/mid-year-2022-sonicwall-cyber-threat-report/
Forestiero. A, "Metaheuristic algorithm for anomaly detection in Internet of Things leveraging on a neural-driven multiagent system," Knowledge-Based Systems, 228, p. 107241, 2021.
Hassan. M, Huda. S, Sharmeen. S, Abawajy. J, and Fortino. G, "An adaptive trust boundary protection for IIoT networks using deep-learning feature-extraction-based semisupervised model," IEEE Transactions on Industrial Informatics, 17(4), pp. 2860-2870, 2021.

상세보기
Vijayakumar. D.S and Ganapathy. S, "Machine Learning Approach to Combat False Alarms in Wireless Intrusion Detection System," Comput. Inf. Sci., 11(3), pp. 67-81, 2018.
Hindy. H, Bayne. E, Bures. M, Atkinson. R, Tachtatzis. C, and Bellekens. X, "Machine learning based IoT intrusion detection system: An MQTT case study (MQTT-IoT-IDS2020 dataset)," in Proc. of 12th International Networking Conference: INC, pp. 73-84, 2021.
Al-Omari. M, Rawashdeh. M, Qutaishat. F, Alshira'H. M, and Ababneh. N, "An intelligent tree-based intrusion detection model for cyber security," Journal of Network and Systems Management, 29, pp. 1-18, 2021.
Vijayakumar. D.S and Ganapathy. S, "Multistage ensembled classifier for wireless intrusion detection system," Wireless Personal Communications, 122(1), pp. 645-668, 2022.

상세보기
de Araujo-Filho. P.F, Kaddoum. G, Campelo. D.R, Santos. A.G, Macedo. D, and Zanchettin. C, "Intrusion detection for cyber-physical systems using generative adversarial networks in fog environment," IEEE Internet of Things Journal, 8(8), pp. 6247-6256, 2021.

상세보기
Chen. Y, Ashizawa. N, Yeo. C.K, Yanai. N, and Yean. S, "Multi-scale self-organizing map assisted deep autoencoding Gaussian mixture model for unsupervised intrusion detection," Knowledge-Based Systems, 224, pp. 107086, 2021.
Ravi. N and Shalinie. S.M, "Semisupervised-learning-based security to detect and mitigate intrusions in IoT network," IEEE Internet of Things Journal, 7(11), pp. 11041-11052, 2020.

상세보기
Suganuma. M, Shirakawa. S, and Nagao. T, "A genetic programming approach to designing convolutional neural network architectures," in Proc. of the genetic and evolutionary computation conference, pp. 497-504, 2017.
Stanley. K.O and Miikkulainen. R, "Evolving neural networks through augmenting topologies," Evolutionary computation, 10(2), pp. 99-127, 2002.

상세보기
Yao. X and Liu. Y, "A new evolutionary system for evolving artificial neural networks," IEEE transactions on neural networks, 8(3), pp. 694-713, 1997.

상세보기
Angeline. P.J, Saunders. G.M, and Pollack. J.B, "An evolutionary algorithm that constructs recurrent neural networks," IEEE transactions on Neural Networks, 5(1), pp. 54-65, 1994.

상세보기
Miikkulainen. R, Liang. J, Meyerson. E, Rawal. A, Fink. D, Francon. O, Raju. B, Shahrzad. H, Navruzyan. A, Duffy. N, and Hodjat. B, "Evolving deep neural networks," in Artificial intelligence in the age of neural networks and brain computing, Academic Press, 2019, pp. 293-312.
Nayak. J, Naik. B, Dash. P.B, Vimal. S, and Kadry. S, "Hybrid Bayesian optimization hyper tuned catboost approach for malicious access and anomaly detection in IoT nomaly framework," Sustainable Computing: Informatics and Systems, 36, p. 100805, 2022.

상세보기
Masum. M, Shahriar. H, Haddad. H, Faruk. M.J.H, Valero. M, Khan. M.A, Rahman. M.A, Adnan. M.I, Cuzzocrea. A, and Wu. F, "Bayesian hyperparameter optimization for deep neural network-based network intrusion detection," in Proc. of 2021 IEEE International Conference on Big Data, pp. 5413-5419, Dec. 2021.
Sharafaldin. I, Lashkari. A.H, and Ghorbani. A.A, "Toward generating a new intrusion detection dataset and intrusion traffic characterization," in Proc. of ICISSp, 1, pp. 108-116, 2018.
Ren. Y, Zhang. L, and Suganthan. P.N, "Ensemble classification and regression-recent developments, applications and future directions," IEEE Computational intelligence magazine, 11(1), pp. 41-53, 2016.
Jabbar. M.A and Aluvalu. R, "RFAODE: A novel ensemble intrusion detection system," Procedia computer science, 115, pp. 226-234, 2017.

상세보기
Gaikwad. D.P and Thool. R.C, "Intrusion detection system using bagging ensemble method of machine learning," in Proc. of 2015 international conference on computing communication control and automation, pp. 291-295, Feb. 2015.
Moustafa. N, Turnbull. B, and Choo. K.K.R, "An ensemble intrusion detection technique based on proposed statistical flow features for protecting network traffic of internet of things," IEEE Internet of Things Journal, 6(3), pp. 4815-4830, 2018.
McLaughlin. N, del Rincon. J.M, Kang. B, Yerima. S, and Miller. P, "Deep android malware detection," in Proc. of the Seventh ACM on Conference on Data and Application Security and Privacy, pp. 301-308, 2017.
Maghrebi. H, Portigliatti. T, and Prouff. E, "Breaking cryptographic implementations using deep learning techniques," in Proc. of Security, Privacy, and Applied Cryptography Engineering: 6th International Conference, SPACE, pp. 3-26, Dec. 2016.
Brun. O, Yin. Y, Gelenbe. E, Kadioglu. Y.M, Augusto-Gonzalez. J, and Ramos. M, "Deep learning with dense random neural networks for detecting attacks against IoT-connected home environments," Procedia Computer Science, vol. 134, pp. 458-463, Feb. 2018.
Diro. A.A and Chilamkurti. N, "Distributed attack detection scheme using deep learning approach for Internet of Things," Future Generation Computer Systems, 82, pp. 761-768, 2018.

상세보기
Torres. P, Catania. C, Garcia. S, and Garino. C.G, "An analysis of recurrent neural networks for botnet detection behavior," in Proc. of 2016 IEEE biennial congress of Argentina (ARGENCON), pp. 1-6, 2016.
Atiga. J, Mbarki. N.E, Ejbali. R, and Zaied. M, "Faulty node detection in wireless sensor networks using a recurrent neural network," in Proc. of Tenth international conference on machine vision (ICMV 2017), Vol. 10696, pp. 711-716, 2018.
Vu. L, Nguyen. Q.U, Nguyen. D.N, Hoang. D.T, and Dutkiewicz. E, "Learning latent representation for iot anomaly detection," IEEE Transactions on Cybernetics, 52(5), pp. 3769-3782, 2022.

상세보기
Abdel-Basset. M, Chang. V, Hawash. H, Chakrabortty. R.K, and Ryan. M, "Deep-IFS: Intrusion detection approach for industrial internet of things traffic in fog environment," IEEE Transactions on Industrial Informatics, 17(11), pp. 7704-7715, 2020.
Imrana. Y, Xiang. Y, Ali. L, and Abdul-Rauf. Z, "A bidirectional LSTM deep learning approach for intrusion detection," Expert Systems with Applications, 185, p. 115524, 2021.

상세보기
Habib. M, Aljarah. I, and Faris. H, "A modified multi-objective particle swarm optimizer-based Levy flight: An approach toward intrusion detection in Internet of Things," Arabian Journal for Science and Engineering, 45, pp. 6081-6108, 2020.

상세보기
Chen. Y, Lin. Q, Wei. W, Ji. J, Wong. K.C, and Coello. C.A.C, "Intrusion detection using multi-objective evolutionary convolutional neural network for Internet of Things in Fog computing," Knowledge-Based Systems, 244, p. 108505, 2022.
Zhu. Y, Liang. J, Chen. J, and Ming. Z, "An improved NSGA-III algorithm for feature selection used in intrusion detection," Knowledge-Based Systems, 116, pp. 74-85, 2017.
Y.H. Yang, H.Z. Huang, Q.N. Shen, Z.H. Wu, and Y. Zhang, "Research on intrusion detection based on incremental GHSOM," Chinese J. Comput. 37(5), pp. 1216-1224, 2014
LeCun. Y and Bengio. Y, "Convolutional networks for images, speech, and time series," The handbook of brain theory and neural networks, 3361(10), p. 1995, 1995.
Kim. J, Shin. Y, and Choi. E, "An intrusion detection model based on a convolutional neural network," Journal of Multimedia Information System, 6(4), pp. 165-172, 2019.

원문보기 상세보기
Kursawe. F, "A variant of evolution strategies for vector optimization," in Proc. of Parallel Problem Solving from Nature: 1st Workshop, pp. 193-197, June. 2005.
Panigrahi. R and Borah. S, "A detailed analysis of CICIDS2017 dataset for designing Intrusion Detection Systems," International Journal of Engineering & Technology, 7(3.24), pp. 479-482, 2018.
Assis. M.V, Carvalho. L.F, Lloret. J, and Proenca Jr. M.L, "A GRU deep learning system against attacks in software defined networks," Journal of Network and Computer Applications, 177, p. 102942, 2021.

상세보기
Vinayakumar. R, Soman. K.P, and Poornachandran. P, "A comparative analysis of deep learning approaches for network intrusion detection systems (N-IDSs): deep learning for N-IDSs," International Journal of Digital Crime and Forensics (IJDCF), 11(3), pp. 65-89, 2019.
Gamage. S and Samarabandu. J, "Deep learning methods in network intrusion detection: A survey and an objective comparison," Journal of Network and Computer Applications, 169, p. 102767, 2020.

상세보기
Lv. Z, Qiao. L, Li. J, and Song. H, "Deep-learning-enabled security issues in the internet of things," IEEE Internet of Things Journal, 8(12), pp. 9531-9538, 2020.
Muhammad. G, Hossain. M.S, and Garg. S, "Stacked autoencoder-based intrusion detection system to combat financial fraudulent," IEEE Internet of Things Journal, vol. 10, no. 3, pp. 2071-2078, 2023.

상세보기
Altunay. H.C and Albayrak. Z, "A hybrid CNN+ LSTM based intrusion detection system for industrial IoT networks," Engineering Science and Technology, an International Journal, 38, p. 101322, 2023.

상세보기
Altunay. H.C and Albayrak. Z, "Network Intrusion Detection Approach Based on Convolutional Neural Network," Avrupa Bilim ve Teknoloji Dergisi, (26), pp. 22-29, 2021.
Kavitha. S, Uma Maheswari. N, and Venkatesh. R, "Intelligent Intrusion Detection System using Enhanced Arithmetic Optimization Algorithm with Deep Learning Model," Tehnicki vjesnik, 30(4), pp. 1217-1224, 2023.
Khan. M.A, Iqbal. N, Jamil. H, and Kim. D.H, "An optimized ensemble prediction model using AutoML based on soft voting classifier for network intrusion detection," Journal of Network and Computer Applications, 212, p. 103560, 2023.

상세보기

표제어: PCR

동의어: Packet Collision Rate

용어 설명 출처 목록 (6)

용어 설명: PCR은 세균 특이성이 있는 primer를 이용하여 적은 수의 세균이 있을지라도 쉽게 검출할 수 있는 유용한 방법이며, 이를 이용하여 구강 내 치면세균막이나 타액에서 직접 세균을 검출할 수 있게 되었다[8].

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증