Deep Autoencoder-Based Integrated Model for Anomaly Detection and Efficient Feature Extraction in IoT Networks
Abstract
:1. Introduction
- We propose a new lightweight IDS system for IoT networks by fully utilizing a deep AE model to do anomaly detection, and feature reduction for multi-classification of the detected cyber-attacks, unlike the existing methods that used AE for either anomaly detection or feature reduction;
- The proposed system is extensively evaluated with real datasets, namely N-BaIoT and MQTTset, which contain normal and malicious network traffic. Classification performances of five IoT devices are evaluated based on accuracy, precision, recall, F1 score, and execution time;
- The effectiveness of the proposed IDS system is compared with state-of-the-art methods.
2. Related Works
3. Methodology
3.1. Autoencoder (AE)
3.2. Proposed System
4. Results and Discussion
4.1. Datasets
4.2. Experiment Setup
4.3. Anomaly Detection
4.4. Feature Extraction
4.5. Multi-Classification
5. Conclusions
Author Contributions
Funding
Data Availability Statement
Conflicts of Interest
References
- Cruz, A.R.S.A.; Gomes, R.L.; Fernandez, M.P. An Intelligent Mechanism to Detect Cyberattacks of Mirai Botnet in IoT Networks. In Proceedings of the 2021 17th International Conference on Distributed Computing in Sensor Systems (DCOSS), Pafos, Cyprus, 14–16 July 2021; pp. 236–243. [Google Scholar] [CrossRef]
- Marzano, A.; Alexander, D.; Fonseca, O.; Fazzion, E.; Hoepers, C.; Jessen, K.S.; Chaves, M.H.; Cunha, I.; Guedes, D.; Meira, W. The evolution of Bashlite and Mirai IoT botnets. In Proceedings of the 2018 IEEE Symposium on Computers and Communications (ISCC), Natal, Brazil, 25–28 June 2018; pp. 00813–00818. [Google Scholar]
- Mira Botnet Source Code. Available online: https://github.com/jgamblin/Mirai-Source-Code (accessed on 1 October 2022).
- Alaghbari, K.A.; Saad, M.H.; Hussain, A.; Alam, M.R. Complex event processing for physical and cyber security in datacenters—Recent progress, challenges and recommendations. J. Cloud Comp. 2022, 11, 65. [Google Scholar] [CrossRef]
- Halbouni, A.; Gunawan, T.S.; Habaebi, M.H.; Halbouni, M.; Kartiwi, M.; Ahmad, R. CNN-LSTM: Hybrid Deep Neural Network for Network Intrusion Detection System. IEEE Access 2022, 10, 99837–99849. [Google Scholar] [CrossRef]
- Alaghbari, K.A.; Saad, M.H.M.; Hussain, A.; Alam, M.R. Activities Recognition, Anomaly Detection and Next Activity Prediction Based on Neural Networks in Smart Homes. IEEE Access 2022, 10, 28219–28232. [Google Scholar] [CrossRef]
- Ullah, I.; Mahmoud, Q.H. Design and Development of a Deep Learning-Based Model for Anomaly Detection in IoT Networks. IEEE Access 2021, 9, 103906–103926. [Google Scholar] [CrossRef]
- Bacha, S.; Aljuhani, A.; Abdellafou, K.B.; Taouali, O.; Liouane, N.; Alazab, M. Anomaly-based intrusion detection system in IoT using kernel extreme learning machine. J. Ambient. Intell. Humaniz. Comput. 2022. [Google Scholar] [CrossRef]
- Hairab, B.I.; Elsayed, M.S.; Jurcut, A.D.; Azer, M.A. Anomaly Detection Based on CNN and Regularization Techniques against Zero-Day Attacks in IoT Networks. IEEE Access 2022, 10, 98427–98440. [Google Scholar] [CrossRef]
- Jia, W.; Sun, M.; Lian, J.; Hou, S. Feature dimensionality reduction: A review. Complex Intell. Syst. 2022, 8, 2663–2693. [Google Scholar] [CrossRef]
- Yeom, S.; Choi, C.; Kim, K. AutoEncoder Based Feature Extraction for Multi-Malicious Traffic Classification. In Proceedings of the 9th International Conference on Smart Media and Applications (SMA 2020), Jeju, Republic of Korea, 17–19 September 2020; Association for Computing Machinery: New York, NY, USA, 2020; pp. 285–287. [Google Scholar] [CrossRef]
- Wang, Y.; Yao, H.; Zhao, S. Auto-Encoder Based Dimensionality Reduction. Neurocomputing 2016, 184, 232–242. [Google Scholar] [CrossRef]
- Meidan, Y.; Bohadana, M.; Mathov, Y.; Mirsky, Y.; Shabtai, A.; Breitenbacher, D.; Elovici, Y. N-BaIoT network-based detection of IoT botnet attacks using deep autoencoders. IEEE Pervasive Comput. 2018, 17, 12–22. [Google Scholar] [CrossRef]
- Alkahtani, H.; Aldhyani, T.H. Botnet Attack Detection by Using CNN-LSTM Model for Internet of Things Applications. Secur. Commun. Netw. 2021, 2021, 3806459. [Google Scholar] [CrossRef]
- Al Dahoul, N.; Abdul Karim, H.; Ba Wazir, A.S. Model fusion of deep neural networks for anomaly detection. J. Big Data 2021, 8, 106. [Google Scholar] [CrossRef]
- Aygun, R.C.; Yavuz, A.G. Network Anomaly Detection with Stochastically Improved Autoencoder Based Models. In Proceedings of the 2017 IEEE 4th International Conference on Cyber Security and Cloud Computing (CSCloud), New York, NY, USA, 26–28 June 2017; pp. 193–198. [Google Scholar] [CrossRef]
- Zavrak, S.; İskefiyeli, M. Anomaly-Based Intrusion Detection from Network Flow Features Using Variational Autoencoder. IEEE Access 2020, 8, 108346–108358. [Google Scholar] [CrossRef]
- Min, B.; Yoo, J.; Kim, S.; Shin, D.; Shin, D. Network Anomaly Detection Using Memory-Augmented Deep Autoencoder. IEEE Access 2021, 9, 104695–104706. [Google Scholar] [CrossRef]
- Kunang, Y.N.; Nurmaini, S.; Stiawan, D.; Zarkasi, A. Automatic Features Extraction Using Autoencoder in Intrusion Detection System. In Proceedings of the 2018 International Conference on Electrical Engineering and Computer Science (ICECOS), Pangkal, Indonesia, 2–4 October 2018; pp. 219–224. [Google Scholar] [CrossRef]
- Mhamdi, L.; McLernon, D.; El-moussa, F.; Zaidi, S.A.R.; Ghogho, M.; Tang, T. A Deep Learning Approach Combining Autoencoder with One-class SVM for DDoS Attack Detection in SDNs. In Proceedings of the 2020 IEEE Eighth International Conference on Communications and Networking (ComNet), Hammamet, Tunisia, 27–30 October 2020; pp. 1–6. [Google Scholar] [CrossRef]
- Chen, Z.; Yeo, C.K.; Lee, B.S.; Lau, C.T. Autoencoder-based network anomaly detection. In Proceedings of the 2018 Wireless Telecommunications Symposium (WTS), Phoenix, AZ, USA, 17–20 April 2018; pp. 1–5. [Google Scholar] [CrossRef]
- Elsayed, M.S.; Le-Khac, N.-A.; Dev, S.; Jurcut, A.D. Network Anomaly Detection Using LSTM Based Autoencoder. In Proceedings of the 16th ACM Symposium on QoS and Security for Wireless and Mobile Networks (Q2SWinet ′20), Alicante, Spain, 16–20 November 2020; Association for Computing Machinery: New York, NY, USA, 2020; pp. 37–45. [Google Scholar] [CrossRef]
- Rao, K.N.; Rao, K.V.; Reddy, P.V.G.D.P. A hybrid intrusion detection system based on sparse autoencoder and deep neural network. Comput. Commun. 2021, 180, 77–88. [Google Scholar] [CrossRef]
- Al-Fawa’reh, M.; Al-Fayoumi, M.; Nashwan, S.; Fraihat, S. Cyber threat intelligence using PCA-DNN model to detect abnormal network behavior. Egypt. Inform. J. 2021, 23, 173–185. [Google Scholar] [CrossRef]
- Xu, W.; Jang-Jaccard, J.; Singh, A.; Wei, Y.; Sabrina, F. Improving Performance of Autoencoder-Based Network Anomaly Detection on NSL-KDD Dataset. IEEE Access 2021, 9, 140136–140146. [Google Scholar] [CrossRef]
- Vaccari, I.; Chiola, G.; Aiello, M.; Mongelli, M.; Cambiaso, E. MQTTset, a New Dataset for Machine Learning Techniques on MQTT. Sensors 2020, 20, 6578. [Google Scholar] [CrossRef] [PubMed]
- Ullah, I.; Mahmoud, Q.H. An Anomaly Detection Model for IoT Networks based on Flow and Flag Features using a Feed-Forward Neural Network. In Proceedings of the 2022 IEEE 19th Annual Consumer Communications & Networking Conference (CCNC), Las Vegas, NV, USA, 8–11 January 2022; pp. 363–368. [Google Scholar] [CrossRef]
IoT | Device Name | Data Size (Samples) | Total | Cyber-Attack | No. Classes | ||
---|---|---|---|---|---|---|---|
Normal | Attack | ||||||
IoT1 | Danmini_Doorbell | 49,548 | 968,750 | 1,018,298 | Mirai (M.): (1) Scan (2) Ack flooding (3) Syn flooding (4) UDP flooding (5) UDPplain | Gafgyt (G.): (1) Scan (2) Junk (3) UDP flooding (4) TCP flooding (5) COMBO | 10 |
IoT2 | Philips_B120N10_Baby_Monitor | 175,240 | 923,437 | 1,098,677 | 10 | ||
IoT3 | SimpleHome_XCS7_1003_WHT_ Security_Camera | 14,646 | 836,180 | 850,826 | 10 | ||
IoT4 | Samsung_SNH_1011_N_Webcam | 52,150 | 323,072 | 375,222 | - | 5 | |
IoT5 | 10 sensors: 2 Motion sensors, Door lock, Fan sensor, Fan speed controller, Smoke, CO-Gas, Temperature, Humidity, Light intensity. | 440,699 | 100,044 | 540,743 | (1) MQTT Publish Flood (2) Brute Force Authentication (3) Flooding DoS (4) Malformed (5) SlowITe | 5 |
Dev. | Classifier | Confusion Matrix Elements | Metrics | Data Size | Time (Second) | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
TP | FN | FP | TN | Acc. | Pre. | Rec | F1.sc | Training/Valid.) | Testing | Training | Testing | ||
IoT1 | OC-SVM | 12,280 | 107 | 0 | 968,750 | 1.00 | 1.00 | 1.00 | 1.00 | (37,161, 115) | (981137, 115) | 2.81 | 83.49 |
iForest | 11,649 | 738 | 0 | 968,750 | 1.00 | 1.00 | 0.97 | 0.98 | 3.22 | 81.91 | |||
AE-2 | 12,310 | 77 | 18 | 968,732 | 1.00 | 1.00 | 1.00 | 1.00 | (27,870, 115) (9291, 115) | 35.22 | 29.3 | ||
AE-3 | 12,304 | 83 | 0 | 968,750 | 1.00 | 1.00 | 1.00 | 1.00 | 30.02 | 26.54 | |||
AE-8 | 12,326 | 61 | 0 | 968,750 | 1.00 | 1.00 | 1.00 | 1.00 | 73.1 | 28.3 | |||
IoT2 | OC-SVM | 43,524 | 286 | 171 | 923,266 | 1.00 | 1.00 | 1.00 | 1.00 | (131,430, 115) | (967247, 115) | 28.79 | 234.94 |
iForest | 42,286 | 1524 | 198,841 | 724,596 | 0.79 | 0.59 | 0.87 | 0.59 | 9.18 | 147.34 | |||
AE-2 | 42,714 | 1096 | 130 | 923,307 | 1.00 | 1.00 | 0.99 | 0.99 | (98,572, 115) (32,858, 115) | 138.26 | 25.85 | ||
AE-3 | 42,525 | 1285 | 136 | 923,301 | 1.00 | 1.00 | 0.99 | 0.99 | 115.61 | 29.41 | |||
AE-8 | 42,606 | 1204 | 149 | 923,288 | 1.00 | 1.00 | 0.99 | 0.99 | 122.42 | 26.53 | |||
IoT3 | OC-SVM | 4861 | 21 | 119 | 831,179 | 1.00 | 0.99 | 1.00 | 0.99 | (14,646, 115) | (836180, 115) | 0.42 | 29.32 |
iForest | 4693 | 189 | 1 | 831,297 | 1.00 | 1.00 | 0.98 | 0.99 | 1.22 | 75.57 | |||
AE-2 | 4758 | 124 | 0 | 831,298 | 1.00 | 1.00 | 0.99 | 0.99 | (10,984, 115) (3662, 115) | 13.74 | 26.91 | ||
AE-3 | 4786 | 96 | 0 | 831,298 | 1.00 | 1.00 | 0.99 | 1.00 | 14.1 | 27.6 | |||
AE-8 | 4793 | 89 | 13 | 831,285 | 1.00 | 1.00 | 0.99 | 0.99 | 17.73 | 29.1 | |||
IoT4 | OC-SVM | 12,863 | 175 | 142 | 322,930 | 1.00 | 0.99 | 0.99 | 0.99 | (39,112, 115) | (336110, 115) | 2.8 | 28 |
iForest | 12,131 | 907 | 30 | 323,042 | 1.00 | 1.00 | 0.97 | 0.98 | 3.14 | 32.3 | |||
AE-2 | 12,062 | 976 | 1 | 323,071 | 1.00 | 1.00 | 0.96 | 0.98 | (29,334, 115) (9778, 115) | 37.3 | 10.74 | ||
AE-3 | 12,430 | 608 | 33 | 323,039 | 1.00 | 1.00 | 0.98 | 0.99 | 32.1 | 8.8 | |||
AE-8 | 12,440 | 598 | 6 | 323,066 | 1.00 | 1.00 | 0.98 | 0.99 | 33.54 | 10.1 | |||
IoT5 | OC-SVM | 109,129 | 1046 | 0 | 100,044 | 1.00 | 0.99 | 1.00 | 1.00 | (330,524, 76) | (210219, 76) | 395.39 | 122 |
iForest | 96,119 | 14,056 | 0 | 100,044 | 0.93 | 0.94 | 0.94 | 0.93 | 14.81 | 13.42 | |||
AE-2 | 109,193 | 982 | 0 | 100,044 | 1.00 | 1.00 | 1.00 | 1.00 | (247,893, 76) (82,631, 76) | 287 | 6.3 | ||
AE-3 | 109,184 | 991 | 0 | 100,044 | 1.00 | 1.00 | 1.00 | 1.00 | 328 | 6.5 | |||
AE-8 | 109,189 | 986 | 0 | 100,044 | 1.00 | 1.00 | 1.00 | 1.00 | 342 | 7 |
Dev. | Dataset | PCA (Compo) | LDA Comp | Data Size | |||
---|---|---|---|---|---|---|---|
2 | 8 | 10 | 2 | 4 | |||
IoT1 | Training () | 4.21 | 7.74 | 8.11 | 18.31 | 21 | (726,563, 115) |
) | 0.27 | 0.29 | 0.18 | 0.44 | 0.55 | (242,188, 115) | |
IoT2 | ) | 5 | 9.11 | 7.1 | 27.3 | 17.61 | (692,577, 115) |
) | 0.37 | 0.83 | 0.16 | 1.11 | 0.8 | (230,860, 115) | |
IoT3 | ) | 4.43 | 5.68 | 6.47 | 20.05 | 15.94 | (623,473, 115) |
) | 0.168 | 0.167 | 0.17 | 0.29 | 0.3 | (207,825, 115) | |
IoT4 | ) | 1.59 | 2 | 2 | 5.2 | 4.86 | (242,304, 115) |
) | 0.06 | 0.06 | 0.07 | 0.17 | 0.09 | (80,768, 115) | |
IoT5 | ) | 0.3 | 0.3 | 0.37 | 0.72 | 0.88 | (75,033, 76) |
) | 0.01 | 0.01 | 0.01 | 0.012 | 0.016 | (25,011, 76) |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | ||||
---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1-sc. | Training | Testing | ||
DT | Without | 0.91 | 0.92 | 0.90 | 0.87 | 28.32 | 0.18 |
PCA-2 | 0.84 | 0.85 | 0.83 | 0.80 | 4.68 | 0.11 | |
PCA-8 | 0.90 | 0.92 | 0.90 | 0.87 | 11.55 | 0.07 | |
PCA-10 | 0.90 | 0.93 | 0.90 | 0.87 | 13 | 0.036 | |
LDA-2 | 0.88 | 0.86 | 0.84 | 0.81 | 1.89 | 0.038 | |
LDA-4 | 0.90 | 0.92 | 0.90 | 0.87 | 2.8 | 0.04 | |
Proposed (AE-2) | 0.73 | 0.77 | 0.75 | 0.72 | 3.42 | 0.092 | |
Proposed (AE-3) | 0.88 | 0.90 | 0.88 | 0.85 | 4.38 | 0.03 | |
Proposed (AE-8) | 0.90 | 0.92 | 0.90 | 0.87 | 10.22 | 0.05 | |
XTree | Without | 0.91 | 0.93 | 0.90 | 0.87 | 0.84 | 0.11 |
PCA-2 | 0.84 | 0.83 | 0.83 | 0.80 | 0.41 | 0.075 | |
PCA-8 | 0.90 | 0.93 | 0.90 | 0.87 | 0.48 | 0.047 | |
PCA-10 | 0.90 | 0.92 | 0.90 | 0.87 | 0.62 | 0.046 | |
LDA-2 | 0.88 | 0.86 | 0.84 | 0.81 | 0.45 | 0.046 | |
LDA-4 | 0.90 | 0.93 | 0.90 | 0.87 | 0.53 | 0.038 | |
Proposed (AE-2) | 0.73 | 0.77 | 0.74 | 0.71 | 0.48 | 0.11 | |
Proposed (AE-3) | 0.86 | 0.87 | 0.87 | 0.84 | 0.36 | 0.071 | |
Proposed (AE-8) | 0.90 | 0.92 | 0.90 | 0.87 | 0.4 | 0.046 | |
RF | Without | 0.91 | 0.93 | 0.90 | 0.87 | 184.33 | 3.23 |
PCA-2 | 0.85 | 0.87 | 0.84 | 0.82 | 149 | 6.4 | |
PCA-8 | 0.90 | 0.92 | 0.90 | 0.87 | 229.69 | 5.74 | |
PCA-10 | 0.90 | 0.93 | 0.90 | 0.87 | 270.06 | 3.93 | |
LDA-2 | 0.88 | 0.89 | 0.85 | 0.82 | 92.56 | 3.52 | |
LDA-4 | 0.90 | 0.93 | 0.90 | 0.87 | 110.08 | 3.3 | |
Proposed (AE-2) | 0.75 | 0.80 | 0.76 | 0.73 | 145.44 | 7.67 | |
Proposed (AE-3) | 0.88 | 0.91 | 0.89 | 0.85 | 110.91 | 5.45 | |
Proposed (AE-8) | 0.90 | 0.93 | 0.90 | 0.87 | 178 | 3.8 | |
NB | Without | 0.86 | 0.82 | 0.80 | 0.75 | 1.91 | 2.94 |
PCA-2 | 0.59 | 0.35 | 0.45 | 0.38 | 0.16 | 0.13 | |
PCA-8 | 0.72 | 0.69 | 0.68 | 0.67 | 0.24 | 0.31 | |
PCA-10 | 0.72 | 0.69 | 0.69 | 0.67 | 0.25 | 0.34 | |
LDA-2 | 0.87 | 0.75 | 0.80 | 0.75 | 0.16 | 0.13 | |
LDA-4 | 0.87 | 0.78 | 0.80 | 0.75 | 0.18 | 0.19 | |
Proposed (AE-2) | 0.61 | 0.58 | 0.52 | 0.47 | 0.14 | 0.14 | |
Proposed (AE-3) | 0.53 | 0.45 | 0.51 | 0.45 | 0.14 | 0.16 | |
Proposed (AE-8) | 0.69 | 0.63 | 0.69 | 0.63 | 0.18 | 0.31 |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | Data Size (Training, Testing) | ||||
---|---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1-sc. | Train. | Test. | |||
CNN-LSTM [14] | Without | 0.91 | 0.93 | 0.91 | 0.88 | - | - | (70%, 30%) |
DNN1 [X,64,32,10] Epochs = 100, batch = 64 | Without | 0.90 | 0.94 | 0.89 | 0.86 | 1356 | 2.8 | (75%, 25%) |
PCA-2 | 0.83 | 0.80 | 0.80 | 0.77 | 993 | 1.97 | ||
PCA-8 | 0.90 | 0.92 | 0.89 | 0.86 | 1083 | 1.35 | ||
PCA-10 | 0.89 | 0.93 | 0.89 | 0.86 | 1221 | 4.8 | ||
LDA-2 | 0.88 | 0.86 | 0.82 | 0.79 | 1054 | 2.15 | ||
LDA-4 | 0.88 | 0.89 | 0.82 | 0.78 | 1429 | 2 | ||
Proposed (AE-2) | 0.74 | 0.70 | 0.73 | 0.71 | 921 | 1.9 | ||
Proposed (AE-3) | 0.86 | 0.90 | 0.85 | 0.82 | 896 | 1.26 | ||
Proposed (AE-8) | 0.90 | 0.91 | 0.89 | 0.86 | 923 | 1.5 | ||
DNN2 [X,64,64,64,10] Epochs = 100 batch = 128 | Without | 0.90 | 0.93 | 0.90 | 0.87 | 1156 | 4.8 | |
PCA-2 | 0.84 | 0.82 | 0.80 | 0.78 | 1061 | 4.2 | ||
PCA-8 | 0.90 | 0.91 | 0.90 | 0.86 | 869 | 3.8 | ||
PCA-10 | 0.90 | 0.92 | 0.89 | 0.87 | 1006 | 4.54 | ||
LDA-2 | 0.88 | 0.87 | 0.81 | 0.78 | 998 | 3.65 | ||
LDA-4 | 0.88 | 0.89 | 0.83 | 0.80 | 1036 | 3.8 | ||
Proposed (AE-2) | 0.74 | 0.79 | 0.72 | 0.70 | 790 | 3 | ||
Proposed (AE-3) | 0.86 | 0.86 | 0.86 | 0.82 | 840 | 2.54 | ||
Proposed (AE-8) | 0.90 | 0.90 | 0.89 | 0.85 | 774 | 3.22 |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | ||||
---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1. | Train. | Test. | ||
CNN-LSTM [14] | Without | 0.91 | 0.93 | 0.92 | 0.89 | - | - |
DT | Without | 0.90 | 0.95 | 0.90 | 0.87 | 19.2 | 0.1 |
PCA-8 | 0.90 | 0.91 | 0.90 | 0.87 | 8.4 | 0.03 | |
LDA-4 | 0.90 | 0.94 | 0.90 | 0.87 | 3.21 | 0.05 | |
Proposed (AE-8) | 0.90 | 0.92 | 0.90 | 0.87 | 8.8 | 0.03 | |
XTree | Without | 0.90 | 0.94 | 0.90 | 0.87 | 0.75 | 0.11 |
PCA-8 | 0.90 | 0.91 | 0.90 | 0.87 | 0.46 | 0.04 | |
LDA-4 | 0.90 | 0.93 | 0.90 | 0.87 | 0.38 | 0.036 | |
Proposed (AE-8) | 0.90 | 0.92 | 0.90 | 0.87 | 0.37 | 0.04 | |
RF | Without | 0.90 | 0.94 | 0.90 | 0.87 | 198.4 | 2.96 |
PCA-8 | 0.90 | 0.92 | 0.90 | 0.87 | 175.1 | 3.41 | |
LDA-4 | 0.90 | 0.94 | 0.90 | 0.87 | 107.4 | 3.03 | |
Proposed (AE-8) | 0.90 | 0.94 | 0.90 | 0.87 | 182.8 | 3.84 | |
DNN2 | Without | 0.90 | 0.94 | 0.90 | 0.87 | 937 | 2.59 |
PCA-8 | 0.90 | 0.92 | 0.89 | 0.86 | 771 | 1.95 | |
LDA-4 | 0.89 | 0.92 | 0.86 | 0.83 | 559 | 1.48 | |
Proposed (AE-8) | 0.90 | 0.92 | 0.90 | 0.87 | 551 | 1.35 |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | ||||
---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1. | Training | Testing | ||
CNN-LSTM [14] | Without | 0.89 | 0.93 | 0.90 | 0.86 | - | - |
DT | Without | 0.88 | 0.95 | 0.90 | 0.87 | 44.2 | 0.2 |
PCA-8 | 0.88 | 0.92 | 0.90 | 0.86 | 14 | 0.07 | |
LDA-4 | 0.88 | 0.92 | 0.90 | 0.87 | 5.6 | 0.04 | |
Proposed (AE-8) | 0.87 | 0.91 | 0.89 | 0.86 | 13.4 | 0.07 | |
Proposed (AE-20) | 0.88 | 0.91 | 0.90 | 0.86 | 20.7 | 0.031 | |
Xtree | Without | 0.88 | 0.94 | 0.90 | 0.87 | 1.1 | 0.18 |
PCA-8 | 0.88 | 0.91 | 0.90 | 0.86 | 0.46 | 0.05 | |
LDA-4 | 0.88 | 0.93 | 0.90 | 0.87 | 0.35 | 0.036 | |
Proposed (AE-8) | 0.87 | 0.92 | 0.89 | 0.86 | 0.41 | 0.045 | |
Proposed (AE-20) | 0.88 | 0.92 | 0.90 | 0.87 | 0.61 | 0.031 | |
RF | Without | 0.88 | 0.94 | 0.90 | 0.87 | 226 | 3.1 |
PCA-8 | 0.88 | 0.92 | 0.90 | 0.87 | 184 | 3.7 | |
LDA-4 | 0.88 | 0.94 | 0.90 | 0.87 | 177 | 4.6 | |
Proposed (AE-8) | 0.88 | 0.92 | 0.90 | 0.87 | 156 | 3.78 | |
Proposed (AE-20) | 0.88 | 0.92 | 0.90 | 0.87 | 253 | 3.2 | |
DNN2 (Epoch = 100) | Without | 0.88 | 0.94 | 0.90 | 0.86 | 1202 | 5.4 |
PCA-8 | 0.87 | 0.91 | 0.87 | 0.85 | 764 | 2.87 | |
LDA-4 | 0.85 | 0.89 | 0.80 | 0.75 | 802 | 2.32 | |
Proposed (AE-8) | 0.87 | 0.91 | 0.89 | 0.86 | 698 | 1.93 | |
Proposed (AE-20) | 0.87 | 0.91 | 0.89 | 0.86 | 602 | 1.52 |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | ||||
---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1. | Training | Testing | ||
DT | Without | 0.70 | 0.88 | 0.80 | 0.74 | 5.4 | 0.045 |
PCA-8 | 0.70 | 0.82 | 0.80 | 0.74 | 1.28 | 0.008 | |
LDA-4 | 0.70 | 0.87 | 0.80 | 0.74 | 0.5 | 0.007 | |
Proposed (AE-8) | 0.70 | 0.84 | 0.80 | 0.74 | 1.3 | 0.008 | |
Proposed (AE-20) | 0.70 | 0.84 | 0.80 | 0.74 | 2.5 | 0.008 | |
XTree | Without | 0.70 | 0.85 | 0.80 | 0.74 | 0.25 | 0.044 |
PCA-8 | 0.70 | 0.87 | 0.80 | 0.74 | 0.10 | 0.01 | |
LDA-4 | 0.70 | 0.87 | 0.80 | 0.74 | 0.092 | 0.01 | |
Proposed (AE-8) | 0.70 | 0.87 | 0.80 | 0.74 | 0.092 | 0.01 | |
Proposed (AE-20) | 0.70 | 0.86 | 0.80 | 0.74 | 0.14 | 0.01 | |
RF | Without | 0.70 | 0.90 | 0.80 | 0.74 | 35.11 | 0.83 |
PCA-8 | 0.70 | 0.88 | 0.80 | 0.74 | 24.2 | 0.8 | |
LDA-4 | 0.70 | 0.90 | 0.80 | 0.74 | 21.7 | 0.76 | |
Proposed (AE-8) | 0.70 | 0.88 | 0.80 | 0.74 | 23.53 | 0.72 | |
Proposed (AE-20) | 0.70 | 0.88 | 0.80 | 0.74 | 36 | 0.68 | |
DNN2 Epoch = 100 | Without | 0.70 | 0.90 | 0.80 | 0.74 | 287 | 1.15 |
PCA-8 | 0.70 | 0.82 | 0.80 | 0.74 | 202 | 0.71 | |
LDA-4 | 0.67 | 0.85 | 0.75 | 0.69 | 234 | 0.52 | |
Proposed (AE-8) | 0.69 | 0.88 | 0.80 | 0.74 | 207 | 0.7 | |
Proposed (AE-20) | 0.70 | 0.87 | 0.80 | 0.74 | 182 | 0.78 | |
CNN (epoch = 50, batch size = 512) | Without | 0.70 | 0.89 | 0.80 | 0.74 | 2247 | 14 |
PCA-8 | 0.70 | 0.85 | 0.80 | 0.74 | 436 | 3.5 | |
LDA-4 | 0.67 | 0.84 | 0.74 | 0.69 | 354 | 5.1 | |
Proposed (AE-8) | 0.69 | 0.85 | 0.79 | 0.73 | 432 | 3.2 | |
Proposed (AE-20) | 0.70 | 0.89 | 0.80 | 0.74 | 877 | 4.5 | |
CNN-LSTM (epoch = 50, batch size = 512) | Without | 0.70 | 0.88 | 0.80 | 0.74 | 13870 | 185 |
PCA-8 | 0.70 | 0.86 | 0.80 | 0.74 | 1523 | 11.53 | |
LDA-4 | 0.66 | 0.84 | 0.74 | 0.68 | 926 | 5.62 | |
Proposed (AE-8) | 0.69 | 0.86 | 0.80 | 0.73 | 1451 | 10.86 | |
Proposed (AE-20) | 0.69 | 0.90 | 0.80 | 0.74 | 3905 | 16.25 |
Classifier | Feature Reduction Technique | Metrics | Time (Second) | ||||
---|---|---|---|---|---|---|---|
Acc. | Prec. | Rec. | F1. | Training | Testing | ||
DT | Without | 1.00 | 0.99 | 0.99 | 0.99 | 0.55 | 0.008 |
PCA-8 | 0.99 | 0.96 | 0.96 | 0.96 | 0.32 | 0.002 | |
LDA-4 | 0.99 | 0.96 | 0.97 | 0.97 | 0.28 | 0.004 | |
Proposed (AE-8) | 0.99 | 0.96 | 0.96 | 0.96 | 0.48 | 0.002 | |
XTree | Without | 1.00 | 0.98 | 0.98 | 0.98 | 0.041 | 0.008 |
PCA-8 | 0.99 | 0.96 | 0.96 | 0.96 | 0.02 | 0.003 | |
LDA-4 | 0.99 | 0.96 | 0.97 | 0.96 | 0.018 | 0.003 | |
Proposed (AE-8) | 0.99 | 0.96 | 0.96 | 0.96 | 0.019 | 0.002 | |
RF | Without | 1.00 | 0.99 | 0.99 | 0.99 | 5.4 | 0.2 |
PCA-8 | 0.99 | 0.97 | 0.97 | 0.97 | 9.57 | 0.27 | |
LDA-4 | 0.99 | 0.97 | 0.97 | 0.97 | 17.24 | 0.4 | |
Proposed (AE-8) | 0.99 | 0.97 | 0.97 | 0.97 | 9.2 | 0.18 | |
DNN2 | Without | 0.99 | 0.98 | 0.97 | 0.97 | 168 | 0.76 |
PCA-8 | 0.98 | 0.94 | 0.92 | 0.93 | 158 | 0.7 | |
LDA-4 | 0.99 | 0.95 | 0.93 | 0.94 | 171 | 0.7 | |
Proposed (AE-8) | 0.99 | 0.94 | 0.93 | 0.94 | 74 | 0.2 | |
CNN | Without | 0.99 | 0.97 | 0.98 | 0.97 | 1454 | 8 |
PCA-8 | 0.98 | 0.94 | 0.92 | 0.93 | 201 | 2.36 | |
LDA-4 | 0.99 | 0.94 | 0.95 | 0.95 | 138 | 1.8 | |
Proposed (AE-8) | 0.98 | 0.93 | 0.94 | 0.94 | 143 | 0.9 | |
CNN-LSTM | Without | 0.99 | 0.96 | 0.96 | 0.96 | 6774 | 32.63 |
PCA-8 | 0.98 | 0.94 | 0.93 | 0.93 | 602 | 5.90 | |
LDA-4 | 0.99 | 0.95 | 0.94 | 0.94 | 347 | 4.2 | |
Proposed (AE-8) | 0.98 | 0.93 | 0.93 | 0.93 | 446 | 2.7 | |
SVM | Without | 0.98 | 0.94 | 0.91 | 0.92 | 12.8 | 17 |
PCA-8 | 0.98 | 0.90 | 0.88 | 0.89 | 7.45 | 16.4 | |
LDA-4 | 0.98 | 0.93 | 0.90 | 0.91 | 3.3 | 11.3 | |
Proposed (AE-8) | 0.97 | 0.93 | 0.88 | 0.89 | 10.8 | 16.2 | |
KNN | Without | 0.99 | 0.96 | 0.96 | 0.96 | 0.02 | 46 |
PCA-8 | 0.99 | 0.96 | 0.96 | 0.96 | 0.24 | 0.85 | |
LDA-4 | 0.99 | 0.97 | 0.97 | 0.97 | 0.13 | 0.90 | |
Proposed (AE-8) | 0.99 | 0.97 | 0.96 | 0.96 | 0.18 | 0.74 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Alaghbari, K.A.; Lim, H.-S.; Saad, M.H.M.; Yong, Y.S. Deep Autoencoder-Based Integrated Model for Anomaly Detection and Efficient Feature Extraction in IoT Networks. IoT 2023, 4, 345-365. https://doi.org/10.3390/iot4030016
Alaghbari KA, Lim H-S, Saad MHM, Yong YS. Deep Autoencoder-Based Integrated Model for Anomaly Detection and Efficient Feature Extraction in IoT Networks. IoT. 2023; 4(3):345-365. https://doi.org/10.3390/iot4030016
Chicago/Turabian StyleAlaghbari, Khaled A., Heng-Siong Lim, Mohamad Hanif Md Saad, and Yik Seng Yong. 2023. "Deep Autoencoder-Based Integrated Model for Anomaly Detection and Efficient Feature Extraction in IoT Networks" IoT 4, no. 3: 345-365. https://doi.org/10.3390/iot4030016
APA StyleAlaghbari, K. A., Lim, H. -S., Saad, M. H. M., & Yong, Y. S. (2023). Deep Autoencoder-Based Integrated Model for Anomaly Detection and Efficient Feature Extraction in IoT Networks. IoT, 4(3), 345-365. https://doi.org/10.3390/iot4030016