Penyulingan Pengetahuan Multi Tahap dengan Konsep Sederhana Baru pada IDS untuk Penerapan Edge Ringan
Abstrak
Penerapan deep learning dalam Intrusion Detection System (IDS) telah menunjukkan kemajuan yang signifikan dalam penelitian-penelitian sebelumnya. Namun, penggunaan edge computing dalam IDS memiliki dua keuntungan, mengurangi latensi dan meningkatkan privasi. Penelitian ini bertujuan untuk mengembangkan Intrusion Detection System (IDS) berbasis edge computing yang memanfaatkan deep neural network (DNN) sebagai model pembelajaran mesin dan arsitektur Multi-Stage Knowledge Distillation (MSKD) yang inovatif dan disederhanakan sebagai metodologi pelatihan. Data set yang digunakan berasal dari Kaggle, yaitu UNSW-NB15 benchmark dataset. Langkah berikutnya adalah menyiapkan data dan melatih proses MSKD. Pada penelitian ini, penulis membandingkan pendekatan MSKD dengan pendekatan Knowledge Distillation (KD) dan mengevaluasi kedua model tersebut. Metode MSKD baru menggunakan tiga model DNN: Teacher Model, Student Model 1, dan Student Model 2. Untuk melihat seberapa baik MSKD bekerja, dilakukan analisis terhadap akurasi, penurunan akurasi, ukuran model, dan tingkat kompresi (CR). Akurasi Teacher Model adalah 0,9955, Student Model 1 mendapatkan akurasi sebesar 0,9897, dan Student Model 2 mendapatkan akurasi sebesar 0,9857. Jika diperhatikan, akurasi Student Model 1 turun sebesar 0,58% dibandingkan dengan Teacher Model, dan akurasi Student Model 2 turun sebesar 0,99% dibandingkan dengan Teacher Model dan 0,40% dibandingkan dengan Student Model 1. Rasio kompresi terbaik untuk Student Model 1 adalah 101,9 kali lebih baik daripada Teacher Model dan 3,0 kali lebih baik daripada Student Model 1. Penelitian ini juga menganalisis bahwa penggunaan soft label menghasilkan rasio kompresi (CR) yang 1,8 kali lebih tinggi antara Student Model 2 dan Student Model 2 yang dilatih menggunakan hard label.
Kata kunci: multi-stage knowledge distillation, intrusion detection system, UNSW-NB15 dataset, edge computing.
Referensi
R. Devi, A. Kumar, V. Kumar, A. Saini, A. Kumari, and V. Kumar, “A Review Paper on IDS in Edge Computing or EoT,” in 2022 International Conference on 4th Industrial Revolution Based Technology and Practices, ICFIRTP 2022, Institute of Electrical and Electronics Engineers Inc., 2022, pp. 30–35. doi: 10.1109/ICFIRTP56122.2022.10059442.
W. AlAqqad, M. Nijim, U. Onyeakazi, and H. Albataineh, “Cyber Edge: Mitigating Cyber-Attacks in Edge Computing Using Intrusion Detection System,” 2024, pp. 292–305. doi: 10.1007/978-3-031-56950-0_25.
T.-M. Hoang, T.-A. Pham, V.-V. Do, V.-N. Nguyen, and M.-H. Nguyen, “A Lightweight DNN-based IDS for Detecting IoT Cyberattacks in Edge Computing,” in 2022 International Conference on Advanced Technologies for Communications (ATC), IEEE, Oct. 2022, pp. 136–140. doi: 10.1109/ATC55345.2022.9943049.
R. K. Vigneswaran, R. Vinayakumar, K. P. Soman, and P. Poornachandran, “Evaluating Shallow and Deep Neural Networks for Network Intrusion Detection Systems in Cyber Security,” in 2018 9th International Conference on Computing, Communication and Networking Technologies (ICCCNT), IEEE, Jul. 2018, pp. 1–6. doi: 10.1109/ICCCNT.2018.8494096.
N. N. Tai, H. Do Hoang, P. T. Duy, and V. H. Pham, “An interpretable approach for trustworthy intrusion detection systems against evasion samples,” CTU Journal of Innovation and Sustainable Development, vol. 15, no. Special Issue (ISDS), pp. 12–19, Oct. 2023, doi: 10.22144/ctujoisd.2023.030.
R. Mishra and H. P. Gupta, “Designing and Training of Lightweight Neural Networks on Edge Devices Using Early Halting in Knowledge Distillation,” IEEE Trans Mob Comput, vol. 23, no. 5, pp. 4665–4677, May 2024, doi: 10.1109/TMC.2023.3297026.
C. Xu et al., “Multiple-Stage Knowledge Distillation,” Applied Sciences, vol. 12, no. 19, p. 9453, Sep. 2022, doi: 10.3390/app12199453.
Z. Hao, J. Guo, K. Han, H. Hu, C. Xu, and Y. Wang, “VanillaKD: Revisit the Power of Vanilla Knowledge Distillation from Small Scale to Large Scale,” May 2023, [Online]. Available: http://arxiv.org/abs/2305.15781
X. Zeng et al., “DSP-KD: Dual-Stage Progressive Knowledge Distillation for Skin Disease Classification,” Bioengineering, vol. 11, no. 1, p. 70, Jan. 2024, doi: 10.3390/bioengineering11010070.
W. Zhou, C. Song, K. Song, N. Wen, X. Sun, and P. Gao, “Surface Defect Detection System for Carrot Combine Harvest Based on Multi-Stage Knowledge Distillation,” Foods, vol. 12, no. 4, p. 793, Feb. 2023, doi: 10.3390/foods12040793.
Y. Du, J. Niu, Y. Wang, and X. Jin, “Multi-stage knowledge distillation for sequential recommendation with interest knowledge,” Inf Sci (N Y), vol. 654, p. 119841, Jan. 2024, doi: 10.1016/j.ins.2023.119841.
Z. Shi, Y. Li, and C. Liu, “Knowledge Distillation-enabled Multi-stage Incremental Learning for Online Process Monitoring in Advanced Manufacturing,” in 2022 IEEE International Conference on Data Mining Workshops (ICDMW), IEEE, Nov. 2022, pp. 860–867. doi: 10.1109/ICDMW58026.2022.00154.
T. Yuliswar, I. Elfitri, and O. W purbo, “Optimization of Intrusion Detection System with Machine Learning for Detecting Distributed Attacks on Server,” INOVTEK Polbeng - Seri Informatika, vol. 10, no. 1, pp. 367–376, Feb. 2025, doi: 10.35314/vem9da98.
P. Illy, G. Kaddoum, P. F. de Araujo-Filho, K. Kaur, and S. Garg, “Retracted: A Hybrid Multistage DNN-Based Collaborative IDPS for High-Risk Smart Factory Networks,” IEEE Transactions on Network and Service Management, vol. 19, no. 4, pp. 4273–4283, Dec. 2022, doi: 10.1109/TNSM.2022.3202801.
A. G. Putrada, N. Alamsyah, M. N. Fauzan, and I. D. Oktaviani, “Pearson Correlation for Efficient Network Anomaly Detection with Quantization on the UNSW-NB15 Dataset,” in 2024 International Conference on ICT for Smart Society (ICISS), IEEE, Sep. 2024, pp. 1–6. doi: 10.1109/ICISS62896.2024.10751550.
S. More, M. Idrissi, H. Mahmoud, and A. T. Asyhari, “Enhanced Intrusion Detection Systems Performance with UNSW-NB15 Data Analysis,” Algorithms, vol. 17, no. 2, p. 64, Feb. 2024, doi: 10.3390/a17020064.
A. Mageed, A. EL-SAYED, M. A. Gaber, and T. Abdelhamid, “A Comprehensive Approach to Skin Cancer Detection: Mixed Dataset, Optimizer Enhancement, and Web Integration,” Menoufia Journal of Electronic Engineering Research, vol. 34, no. 1, pp. 26–39, Jan. 2025, doi: 10.21608/mjeer.2025.316087.1093.
S. Moualla, K. Khorzom, and A. Jafar, “Improving the Performance of Machine Learning‐Based Network Intrusion Detection Systems on the UNSW‐NB15 Dataset,” Comput Intell Neurosci, vol. 2021, no. 1, Jan. 2021, doi: 10.1155/2021/5557577.
A. Gupta and R. Simon, “Enhancing Security in Cloud Computing With Anomaly Detection Using Random Forest,” in 2024 11th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO), IEEE, Mar. 2024, pp. 1–6. doi: 10.1109/ICRITO61523.2024.10522227.
I. Tavares et al., “Comparison of PV Power Generation Forecasting in a Residential Building using ANN and DNN,” IFAC-PapersOnLine, vol. 55, no. 9, pp. 291–296, 2022, doi: 10.1016/j.ifacol.2022.07.051.
M. N. Adiputra, A. G. Putrada, and I. D. Oktaviani, “DistilQ-NILM: A Hybrid Quantization-Knowledge Distillation LSTM Model for Edge-Based Load Monitoring,” in 2024 International Conference on Intelligent Cybernetics Technology & Applications (ICICyTA), IEEE, Dec. 2024, pp. 284–289. doi: 10.1109/ICICYTA64807.2024.10913187.
B. Fan, H. Ma, Y. Liu, X. Yuan, and W. Ke, “KDTM: Multi-Stage Knowledge Distillation Transfer Model for Long-Tailed DGA Detection,” Mathematics, vol. 12, no. 5, p. 626, Feb. 2024, doi: 10.3390/math12050626.
A. Gautama, P. #1, N. Alamsyah, M. N. Fauzan, S. Prabowo, and I. D. Oktaviani, “QUIDS: A Novel Edge-Based Botnet Detection with Quantization for IoT Device Pairing”, doi: 10.34818/indojc.2023.8.3.878.
A. G. Putrada and D. A. S. Ilhami, “G-Mean for Optimum Threshold in Anomaly Detection with Autoencoder: Cyber Security on The RT-IOT2022 Dataset,” in 2024 IEEE 22nd Student Conference on Research and Development (SCOReD), IEEE, Dec. 2024, pp. 433–438. doi: 10.1109/SCOReD64708.2024.10872727.
S. Mandal, X. Lin, and R. Srikant, “A Theoretical Analysis of Soft-Label vs Hard-Label Training in Neural Networks,” Dec. 2024, [Online]. Available: http://arxiv.org/abs/2412.09579
