تشخیص نفوذ با استفاده از یادگیری عمیق در شبکه های حسگر بی سیم بدنی
محورهای موضوعی : پردازش چند رسانه ای، سیستمهای ارتباطی، سیستمهای هوشمندالهام حاجیان 1 * , نوید اسدی 2
1 - استادیار گروه مهندسی کامپیوتر، دانشکده فنی و مهندسی، دانشگاه بجنورد، بجنورد، ایران
2 - دانشجوی کارشناسی ارشد، دانشکده فنی و مهندسی، دانشگاه بجنورد، بجنورد، ایران
کلید واژه: یادگیری عمیق, تشخیص نفوذ, حملات سایبری, شبکة حسگر بی¬سیم بدنی, دقت, صحت,
چکیده مقاله :
فراگیری استفاده از فناوری اطلاعات و شبکههای کامپیوتری باعث بروز حملات متعددی شده است که هدف عمده این حملات، به خطر انداختن امنیت شبکه و پایگاه دادههاست. شبکههای حسگر بیسیم بدنی که نوعی از فناوری جدید برای پیگیری وضعیت بیمار میباشد، نیز از این جهت مستثنی نیست. این شبکهها بدلیل کاربردهای حساس پزشکی از اهمیت ویژهای برخوردار هستند. هر گونه حمله و نفوذ به این شبکهها خسارات جبران ناپذیری را برای بیمار به همراه دارد. به همین منظور میتوان از سیستمهای تشخیص نفوذ به عنوان مکمل امنیت در ارتباطات شبکهای حسگر بدن، بهره برد. یکی از روشهای تأمین امنیّت استفاده از سامانه های
تشخیص نفوذ به عنوان یک دفاع خط دوم است. از این رو که تکنیکهای متداول مخرب به طور تصادفی در حال افزایش است، روشهای سنتی قادر به جوابگویی در برابر حملات نیست. از وظائف سیستمهای تشخیص نفوذ در شبکههای حسگر بدن، علاوه بر شناسایی حملات، یادگیری الگوی رفتاری حمله در سیستم میباشد. یکی از موضوعات پرچالش در این سیستمها، دقت میباشد. روشهای جدیدی به منظور بهبود نرخ تشخیص درست و به حداقل رساندن نرخ تشخیص اشتباه ابداع شدهاند که با بهبود هر چه بیشتر دقت، کارایی سیستم افزایش مییابد. در این پژوهش، افزایش دقت که با استفاده از شبکه های پرسپترون چند لایه که یکی از روشهای یادگیری عمیق میباشد، انجام شده است. با افزایش تعداد لایههای پنهان، یادگیری کارآمدتر در این شبکهها انجام میشود. مجموعه داده WBAN rssi که از Kaggle گرفته شده است، در 3 کلاس مختلف نرمال، حملة نوع 1 و حملة نوع 2 بررسی میشود. سپس نمودار الگوریتم پیشنهادی برای دقت، درستی و فراخوانی و امتیاز F1 با استفاده از مجموعة داده به تنهایی و در 3 کلاس مختلف ترسیم شده است که نشان از دقت 0.72 میدهد.در هر کلاس و نوع حمله، میزان دقت و درستی و امتیاز F1 بررسی می شود و در انتها برای هر کدام جدول گزارش نشان داده می شود.
Abstract
The widespread use of information technology and computer networks has led to the emergence of numerous attacks, the main purpose of which is to compromise the security of networks and databases. Wireless body sensor networks, which are a new technology for tracking patient status, are no exception. These networks are of particular importance due to their sensitive medical applications. Any attack and intrusion into these networks will cause irreparable damage to the patient. For this purpose, intrusion detection systems can be used as a security supplement in body sensor network communications. Since common destructive techniques are increasing randomly, traditional methods are unable to respond to attacks. In addition to identifying attacks, the tasks of intrusion detection systems in body sensor networks include learning the behavioral pattern of attacks in the system. One of the challenging issues in these systems is accuracy. New methods have been developed to improve the correct detection rate and minimize the false detection rate, which increases the efficiency of the system by improving the accuracy. In this study, the accuracy increase is done using multilayer perceptron networks, which is one of the deep learning methods. By increasing the number of hidden layers, more efficient learning is done in these networks. The WBAN RSSI dataset, which is taken from Kaggle, is examined in 3 different classes: normal, type 1 attack, and type 2 attack. Then, the proposed algorithm is plotted for precision, accuracy, recall, and F1 score using the dataset alone and in 3 different classes, which shows an accuracy of 0.72.
Introduction: This paper examines intrusion detection in wireless body area networks. A wireless body area network is a network that sends a lot of clinical data remotely to a server for further processing and then to the doctor for further review. Intrusion due to data diversion in a medical system can have dangerous consequences. Therefore, a mechanism is needed to detect and prevent it.
Method: Intrusion detection in this research has been done using deep learning. By increasing the number of hidden layers in the neural network, data processing and learning are increased and they give more accurate results. Each layer has an activation function. The output layer has 3 classes, which are related to the normal class and types of attacks. The most likely class related to these classes is the prediction of this method for the input data, which attacks this data is most exposed to.
Results: Given the data set considered for testing, there are 3 different classes with different precisions. Class 0(normal data) has the highest precision. Class 0 also has the highest F1 score, indicating good performance in detecting normal data. Class 1 has lower recall, meaning it has difficulty identifying some examples of this class. Class 2 has good recall and lower precisions, indicating some false positives in this class.
Discussion: Other improvements were also made to the model in this regard. These improvements include Hyperparameter tuning: can be tested with different learning rates, batch sizes, and optimal number of epochs. Class balance: handling unbalanced datasets can improve recall of minority classes. Advanced architectures: can be tested and researched using recurrent neural networks or convolutional neural networks to improve model performance.
[1] K. Heshan, K. Ibrahim, T. Zahir, Z. Alber, “Distributed anomaly detection for industrial wireless sensor networks based on fuzzy data modelling”, parallel and distributed computing, vol.73,no. 6, pp. 790-806, 2013.
[2] S. Mambwe Kasongo, “A deep learning technique for intrusion detection system using a Recurrent Neural Networks based framework” Computer Communications,Volume 199, Pages 113-125,2023.
[3]M. Usman, M. R. Asghar, I. S. Ansari, and M.Qaraqe, "Security in Wireless Body Area Networks: From In-Body to Off-Body Communications," IEEE Access, vol. 6, pp.58064-58074, 2018.
[4]O. Salem, A. Serhrouchni, A. Mehaoua, and R.Boutaba, "Event Detection in Wireless Body Area Networks using Kalman Filter and Power Divergence," IEEE Transactions on Network and Service Management, 2018.
[5] N. K. Jha, A. Raghunathan, and M. Zhang,"Securing medical devices through wireless monitoring and anomaly detection," ed: Google Patents, 2018.
[6] A. Rani, A. Viswasa and E. Baburaj, "Secure and intelligent architecture for cloud-based healthcare applications in wireless body sensor networks", Int. J. Biomed. Eng. Technol., vol. 29, no. 2, pp. 186-199, 2019.
[7] A. Alabdulatif, I. Khalil, A. R. M. Forkan and M. Atiquzzaman, "Real-time secure health surveillance for smarter health communities", IEEE Commun. Mag., vol. 57, no. 1, pp. 122-129, Jan. 2019.
[8] S. Carta, S.; Podda, A.S.; Reforgiato Recupero, D.R.; Saia, R. “A Local Feature Engineering Strategy to Improve Network Anomaly Detection”. Future Internet vol. 12, no.177.2020.
[9] B. A. Alzahrani, S. A. Chaudhry, A. Barnawi, A. Al-Barakati and M. H. Alsharif, "A privacy preserving authentication scheme for roaming in IoT-based wireless mobile networks", Symmetry, vol. 12, no. 2, pp. 287, Feb. 2020.
[10] V. Odelu, S. Saha, R. Prasath, L. Sadineni, M. Conti and M. Jo, "Efficient privacy preserving device authentication in WBANs for industrial e-health applications", Comput. Secur., vol. 83, pp. 300-312, Jun. 2019.
[11] G. Thangarasu, K. R. Alla and K. N. Kannan, "Denial of Service Mitigation in Wireless Body Area Network Using Deep Learning," 2024 IEEE 6th Symposium on Computers & Informatics (ISCI), Kuala Lumpur, Malaysia, 2024, pp. 328-332, doi: 10.1109/ISCI62787.2024.10668017.
[12] H. Shaker, M. Nariman, J. Q. Ahmed, D. Radhi. (2023). “A Deep learning approach for trust-untrust nodes classification problem in WBAN. Periodicals of Engineering and Natural Sciences (PEN)”, vol.1, no.3. doi: 10.21533/pen.v11i3.3579.
[13] B. Liya, S. Krishnamoorthy, R. Arun, S. “An enhanced deep learning-based disease detection model in wireless body area network with energy efficient routing protocol”. Wireless Netw , vol.30, pp. 2961–2986, 2024 https://doi.org/10.1007/s11276-024-03717-1.
[14] R. Arthi, S. Krishnaveni and S. Zeadally, "An Intelligent SDN-IoT Enabled Intrusion Detection System for Healthcare Systems Using a Hybrid Deep Learning and Machine Learning Approach," in China Communications, vol. 21, no. 10, pp. 1-21, Oct. 2024, doi: 10.23919/JCC.ja.2022-0681.
[15] C. Iwend, JH.Anajemba, C. Biamba, D. Ngabo. “Security of Things Intrusion Detection System for Smart Healthcare”, Electronics. Vol.10, no.12 2021. https://doi.org/10.3390/electronics10121375.
[16] A. Singh, K.Chatterjee, and S.Chandra Satapathy “TrIDS: an intelligent behavioural trust based IDS for smart healthcare system”. Cluster Computing, vol.26, no.2, pp.903–925, 2022. https://doi.org/10.1007/s10586-022-03614-2.
[17]M.R.Erfaneh noroozi,”Presenting A Hybrid Method of Deep Neural Networks to Prevent Intrusion in Computer Networks”, Intelligent Multimedia Processing and Communication Systems(IMPCS),no.4, p.65, 2023.
[18]A.K..M.K.SeyedReza Kamel,”An Efficient and Light Weight Intrusion Detection for IoT Based on Fog and Cloud Using KNN Classification”, Intelligent Multimedia Processing and Communication Systems(IMPCS),no.2, p.64, 2023.
[19]M.Yaghoubi, K. Ahmed, Y. Miao. “Wireless Body Area Network (WBAN): A Survey on Architecture, Technologies, Energy Consumption, and Security Challenges”. Journal of Sensor and Actuator Networks.;vol. 11, no. 4 , 2022. https://doi.org/10.3390/jsan11040067.
[20] P. Pijush, K. Dutta, N. Anand, P.Gaurav, D. Hemanth B. Valentina Emilia, “WBAN: Driving e-healthcare Beyond Telemedicine to Remote Health Monitoring: Architecture and Protocols,Telemedicine”, Technologies,Academic Press, Volume 89,NO 119, 2019.
[21] A. Vidyadhar Jinnappa, D. Vijaypal Singh, P. Anubha, k. Sunil, R. Imad , “Internet of Things in healthcare: A survey on protocol standards”, enabling technologies, WBAN architectures and open issues, Physical Communication,Vol. 60, No 102103,2023.
[22] S. Karchowdhury and M. Sen, "Survey on attacks on wireless body area network," International Journal of Computational Intelligence & IoT, Forthcoming, 2019.