
		<paper>
			<loc>https://jjcit.org/paper/255</loc>
			<title>INTRUSION DETECTION SYSTEM FOR INTERNET OF MEDICAL THINGS USING GRU WITH ATTENTION MECHANISM-BASED HYBRID DEEP LEARNING TECHNIQUE</title>
			<doi>10.5455/jjcit.71-1725609265</doi>
			<authors>Naveen Saran,Nishtha Kesswani</authors>
			<keywords>Internet  of medical things,Intrusion-detection system,Principal-component analysis,Deep learning,Gated recurrent unit,Attention mechanism,K-fold cross validation</keywords>
			<citation>18</citation>
			<views>4814</views>
			<downloads>1015</downloads>
			<received_date>6-Sep.-2024</received_date>
			<revised_date>  8-Nov.-2024</revised_date>
			<accepted_date>  21-Nov.-2024</accepted_date>
			<abstract>The  proliferation  of  Internet  of  Things  devices  in  healthcare,  specifically  the  Internet  of  Medical  Things,  has 
revolutionized  patient  care  and  health-monitoring  systems.  Integrating  these interconnected  medical  devices 
introduces  unprecedented  security  challenges,  necessitating  robust  Intrusion  Detection  Systems  (IDSs)  to 
safeguard  patient  data  and  healthcare  infrastructure.  To  protect  the  IoMT  devices  from  numerous  malicious 
attacks,  researchers  have  developed  numerous  Intrusion  Detection  Systems,  but  the  development  of  an  effective 
and real-time IDS remains a challenge. Our proposed IDS addresses this gap and surpasses state-of-the-art IDS 
techniques  for  IoMT  networks.  In  this  research  paper,  we  have  proposed  a  novel  IDS  approach  for  IoMT, 
leveraging  a  Hybrid  Deep  Learning  technique  to  enhance  detection  accuracy  and  efficiency.  By  combining  the 
strengths  of  Gated  Recurrent  Unit  (GRU)  and  Attention  Mechanism,  the  proposed  IDS  achieves  superior 
performance  in  detecting  anomalous  activities  in  medical  networks.  We  evaluated  the  proposed  IDS  model  on 
two  publicly  available  benchmark  intrusion  datasets  and  achieved 99.99  %  accuracy  on  the  ICU  Healthcare 
Dataset  and  98.94  %  on  the  NF-TON-IoT  Dataset.  Precision,  Recall,  F1-score  metrics and ROC-AUC  for  the 
proposed model are promising. We also added Noise to the features to show how effectively the model performed 
in noisy environments. Moreover, we used the K-Fold Cross Validation Technique to cross-va�O�L�Ga�W�H �W�K�H �Po�G�H�O’�V 
performance  on  both  datasets,  ensuring  the  reliability  and  applicability  of  the  suggested  IDS  model  for  IoMT 
networks.</abstract>
		</paper>


