Related Experiment Videos
Reservoir computing for network intrusion classification
Khorshed Alam1,2, Mahbubul Haq Bhuiyan1, Mohammad Ashraful Hoque1
1Department of Computer Science and Engineering, Southeast University, Dhaka, Bangladesh.
Plos One
|August 14, 2026
Summary
Reservoir computing models like Echo State Networks (ESNs) and Liquid State Machines (LSMs) offer efficient, lightweight network intrusion detection for IoT. These models achieve performance comparable to deep learning methods with reduced computational demands.
Area of Science:
- Computer Science
- Cybersecurity
- Artificial Intelligence
Background:
- Network Intrusion Detection Systems (NIDS) are crucial for IoT security.
- Resource constraints in IoT necessitate lightweight detection solutions.
- Conventional deep learning models (CNN, LSTM) are computationally intensive for IoT.
Purpose of the Study:
- To investigate reservoir computing models (ESNs, LSMs) as lightweight NIDS alternatives for IoT.
- To develop custom ESN and LSM architectures for efficient temporal feature learning and attack classification.
- To evaluate the performance and resource efficiency of reservoir computing models against benchmarks.
Main Methods:
- Utilized Echo State Networks (ESNs) and Liquid State Machines (LSMs) for temporal feature learning and attack classification.
- Designed custom ESN and LSM architectures with low computational complexity.
- Employed the NF-ToN-IoT dataset for evaluating NIDS performance on diverse attack categories.
Main Results:
- Proposed ESN and LSM models demonstrated performance comparable to CNN and LSTM benchmarks.
- Achieved substantial reductions in resource usage compared to conventional deep learning models.
- Validated the effectiveness of reservoir computing for real-time intrusion detection in IoT networks.
Conclusions:
- Reservoir computing (ESN, LSM) presents a viable, efficient, and scalable alternative for lightweight NIDS in IoT.
- These models are suitable for real-time deployment due to their low computational overhead.
- Further exploration of reservoir computing in NIDS applications is warranted.
