The rapid adoption of the Internet of Medical Things (IoMT) is transforming healthcare by enabling seamless connectivity among medical devices, systems, and services. However, it also introduces serious cybersecurity and patient safety concerns as attackers increasingly exploit new methods and emerging vulnerabilities to infiltrate IoMT networks. This paper proposes a novel Tsetlin Machine (TM)-based Intrusion Detection System (IDS) for detecting a wide range of cyberattacks targeting IoMT networks. The TM is a rule-based and interpretable machine learning (ML) approach that models attack patterns using propositional logic. Extensive experiments conducted on the CICIoMT-2024 dataset, which includes multiple IoMT protocols and cyberattack types, demonstrate that the proposed TM-based IDS outperforms traditional ML classifiers. The proposed model achieves an accuracy of 99.5% in binary classification and 90.7% in multi-class classification, surpassing existing state-of-the-art approaches. Moreover, to enhance model trust and interpretability, the proposed TM-based model presents class-wise vote scores and clause activation heatmaps, providing clear insights into the most influential clauses and the dominant class contributing to the final model decision.
Figures & tables
Fig. 1: A simple illustration of IoMT working in healthcare.
Fig. 2: The overall architecture of the proposed TM-based IDS.
Protocol: Bluetooth
S.No.
Feature name
S.No.
Feature name
S.No.
Feature name
S.No.
Feature name
1
Header_Length
2
Protocol_Type
3
Packet_Type
4
Rate
5
HCI_Command
6
HCI_Event
7
HCI_ACL_Data
8
HCI_SCO_Data
9
Command_Complete
10
Command_Status
11
LE_Meta
12
Connection_Complete
13
Disconnection_Complete
14
Inquiry_Complete
15
Advertising_Report
16
Read_Remote_Features
17
Encryption_Change
18
Number_Completed_Packets
19
Tot_sum
20
Min
TABLE I: List of features present in CICIoMT24.
IoMT Protocol
Attack type
Number of samples
Training
Testing
Bluetooth
Benign
21750
6533
DoS
99840
25171
Total samples
121590
31704
MQTT and Wi-Fi
Benign
19291
37607
DoS
1805529
416676
TABLE II: Different attacks present in CICIoMT24.
Fig. 3: Class imbalance in Scenario 1: Binary classification.
Fig. 4: Balanced training class in Scenario 1.
Model
Parameters
TM
Binarizer: KBinsDiscretizer, n_bins =5,
encode= onehot-dense , strategy=quantile
number_of_clauses =100, T =10, s =2,
weighted_clauses =False, Epochs=10
DT
class_weight=balanced
RF
class_weight=balanced
TABLE III: Parameters used for models in Scenario 1.
Model
Accuracy
Precision
Recall
F1-score
Inference time
(in %)
(in %)
(in %)
(in %)
(in microseconds \SIUnitSymbolMicros )
TM
0.995
0.999
0.991
0.995
0.743
DT
0.997
0.997
0.997
0.997
0.045
RF
0.998
0.999
0.996
0.998
1.930
XGBoost
0.998
0.999
0.997
0.998
0.140
LGBM
0.998
0.999
0.996
0.998
0.516
TABLE IV: Model performance in Scenario 1.
Fig. 5: TM confusion matrix in Scenario 1 (two classes).
Fig. 6: Class-wise votes of a Benign sample in Scenario 1.
Fig. 7: Clause activation heatmap of the same Benign sample.
Fig. 8: Class imbalance in Scenario 2: Multi-class (six-class) classification.
Fig. 9: Balanced training class in Scenario 2.
Model
Parameters
TM
Binarizer: The same settings as in Table III
number_of_clauses =100, T =10, s =5,
weighted_clauses =False, Epochs=15
DT, RF
compute_class_weight method with class_weight=balanced
XGBoost
objective=multi:softprob, eval_metric =mlogloss,
tree_method =hist, learning_rate =0.1,
TABLE V: Parameters used for models in Scenario 2.
Model
Accuracy
Precision
Recall
F1-score
Inference time
(in %)
(in %)
(in %)
(in %)
(in microseconds \SIUnitSymbolMicros )
TM
0.907
0.913
0.907
0.906
4.056
DT
0.803
0.858
0.856
0.857
0.217
RF
0.816
0.869
0.875
0.872
17.588
XGBoost
0.862
0.868
0.927
0.890
1.498
LGBM
0.860
0.866
0.925
0.888
3.953
TABLE VI: Model performance in Scenario 2.
Fig. 10: Training and Testing accuracy of the TM model.
Fig. 11: TM confusion matrix in Scenario 2 (six classes).
Fig. 12: Class-wise votes of a Recon sample in Scenario 2.
Fig. 13: Clause activation heatmap of the same Recon sample.
Fig. 14: Class imbalance in Scenario 3: Seven classes.
Model
Parameters
TM
Binarizer: The same settings as in Table V
number_of_clauses =120, T =15, s =2, Epochs=15
DT, RF, XGBoost,
The same settings as in Table V
LGBM, KNN, NB, LR
NN
The same settings as in Table V except
input layer=11 neurons, output layer=7 neurons
TABLE VII: Parameters used for models in Scenario 3.
Model
Accuracy
Precision
Recall
F1-score
Inference time
(in %)
(in %)
(in %)
(in %)
(in microseconds \SIUnitSymbolMicros )
TM
0.884
0.889
0.884
0.884
5.413
DT
0.785
0.863
0.863
0.863
0.180
RF
0.800
0.876
0.878
0.877
15.470
XGBoost
0.851
0.882
0.920
0.895
1.650
LGBM
0.788
0.807
0.888
0.820
4.319
TABLE VIII: Model performance in Scenario 3.
Fig. 15: Class-wise votes of a Benign sample in Scenario 3.
Fig. 16: Class-wise votes of a DDoS sample in Scenario 3.
Model
Class
Accuracy
Precision
Recall
F1-score
Paper [ 11 ]
DNN/
Binary
0.99
0.99
0.99
0.99
LSTM
Six
0.78
0.77
0.77
0.75
Paper [ 12 ]
ML
Binary
0.996
0.971
0.951
0.961
Classifier
Six
0.735
0.735
0.713
0.676
Proposed
Tsetlin
Binary
0.995
0.999
0.991
0.995
Machine
Six
0.907
0.913
0.907
0.906
TABLE IX: Comparison with the state-of-the-art methods.
The rapid evolution of digital health technologies is redefining healthcare services worldwide. The integration of wireless communication and Internet-enabled medical devices within Internet of Medical Things (IoMT) networks enables continuous, real-time patient monitoring. However, this increased connectivity raises cybersecurity and patient safety risks due to increasingly sophisticated cyberattacks. This paper proposes a novel on-device, interpretable Tsetlin Machine (TM)-based Intrusion Detection System (IDS) to identify various phases of cyberattacks in IoMT environments. The TM is a rule-driven and transparent machine learning (ML) approach that represents attack patterns using propositional logic. Extensive evaluations on the MedSec-25 dataset, encompassing various phases of realistic cyberattacks, show that the proposed model outperforms ML models, attaining an F1-score (macro) of 97.83%. Moreover, the proposed model offers explicit explanations of its decisions to enhance transparency using feature-level contributions, class-wise vote scores, and clause activation heatmaps. Edge deployment (Raspberry Pi) further supports real-time on-device inference and intrusion detection. The combination of interpretability and high performance makes the proposed model well-suited for IoMT healthcare, where trust, reliability, safety, and timely decision-making are critical.
The rapid proliferation of Internet of Medical Things (IoMT) devices introduces critical cybersecurity vulnerabilities in healthcare environments where resource-constrained medical devices operate under strict latency requirements and stringent data-privacy regulations. To address these challenges, this paper presents the Lightweight Digital Twin and Federated Reinforcement Learning (LDT-FRL) framework, a privacy-preserving defense architecture integrating four complementary mechanisms: a Temporal Attention Encoder (TAE) built on a GRU backbone with learned temporal self-attention for flow-level threat classification; lightweight LSTM-based Digital Twins trained on normal-class traffic to generate per-device anomaly scores that gate the TAE classifier through a learned sigmoid coupling; a Federated Proximal Policy Optimization (PPO) agent selecting among ALLOW, ISOLATE, and HONEYPOT_REDIRECT actions based on a seven-dimensional state; and an intelligent honeypot layer that converts redirected suspicious traffic into actionable threat intelligence. A federated aggregation strategy employing EMA-smoothed per-client validation losses as inverse-weighted FedAvg coefficients stabilizes global model updates under non-IID client distributions. Evaluated on CICDDoS 2019 and TON-IoT benchmarks, LDT-FRL achieves 99.66% and 99.95% test accuracy respectively, with macro-F1 scores of 0.9913 and 0.9995, converging 81% faster than the DTFL-CD baseline while attaining perfect F1=1.000 on the severely imbalanced MITM class. Explainability analysis via SHAP, LIME, Grad-CAM, and counterfactual methods confirms that the TAE focuses on semantically meaningful flow features, providing interpretable evidence for each defense decision.
Syed Zeeshan Haider, Anwar Shah, Muneeb Arif +2
FAST National University of Computer and Emerging Sciences, Islamabad, Pakistan
Internet of Medical Things (IoMT) networks are hard to protect: devices are heterogeneous, computing resources are scarce, and traffic must be analyzed in real time. We present an intrusion detection system that addresses these constraints through feature selection. A Pearson correlation filter first removes redundant attributes; a hybrid strategy then combines model-based feature importance with SHAP attribution to pick a compact subset, on which we train Random Forest and LightGBM classifiers. SHAP and LIME explain what each retained feature contributes to the decisions. On CIC-IoMT 2024 and CIC-IDS 2017, the method cuts the feature space by up to 88% - from 40 to as few as 5 features - and accuracy and F1-score stay within a few points of models trained on all features. Compact, interpretable detectors of this kind are practical candidates for deployment on resource-limited medical networks.