Keecheon Kim
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Keecheon Kim has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- DongJun Seo
Department of Computer Science and Engineering, Konkuk University, Seoul 05029, Republic of Korea
2 shared publications Ung Heo
Department of IT Convergence Information Security; Konkuk University; Seoul Korea
2 shared publicationsYewon Oh
Konkuk University
2 shared publicationsYoungse Kim
Department of IT Convergence Information Security; Konkuk University; Seoul Korea
2 shared publicationsGreen Kim
Department of IT Convergence Information Security Konkuk University Seoul Korea
1 shared publication- Gyou Beom Kim
Department of Unmanned Aircraft System Engineering, Catholic Kwandong University, 579 beon-gil, Gangneung-si 25601, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 9 June 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (12) , 5809
Deployment-Oriented Interpretable Fraud Detection via Hybrid Explainable Boosting Machines with Concept–Raw Fusion on the IEEE-CIS Benchmark
Fraud detection models often achieve a strong ranking performance through black-box ensembles, but operational deployment also requires calibration, low explanation cost, and auditable scoring logic. This study develops an interpretable fraud-detection pipeline for IEEE-CIS by combining a 63-variable causal concept bank with teacher-guided additive Explainable Boosting Machine (EBM) students. The concept bank summarizes the temporal state, entity history, novelty/reuse, identity missingness, and aggregate deviation. Experiments use a chronological out-of-time split and a stricter pseudo-entity-disjoint holdout. In the main three-seed evaluation, the CatBoost predictive ceiling and XGBoost teacher achieved PR-AUC 0.489 ± 0.001 and 0.478 ± 0.003, respectively. Among interpretable models, concept-only EBM reached 0.189 ± 0.000, raw-only EBMs reached 0.372 ± 0.005 (top-k = 8) and 0.383 ± 0.002 (top-k = 12), and hybrid EBMs reached 0.407 ± 0.003 (top-k = 8) and 0.407 ± 0.004 (top-k = 12), consistently improving over matched raw-only additive baselines. The final top-k = 8 hybrid reduced input features from 154 to 71, achieved about 9.7× faster inference than XGBoost, remained close to XGBoost in ECE-15 calibration (0.01587 vs. 0.01611) while having a higher Brier score, and produced native local explanations far faster than XGBoost + SHAP. The results position CatBoost as the predictive ceiling and hybrid EBM as a benchmark-supported, deployment-relevant interpretable compromise for applied financial risk-screening workflows, rather than as a production-validated fraud-monitoring system.
- 9 June 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (12) , 5809
Open AccessDeployment-Oriented Interpretable Fraud Detection via Hybrid Explainable Boosting Machines with Concept–Raw Fusion on the IEEE-CIS Benchmark
Fraud detection models often achieve a strong ranking performance through black-box ensembles, but operational deployment also requires calibration, low explanation cost, and auditable scoring logic. This study develops an interpretable fraud-detection pipeline for IEEE-CIS by combining a 63-variable causal concept bank with teacher-guided additive Explainable Boosting Machine (EBM) students. The concept bank summarizes the temporal state, entity history, novelty/reuse, identity missingness, and aggregate deviation. Experiments use a chronological out-of-time split and a stricter pseudo-entity-disjoint holdout. In the main three-seed evaluation, the CatBoost predictive ceiling and XGBoost teacher achieved PR-AUC 0.489 ± 0.001 and 0.478 ± 0.003, respectively. Among interpretable models, concept-only EBM reached 0.189 ± 0.000, raw-only EBMs reached 0.372 ± 0.005 (top-k = 8) and 0.383 ± 0.002 (top-k = 12), and hybrid EBMs reached 0.407 ± 0.003 (top-k = 8) and 0.407 ± 0.004 (top-k = 12), consistently improving over matched raw-only additive baselines. The final top-k = 8 hybrid reduced input features from 154 to 71, achieved about 9.7× faster inference than XGBoost, remained close to XGBoost in ECE-15 calibration (0.01587 vs. 0.01611) while having a higher Brier score, and produced native local explanations far faster than XGBoost + SHAP. The results position CatBoost as the predictive ceiling and hybrid EBM as a benchmark-supported, deployment-relevant interpretable compromise for applied financial risk-screening workflows, rather than as a production-validated fraud-monitoring system.
- Open Access
- 9 June 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (12) , 5811
An Integrated Pipeline for Intent-Based Zero-Touch Networks: From Intent Translation to Minimal-Modification Reconfiguration
To support Industry 5.0 smart factories that require ultra-low latency and high reliability, this paper proposes a three-layer Intent-Based Zero-Touch Networking (IBZTN) pipeline. Existing Intent-Based Networking (IBN)/Zero-Touch Networking (ZTN) studies often remain conceptual, while Graph Neural Network (GNN)-based Quality of Service (QoS) prediction and Deep Reinforcement Learning (DRL)-based reconfiguration are usually developed as separate modules. The proposed pipeline connects natural-language intent translation, feasibility prediction, and minimal-modification reconfiguration through a validated QoS contract and feasibility-aware closed-loop structure. Layer 1 converts intents into quantitative QoS profiles by combining Retrieval-Augmented Generation (RAG) with schema- and rule-based validation. Layer 2 evaluates feasibility using a Graph Isomorphism Network with Edge features (GINE)-based binary classifier. Layer 3 recovers infeasible states using a Behavior Cloning (BC) Proximal Policy Optimization (PPO) agent with Smart Traffic Engineering (TE) masking. In experiments with 300 natural language intents, RAG+Validator reduced Layer 1 constraint violations to 0.0% for most evaluated cloud and local Large Language Models (LLMs). The Layer 2 predictor achieved a 93.9% F1-score, and Layer 3 achieved an 87.8% recovery success rate with 9.8 average modifications and 5.56 ms inference latency. These results demonstrate the simulation-level potential of IBZTN and motivate future hardware-in-the-loop validation in industrial networks.
- 9 June 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (12) , 5811
Open AccessAn Integrated Pipeline for Intent-Based Zero-Touch Networks: From Intent Translation to Minimal-Modification Reconfiguration
To support Industry 5.0 smart factories that require ultra-low latency and high reliability, this paper proposes a three-layer Intent-Based Zero-Touch Networking (IBZTN) pipeline. Existing Intent-Based Networking (IBN)/Zero-Touch Networking (ZTN) studies often remain conceptual, while Graph Neural Network (GNN)-based Quality of Service (QoS) prediction and Deep Reinforcement Learning (DRL)-based reconfiguration are usually developed as separate modules. The proposed pipeline connects natural-language intent translation, feasibility prediction, and minimal-modification reconfiguration through a validated QoS contract and feasibility-aware closed-loop structure. Layer 1 converts intents into quantitative QoS profiles by combining Retrieval-Augmented Generation (RAG) with schema- and rule-based validation. Layer 2 evaluates feasibility using a Graph Isomorphism Network with Edge features (GINE)-based binary classifier. Layer 3 recovers infeasible states using a Behavior Cloning (BC) Proximal Policy Optimization (PPO) agent with Smart Traffic Engineering (TE) masking. In experiments with 300 natural language intents, RAG+Validator reduced Layer 1 constraint violations to 0.0% for most evaluated cloud and local Large Language Models (LLMs). The Layer 2 predictor achieved a 93.9% F1-score, and Layer 3 achieved an 87.8% recovery success rate with 9.8 average modifications and 5.56 ms inference latency. These results demonstrate the simulation-level potential of IBZTN and motivate future hardware-in-the-loop validation in industrial networks.
- Open Access
- 26 February 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (5) , 2284
Operationally Constrained Zero-Day Intrusion Detection with Target-FPR Calibration and Similarity Graph Construction
Intrusion detectors are often evaluated using average metrics at unconstrained thresholds, yet deployments require explicit control over false alarms. We investigate zero-day (out-of-distribution, OOD) intrusion detection under a target-FPR calibrated protocol, where a threshold is set on benign validation traffic to satisfy a target false positive rate α and transferred, unchanged, to a seen-test and OOD-test. Using CICIDS2017-derived host-session nodes aggregated in 1 min and 5 min windows, we compare tabular baselines, message-passing GNNs on a rule-based graph, and employ a method that builds a k-nearest-neighbor similarity graph with lightweight feature pre-smoothing. Robustness is measured using the OOD violation ratio, percentile tail risk, and feasibility under explicit false-alarm budgets. Base-graph GNNs exhibit heavy-tailed false-alarm amplification under OOD shifts: at α = 0.001, the p95 violation ratio reaches 68.50 (1 m) and 67.95 (5 m). In contrast, the proposed method reduces p95 to 3.41 (1 m) and 1.15 (5 m) and improves budget feasibility. We further verify robustness beyond a single held-out family by evaluating additional unseen-family splits (e.g., DDoS and DDoS+DoS) under the same calibrated operating point. We also quantify deployment-oriented cost via edge-list size and practical parsing/loading time. These findings suggest that similarity-based graphs with light pre-smoothing improve deployability under distribution shifts.
- 26 February 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (5) , 2284
Open AccessOperationally Constrained Zero-Day Intrusion Detection with Target-FPR Calibration and Similarity Graph Construction
Intrusion detectors are often evaluated using average metrics at unconstrained thresholds, yet deployments require explicit control over false alarms. We investigate zero-day (out-of-distribution, OOD) intrusion detection under a target-FPR calibrated protocol, where a threshold is set on benign validation traffic to satisfy a target false positive rate α and transferred, unchanged, to a seen-test and OOD-test. Using CICIDS2017-derived host-session nodes aggregated in 1 min and 5 min windows, we compare tabular baselines, message-passing GNNs on a rule-based graph, and employ a method that builds a k-nearest-neighbor similarity graph with lightweight feature pre-smoothing. Robustness is measured using the OOD violation ratio, percentile tail risk, and feasibility under explicit false-alarm budgets. Base-graph GNNs exhibit heavy-tailed false-alarm amplification under OOD shifts: at α = 0.001, the p95 violation ratio reaches 68.50 (1 m) and 67.95 (5 m). In contrast, the proposed method reduces p95 to 3.41 (1 m) and 1.15 (5 m) and improves budget feasibility. We further verify robustness beyond a single held-out family by evaluating additional unseen-family splits (e.g., DDoS and DDoS+DoS) under the same calibrated operating point. We also quantify deployment-oriented cost via edge-list size and practical parsing/loading time. These findings suggest that similarity-based graphs with light pre-smoothing improve deployability under distribution shifts.
- Open Access
- 19 November 2025
- Published by MDPI AG in Electronics
- Vol. 14 (22) , 4524
Offline Metrics vs. Online Performance in SDN: A Performance Reversal Study of MLP and GraphSAGE
Software-Defined Networking (SDN) provides centralized control over routing paths through a logically centralized controller. Although Graph Neural Networks (GNNs) such as GraphSAGE have shown strong potential for network topology analysis, their superiority over simpler models like the Multi-Layer Perceptron (MLP) in dynamic SDN control remains unclear. In this study, we compare MLP and GraphSAGE using three training data volumes (70, 100, and 140) and spatio-temporal features that integrate spatial and temporal characteristics of each node. Experimental results reveal a distinct discrepancy between offline classification metrics and online SDN performance. Offline evaluation showed that MLP achieved a slightly higher F1-score (0.62) than GraphSAGE (0.59). However, when deployed in a SDN controller, GraphSAGE reduced latency by 17%, increased throughput by 8%, and improved jitter by 31%. These results demonstrate that higher offline accuracy does not necessarily translate into better real-time control performance, since offline metrics fail to capture topology-aware routing, congestion recovery, and dynamic adaptation effects. The findings provide a practical guideline for SDN-oriented AI model evaluation, emphasizing end-to-end system performance over isolated offline metrics.
- 19 November 2025
- Published by MDPI AG in Electronics
- Vol. 14 (22) , 4524
Open AccessOffline Metrics vs. Online Performance in SDN: A Performance Reversal Study of MLP and GraphSAGE
Software-Defined Networking (SDN) provides centralized control over routing paths through a logically centralized controller. Although Graph Neural Networks (GNNs) such as GraphSAGE have shown strong potential for network topology analysis, their superiority over simpler models like the Multi-Layer Perceptron (MLP) in dynamic SDN control remains unclear. In this study, we compare MLP and GraphSAGE using three training data volumes (70, 100, and 140) and spatio-temporal features that integrate spatial and temporal characteristics of each node. Experimental results reveal a distinct discrepancy between offline classification metrics and online SDN performance. Offline evaluation showed that MLP achieved a slightly higher F1-score (0.62) than GraphSAGE (0.59). However, when deployed in a SDN controller, GraphSAGE reduced latency by 17%, increased throughput by 8%, and improved jitter by 31%. These results demonstrate that higher offline accuracy does not necessarily translate into better real-time control performance, since offline metrics fail to capture topology-aware routing, congestion recovery, and dynamic adaptation effects. The findings provide a practical guideline for SDN-oriented AI model evaluation, emphasizing end-to-end system performance over isolated offline metrics.
- Open Access
- 17 April 2025
- Published by MDPI AG in Electronics
- Vol. 14 (8) , 1629
V2X Network-Based Enhanced Cooperative Autonomous Driving for Urban Clusters in Real Time: A Model for Control, Optimization and Security
For the commercialization of connected vehicles and smart cities, extensive research is carried out on autonomous driving, Vehicle-to-Everything (V2X) communication, and platooning. However, limitations remain, such as restrictions to highway environments, and studies are conducted separately due to challenges in ensuring reliability and real-time performance under external influences. This paper proposes a cooperative autonomous driving system based on V2X network implemented in the CARLA simulator, which simulates an urban environment to optimize vehicle-embedded systems and ensure safety and real-time performance. First, the proposed Throttle–Steer–Brake (TSB) driving technique reduces the computational overhead for following vehicles by utilizing the control commands of a leading vehicle. Second, a V2X network is designed to support object perception, cluster escape, and joining. Third, an urban perception system is developed and validated for safety. Finally, pseudonymized vehicle identifiers, Advanced Encryption Standard (AES), and the Edwards-curve Digital Signature Algorithm (EdDSA) are employed for data reliability and security. The system is validated in processing time and accuracy, confirming feasibility for real-world application. TSB driving demonstrates a computation speed approximately 466 times faster than conventional waypoints-based driving. Accurate urban perception and V2X communication enable safe cluster escape and joining, establishing a foundation for cooperative autonomous driving with improved safety and real-time capabilities.
- 17 April 2025
- Published by MDPI AG in Electronics
- Vol. 14 (8) , 1629
Open AccessV2X Network-Based Enhanced Cooperative Autonomous Driving for Urban Clusters in Real Time: A Model for Control, Optimization and Security
For the commercialization of connected vehicles and smart cities, extensive research is carried out on autonomous driving, Vehicle-to-Everything (V2X) communication, and platooning. However, limitations remain, such as restrictions to highway environments, and studies are conducted separately due to challenges in ensuring reliability and real-time performance under external influences. This paper proposes a cooperative autonomous driving system based on V2X network implemented in the CARLA simulator, which simulates an urban environment to optimize vehicle-embedded systems and ensure safety and real-time performance. First, the proposed Throttle–Steer–Brake (TSB) driving technique reduces the computational overhead for following vehicles by utilizing the control commands of a leading vehicle. Second, a V2X network is designed to support object perception, cluster escape, and joining. Third, an urban perception system is developed and validated for safety. Finally, pseudonymized vehicle identifiers, Advanced Encryption Standard (AES), and the Edwards-curve Digital Signature Algorithm (EdDSA) are employed for data reliability and security. The system is validated in processing time and accuracy, confirming feasibility for real-world application. TSB driving demonstrates a computation speed approximately 466 times faster than conventional waypoints-based driving. Accurate urban perception and V2X communication enable safe cluster escape and joining, establishing a foundation for cooperative autonomous driving with improved safety and real-time capabilities.
- Open Access
Estimating Drone Visual Line-of-Sight Distance Using Machine Learning Approaches
In this study, we conducted flight tests to establish a clear standard for the visual line-of-sight (VLOS) distance of drones using machine learning models, as outlined in the Aviation Safety Act. Various machine learning models were applied and compared to predict the VLOS distance based on flight data. The analysis revealed that factors such as flight altitude, drone size, and observer’s vision significantly influence the VLOS distance. In particular, drone volume and observer’s vision were identified as the most important factors in predicting VLOS distance. The Random Forest Regression model demonstrated the best predictive performance, followed by the Polynomial Regression model. This study provides fundamental data to ensure safe drone operations and compliance with aviation regulations. The findings can also serve as practical resources for drone operators in planning safe flights. Future works should focus on collecting data from diverse environmental conditions to improve the generalization of prediction models. Additional research is also needed on beyond visual line-of-sight (BVLOS) and night flights, as these are critical areas for drone commercialization and require new predictive models and technological advancements to ensure safety.
Open AccessEstimating Drone Visual Line-of-Sight Distance Using Machine Learning Approaches
In this study, we conducted flight tests to establish a clear standard for the visual line-of-sight (VLOS) distance of drones using machine learning models, as outlined in the Aviation Safety Act. Various machine learning models were applied and compared to predict the VLOS distance based on flight data. The analysis revealed that factors such as flight altitude, drone size, and observer’s vision significantly influence the VLOS distance. In particular, drone volume and observer’s vision were identified as the most important factors in predicting VLOS distance. The Random Forest Regression model demonstrated the best predictive performance, followed by the Polynomial Regression model. This study provides fundamental data to ensure safe drone operations and compliance with aviation regulations. The findings can also serve as practical resources for drone operators in planning safe flights. Future works should focus on collecting data from diverse environmental conditions to improve the generalization of prediction models. Additional research is also needed on beyond visual line-of-sight (BVLOS) and night flights, as these are critical areas for drone commercialization and require new predictive models and technological advancements to ensure safety.
- Open Access
- 11 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 50641-50650
Enhancing Reinforcement Learning Performance in Delayed Reward System Using DQN and Heuristics
This paper suggests and implements how to apply the reinforcement learning on delayed reward system which is known to be complex to apply the machine learning technology such as Q-learning. Such games as Tetris game is known to be a delayed reward system because of its characteristics of generating sparse reward in learning process. Tetris game requires the actor’s quick judgment ability and speed of response because the blocks must be stacked in an optimal location quickly, considering the random shape and rotation of appearing blocks. Also, since the number of cases is very large due to the various block types and order, if a human-being is playing the game, the performance is limited by simply relying on human memorization capability. Therefore, we applied a reinforcement learning implemented in this study for this delayed reward system. We find that the general legacy reinforcement learning method shows its limitation in improving the performance. Hence, we apply the heuristic to increase the decision accuracy as the weighting method of reward. As a result, we were able to obtain high scores in games. Although it is not yet possible to say that this heuristic(rule-based) approach has completely conquered the game. In several experiments, this hybrid reinforcement learning shows better playability than human in terms of learning speed, as well as high scores. In this paper, it is shown that general Q-learning is not suitable for delayed reward system. And a hybrid learning that adds prioritized experience replay tactics, and the related techniques and algorithms are introduced to increase the reinforcement learning performance.
- 11 May 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 50641-50650
Open AccessEnhancing Reinforcement Learning Performance in Delayed Reward System Using DQN and Heuristics
This paper suggests and implements how to apply the reinforcement learning on delayed reward system which is known to be complex to apply the machine learning technology such as Q-learning. Such games as Tetris game is known to be a delayed reward system because of its characteristics of generating sparse reward in learning process. Tetris game requires the actor’s quick judgment ability and speed of response because the blocks must be stacked in an optimal location quickly, considering the random shape and rotation of appearing blocks. Also, since the number of cases is very large due to the various block types and order, if a human-being is playing the game, the performance is limited by simply relying on human memorization capability. Therefore, we applied a reinforcement learning implemented in this study for this delayed reward system. We find that the general legacy reinforcement learning method shows its limitation in improving the performance. Hence, we apply the heuristic to increase the decision accuracy as the weighting method of reward. As a result, we were able to obtain high scores in games. Although it is not yet possible to say that this heuristic(rule-based) approach has completely conquered the game. In several experiments, this hybrid reinforcement learning shows better playability than human in terms of learning speed, as well as high scores. In this paper, it is shown that general Q-learning is not suitable for delayed reward system. And a hybrid learning that adds prioritized experience replay tactics, and the related techniques and algorithms are introduced to increase the reinforcement learning performance.
- Open Access
- 30 March 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (7) , 3520
Multi-Agent Deep Q Network to Enhance the Reinforcement Learning for Delayed Reward System
This study examines various factors and conditions that are related with the performance of reinforcement learning, and defines a multi-agent DQN system (N-DQN) model to improve them. N-DQN model is implemented in this paper with examples of maze finding and ping-pong as examples of delayed reward system, where delayed reward occurs, which makes general DQN learning difficult to apply. The implemented N-DQN shows about 3.5 times higher learning performance compared to the Q-Learning algorithm in the reward-sparse environment in the performance evaluation, and compared to DQN, it shows about 1.1 times faster goal achievement speed. In addition, through the implementation of the prioritized experience replay and the implementation of the reward acquisition section segmentation policy, such a problem as positive-bias of the existing reinforcement learning models seldom or never occurred. However, according to the characteristics of the architecture that uses many numbers of actors in parallel, the need for additional research on light-weighting the system for further performance improvement has raised. This paper describes in detail the structure of the proposed multi-agent N_DQN architecture, the contents of various algorithms used, and the specification for its implementation.
- 30 March 2022
- Published by MDPI AG in Applied Sciences
- Vol. 12 (7) , 3520
Open AccessMulti-Agent Deep Q Network to Enhance the Reinforcement Learning for Delayed Reward System
This study examines various factors and conditions that are related with the performance of reinforcement learning, and defines a multi-agent DQN system (N-DQN) model to improve them. N-DQN model is implemented in this paper with examples of maze finding and ping-pong as examples of delayed reward system, where delayed reward occurs, which makes general DQN learning difficult to apply. The implemented N-DQN shows about 3.5 times higher learning performance compared to the Q-Learning algorithm in the reward-sparse environment in the performance evaluation, and compared to DQN, it shows about 1.1 times faster goal achievement speed. In addition, through the implementation of the prioritized experience replay and the implementation of the reward acquisition section segmentation policy, such a problem as positive-bias of the existing reinforcement learning models seldom or never occurred. However, according to the characteristics of the architecture that uses many numbers of actors in parallel, the need for additional research on light-weighting the system for further performance improvement has raised. This paper describes in detail the structure of the proposed multi-agent N_DQN architecture, the contents of various algorithms used, and the specification for its implementation.
- Open Access
- 28 December 2020
- Published by MDPI AG in Electronics
- Vol. 10 (1) , 36
Traffic Type Recognition Method for Unknown Protocol—Applying Fuzzy Inference
In this paper, we propose a system that can recognize traffic types without prior knowledge of static features such as protocol header information by combining protocol analysis based on an ecological sequence alignment algorithm in a bioinformatics and fuzzy inference system. The algorithm proposed in this paper obtained up to a 91% level of performance at a similar level to several existing algorithms in experiments using datasets containing various types of traffic. In addition, it showed an excellent accuracy of 82.5% or more even under severe conditions that lowered the amount of data to a level of at least 40% or only included data in the middle of the traffic. This shows that the problem of dependence on initial data that frequently occurs in existing machine learning and deep learning-based traffic classification algorithms does not appear in the proposed algorithm. Furthermore, based on the ability to directly extract traffic characteristics without being dependent on static field values, it has secured the ability to respond with a small number of data by taking advantage of the flexibility of the membership function of the fuzzy inference engine. Through this, the applicability to low-power and low-performance environments such as IoT networks was confirmed. In this paper, we describe in detail the theoretical background for constructing such an algorithm and relevant experiments and considerations for actual verification.
- 28 December 2020
- Published by MDPI AG in Electronics
- Vol. 10 (1) , 36
Open AccessTraffic Type Recognition Method for Unknown Protocol—Applying Fuzzy Inference
In this paper, we propose a system that can recognize traffic types without prior knowledge of static features such as protocol header information by combining protocol analysis based on an ecological sequence alignment algorithm in a bioinformatics and fuzzy inference system. The algorithm proposed in this paper obtained up to a 91% level of performance at a similar level to several existing algorithms in experiments using datasets containing various types of traffic. In addition, it showed an excellent accuracy of 82.5% or more even under severe conditions that lowered the amount of data to a level of at least 40% or only included data in the middle of the traffic. This shows that the problem of dependence on initial data that frequently occurs in existing machine learning and deep learning-based traffic classification algorithms does not appear in the proposed algorithm. Furthermore, based on the ability to directly extract traffic characteristics without being dependent on static field values, it has secured the ability to respond with a small number of data by taking advantage of the flexibility of the membership function of the fuzzy inference engine. Through this, the applicability to low-power and low-performance environments such as IoT networks was confirmed. In this paper, we describe in detail the theoretical background for constructing such an algorithm and relevant experiments and considerations for actual verification.