Hossein Ahmadi

Dr. Hossein Ahmadi

DET, Politecnico di Torino, Turin, Piedmont, Italy
orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 January 2023 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
MFMoslem Forouzesh
Moslem Forouzesh
FKFarid Samsami Khodadad
Farid Samsami Khodadad
PAPaeiz Azmi
Paeiz Azmi
AKAli Kuhestani
Ali Kuhestani
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
In practice, Internet of Things (IoT) applications may face two different hostile attacks, i.e., overhearing the broadcasted data and detecting the presence of data communication. Previous works have assumed that only one of these attacks is present in the network and have addressed these attacks separately. In this article, we investigate the security of a simple but practical IoT-based point-to-point communication which is facing with the above-mentioned two attacks under some realistic assumptions. In other words, we aim to protect the data against two adversary nodes, namely Eve (who tries to capture the broadcasted data) and warden Willie (who tries to detect the presence of data communication between legitimate nodes). To tackle these attacks, we jointly employ the information-theoretic security and covert communication techniques. To hide information from Willies, we force the source to transmit its message in the selected time slot and the multiple-antennas cooperative jammer to inject jamming during all time slots contentiously which leads to deceive Eves and Willies jointly. To enhance the performance of this security design, we formulate an optimization problem, with the goal of maximizing the secrecy rate subject to power constraints at the source and jammer, while satisfying a covert communication requirement. Since the mentioned optimization problem is nonconvex, we propose an algorithm to solve it. Then, to obtain further insights, we extend our proposed system model to noncolluding and colluding Willies. Next, we study a practical communication scenario, where the source has uncertainty about Willie’s location and knows the channel state information (CSI) of Eves imperfectly. Our numerical results highlight that a more accurate estimation of Willie’s location has a more positive efficacy on the secrecy performance compared to the case with more accuracy of Eves’ CSI estimation.

10 March 2024 in Electronics
Verified
Journal Article
Open access
PDF file
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
Luca MesinLuca Mesin
Luca Mesin
In the evolving field of Brain–Computer Interfaces (BCIs), accurately classifying Electroencephalography (EEG) signals for Motor Imagery (MI) tasks is challenging. We introduce the Correlation-Optimized Weighted Stacking Ensemble (COWSE) model, an innovative ensemble learning framework designed to improve MI EEG signal classification. The COWSE model integrates sixteen machine learning classifiers through a weighted stacking approach, optimizing performance by balancing the strengths and weaknesses of each classifier based on error correlation analysis and performance metrics evaluation across benchmark datasets. The COWSE model’s development involves selecting base classifiers, dynamically assigning weights according to performance, and employing a meta-classifier trained on these weighted predictions. Testing on the BNCI2014-002 dataset, the COWSE model achieved classification accuracy exceeding 98.16%, marking a significant advancement in MI EEG classification. This study highlights the potential of integrating multiple machine learning classifiers to address the complex challenges of EEG signal classification. By achieving new benchmarks and showcasing enhanced classification capabilities, the COWSE model contributes significantly to BCI research, encouraging further exploration into advanced ensemble learning strategies.

29 July 2024 in IEEE Access
Verified
Journal Article
Open access
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
Luca MesinLuca Mesin
Luca Mesin
Electroencephalography (EEG) based Brain-Computer Interfaces (BCIs) are vital for various applications, yet achieving accurate EEG signal classification, particularly for Motor Imagery (MI) tasks, remains a significant challenge. This study introduces a novel Weighted and Stacked Adaptive Integrated Ensemble Classifier (WS-AIEC), employing a comprehensive approach across six MI EEG datasets with 16 diverse Machine Learning (ML) classifiers. Through evaluations that encompass metric-based comparisons and learning curve analyses, we systematically ranked and clustered the classifiers. The WS-AIEC integrates the top-performing classifiers from each cluster and employs a unique blend of weighted and stacked ensemble techniques. Our results demonstrate the WS-AIEC’s superior performance, achieving an exceptional accuracy of 99.58% on the BNCI2014-002 dataset and an average improvement of 20.23% in accuracy over the top-performing individual classifiers across all datasets. This significant enhancement underscores the innovative approach of our WS-AIEC in EEG signal classification for BCIs, setting a new benchmark for accuracy and reliability in the field.

6 May 2025 in Journal of Neural Engineering
Verified
Journal Article
Open access
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
Luca MesinLuca Mesin
Luca Mesin
Objective. Extracting universal, task-independent semantic features from electroencephalography (EEG) signals remains an open challenge. Traditional approaches are often task-specific, limiting their generalization across different EEG paradigms. This study aims to develop a robust, unsupervised framework for learning high-level, task-independent neural representations. Approach. We propose a novel framework integrating convolutional neural networks, AutoEncoders, and Transformers to extract both low-level spatiotemporal patterns and high-level semantic features from EEG signals. The model is trained in an unsupervised manner to ensure adaptability across diverse EEG paradigms, including motor imagery (MI), steady-state visually evoked potentials (SSVEPs), and event-related potentials (ERPs, specifically P300). Extensive analyses, including clustering, correlation, and ablation studies, are conducted to validate the quality and interpretability of the extracted features. Main results. Our method achieves state-of-the-art performance, with average classification accuracies of 83.50% and 84.84% on MI datasets (BCICIV_2a and BCICIV_2b), 98.41% and 99.66% on SSVEP datasets (Lee2019-SSVEP and Nakanishi2015), and an average AUC of 91.80% across eight ERP datasets. t-distributed stochastic neighbor embedding and clustering analyses reveal that the extracted features exhibit enhanced separability and structure compared to raw EEG data. Correlation studies confirm the framework’s ability to balance universal and subject-specific features, while ablation results highlight the near-optimality of the selected model configuration. Significance. This work establishes a universal framework for task-independent semantic feature extraction from EEG signals, bridging the gap between conventional feature engineering and modern deep learning methods. By providing robust, generalizable representations across diverse EEG paradigms, this approach lays the foundation for advanced brain–computer interface applications, cross-task EEG analysis, and future developments in semantic EEG processing.

20 January 2025 in IEEE Access
Verified
Journal Article
Open access
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
AKAli Kuhestani
Ali Kuhestani
MKMohammadreza Keshavarzi
Mohammadreza Keshavarzi
Luca MesinLuca Mesin
Luca Mesin
In this study, we investigate the effects of Adversarial Neural Network Training (ANNT) on the robustness and effectiveness of Brain-to-Brain Communication (B2B-C) systems using Steady-State Visually Evoked Potentials (SSVEP) EEG data. We utilized a combined Convolutional Neural Network-Temporal Convolutional Network (CNN-TCN) architecture to classify the data and assessed the system’s resistance to various adversarial strategies, including Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Basic Iterative Method (BIM), Carlini & Wagner (C&W), and Momentum Iterative Method (MIM). By analyzing publicly accessible datasets, specifically Lee2019_SSVEP and Nakanishi2015, we observed significant enhancements in both accuracy and AUC metrics when ANNT was applied. Specifically, the Lee2019_SSVEP dataset exhibited a 24% increase in accuracy and a 0.23-point improvement in AUC, while the Nakanishi2015 dataset demonstrated improvements of 9% and 0.07 points, respectively. Our results indicate that PGD posed the greatest challenge to the model, significantly reducing accuracy and AUC across various scenarios, whereas FGSM was the least impactful. These findings highlight ANNT’s potential in fortifying the security and stability of B2B-C systems against diverse adversarial conditions.

13 March 2024 in IEEE Access
Verified
Journal Article
Open access
Hossein AhmadiHossein Ahmadi
Hossein Ahmadi
AKAli Kuhestani
Ali Kuhestani
Luca MesinLuca Mesin
Luca Mesin
In the burgeoning field of brain-to-brain communication, ensuring the security and integrity of transmitted information is paramount. This study aims to enhance the robustness and security of brain-to-brain communication systems against adversarial attacks using electroencephalogram data through adversarial neural network training. By focusing on event-related potentials and employing eight diverse datasets, our objectives were to rigorously evaluate the model’s defence mechanisms against adversarial manipulations and optimise trial durations and sampling rates for maximum security. The results signify a notable advancement in system defence, evidenced by an average increase in adversarial accuracy by 17% and an improvement in the area under the curve by 0.12 points, demonstrating the effectiveness of our approach in strengthening brain-to-brain communication systems against sophisticated cyber threats.