Mohamed El Ghazouani
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Mohamed El Ghazouani has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
Latifa Er-Rajy
6 shared publicationsMy Ahmed El Kiram
5 shared publicationsCharafeddine Ait Zaouiat
4 shared publicationsLayla Aziz
Polydisciplinary Faculty of Sidi Bennour, Chouaib Doukkali University, El Jadida, Morocco
4 shared publications- Hanan Moufid
LISI Laboratory, Faculty of Sciences Semlalia, Cadi Ayyad University, Marrakech 40000, Morocco
3 shared publications - Latifa Er-Rajy
University Cadi Ayyad
3 shared publications
Latest Publications
- Open Access
Artificial Intelligence for Intrusion Detection and Prevention: A Systematic Review of Methods, Challenges, and Future Directions
This systematic review synthesizes AI-based techniques for detecting and preventing cyberattacks published between 2019 and 2025. Following PRISMA 2020, we searched IEEE Xplore, Scopus, and the ACM Digital Library (550 records) and included 74 studies after independent dual screening; study quality was appraised with a modified JBI checklist, the risk of bias of the review was assessed with ROBIS, and the review was not registered. We include a higher-level taxonomy of present methods divided into seven categories: (1) traditional machine learning algorithms; (2) deep learning models; (3) hybrid intelligence frameworks; (4) blockchain-based systems; (5) reinforcement learning approaches; (6) ensemble learning methods; and (7) explainable AI (XAI) techniques. Reported accuracies are high in all seven categories (category medians between 95.3% and 99.9%), but individual results range from 58.17% to 100% and were obtained with different datasets, data splits, and evaluation protocols, so they do not support a ranking of the categories. Latency and resource consumption are rarely reported and, when reported, were measured under non-comparable conditions. Automated prevention actions (blocking, rate limiting, or network reconfiguration) are evaluated far less often than detection, although false positives are more costly when they trigger blocking. Targeted machine learning methods report strong performance in phishing detection (96.83–98.3% accuracy in the tabulated studies). This study also identifies several major areas needing improvement, including a predominant reliance on legacy benchmark datasets (NSL-KDD, KDD99, UNSW-NB15), which were used by 25 of the 48 studies with identifiable datasets (52.1%) and a limited capability to interpret “black box” AI models. Accordingly, we suggest that future research use more realistic test environments and recent datasets, develop lightweight and adaptive AI models, and adopt a standardized minimum reporting protocol (accuracy, precision, recall, F1-score, false positive rate, latency, resource usage, and statistical validation), for which we provide a concrete proposal and a list of recent datasets. The intention of this study is to assist cybersecurity practitioners in identifying and implementing effective AI-based security solutions in relation to present-day threats.
Open AccessArtificial Intelligence for Intrusion Detection and Prevention: A Systematic Review of Methods, Challenges, and Future Directions
This systematic review synthesizes AI-based techniques for detecting and preventing cyberattacks published between 2019 and 2025. Following PRISMA 2020, we searched IEEE Xplore, Scopus, and the ACM Digital Library (550 records) and included 74 studies after independent dual screening; study quality was appraised with a modified JBI checklist, the risk of bias of the review was assessed with ROBIS, and the review was not registered. We include a higher-level taxonomy of present methods divided into seven categories: (1) traditional machine learning algorithms; (2) deep learning models; (3) hybrid intelligence frameworks; (4) blockchain-based systems; (5) reinforcement learning approaches; (6) ensemble learning methods; and (7) explainable AI (XAI) techniques. Reported accuracies are high in all seven categories (category medians between 95.3% and 99.9%), but individual results range from 58.17% to 100% and were obtained with different datasets, data splits, and evaluation protocols, so they do not support a ranking of the categories. Latency and resource consumption are rarely reported and, when reported, were measured under non-comparable conditions. Automated prevention actions (blocking, rate limiting, or network reconfiguration) are evaluated far less often than detection, although false positives are more costly when they trigger blocking. Targeted machine learning methods report strong performance in phishing detection (96.83–98.3% accuracy in the tabulated studies). This study also identifies several major areas needing improvement, including a predominant reliance on legacy benchmark datasets (NSL-KDD, KDD99, UNSW-NB15), which were used by 25 of the 48 studies with identifiable datasets (52.1%) and a limited capability to interpret “black box” AI models. Accordingly, we suggest that future research use more realistic test environments and recent datasets, develop lightweight and adaptive AI models, and adopt a standardized minimum reporting protocol (accuracy, precision, recall, F1-score, false positive rate, latency, resource usage, and statistical validation), for which we provide a concrete proposal and a list of recent datasets. The intention of this study is to assist cybersecurity practitioners in identifying and implementing effective AI-based security solutions in relation to present-day threats.
- Open Access
- 10 September 2026
- Published by MDPI AG in Journal of Cybersecurity and Privacy
- Vol. 6 (5) , 160
Metadata Compressibility and Evaluation Bias in Malicious Package Detection for NPM and PyPI
Malicious packages in open source-software supply chains are a growing security concern, yet machine learning detectors built on registry metadata are difficult to interpret and are typically evaluated under protocols susceptible to data leakage. We construct a dataset of 3330 package versions from NPM and PyPI in which every attribute is reconstructed at its exact publication timestamp, and we partition the data so that all versions of a package, all packages of a maintainer, and all members of a name-similarity family reside within a single fold. The primary analysis is restricted to NPM, where 18 of 456 point-in-time attributes attain an AU-PR of 0.9972 against a positive-class baseline of 0.8755 and an AUC-ROC of 0.9828 on 763 held-out versions. A secondary pooled analysis over both ecosystems is reported: the class priors differ by nearly a factor of nine, a classifier using only the ecosystem of origin attains an AUC-ROC of 0.8036, and the pooled figures should be read with that composition in mind. Confidence intervals are obtained by cluster-bootstrap resampling at the composite-group level and are 3.7-times wider than version-level intervals in AU-PR (0.0201 against 0.0055); at that scale, none of the differences between feature subsets, selection-stage orderings, or metadata families reported here is distinguishable from sampling variation. At an operational prevalence of 0.1%, the positive predictive value is 4.3% and recall at the selected threshold is 0.537, indicating that metadata-based detection functions as a triage filter rather than a definitive verdict.
- 10 September 2026
- Published by MDPI AG in Journal of Cybersecurity and Privacy
- Vol. 6 (5) , 160
Open AccessMetadata Compressibility and Evaluation Bias in Malicious Package Detection for NPM and PyPI
Malicious packages in open source-software supply chains are a growing security concern, yet machine learning detectors built on registry metadata are difficult to interpret and are typically evaluated under protocols susceptible to data leakage. We construct a dataset of 3330 package versions from NPM and PyPI in which every attribute is reconstructed at its exact publication timestamp, and we partition the data so that all versions of a package, all packages of a maintainer, and all members of a name-similarity family reside within a single fold. The primary analysis is restricted to NPM, where 18 of 456 point-in-time attributes attain an AU-PR of 0.9972 against a positive-class baseline of 0.8755 and an AUC-ROC of 0.9828 on 763 held-out versions. A secondary pooled analysis over both ecosystems is reported: the class priors differ by nearly a factor of nine, a classifier using only the ecosystem of origin attains an AUC-ROC of 0.8036, and the pooled figures should be read with that composition in mind. Confidence intervals are obtained by cluster-bootstrap resampling at the composite-group level and are 3.7-times wider than version-level intervals in AU-PR (0.0201 against 0.0055); at that scale, none of the differences between feature subsets, selection-stage orderings, or metadata families reported here is distinguishable from sampling variation. At an operational prevalence of 0.1%, the positive predictive value is 4.3% and recall at the selected threshold is 0.537, indicating that metadata-based detection functions as a triage filter rather than a definitive verdict.
- Open Access
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 72554-72564
Balancing Sensor Communication Performances With MQTT/CoAP and Data Security With TLS/DTLS in Invasive Cardiac Internet of Medical Things
Security is a critical concern in the Internet of Medical Things (IoMT), where protecting sensitive patient data and the integrity of device communication are paramount. Medical devices such as pacemakers and defibrillators must ensure the confidentiality and authenticity of the data while maintaining efficient communication in resource-constrained environments. Lightweight protocols such as MQTT (Message Queuing Telemetry Transport) and CoAP (Constrained Application Protocol) are widely used in IoMT to support secure and efficient data transmission. This paper evaluates the performance of MQTT and CoAP in terms of latency, efficiency, and energy consumption under varying packet sizes, considering the impact of security mechanisms such as Transport Layer Security (TLS) for MQTT and Datagram Transport Layer Security (DTLS) for CoAP. The results show that CoAP, with its lightweight UDP-based design, achieves 20 to 35% lower latency and 15 to 25% higher efficiency compared toMQTTin packet sizes ranging from 50 to 500 bytes. When security layers are added, CoAP-DTLS outperforms MQTT-TLS by consuming 20–30% less energy while maintaining comparable latency and efficiency. These findings suggest that CoAP-DTLS is the preferred protocol for energy-constrained IoMT devices, ensuring real-time performance and extended battery life while supporting robust security.
- 30 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 72554-72564
Open AccessBalancing Sensor Communication Performances With MQTT/CoAP and Data Security With TLS/DTLS in Invasive Cardiac Internet of Medical Things
Security is a critical concern in the Internet of Medical Things (IoMT), where protecting sensitive patient data and the integrity of device communication are paramount. Medical devices such as pacemakers and defibrillators must ensure the confidentiality and authenticity of the data while maintaining efficient communication in resource-constrained environments. Lightweight protocols such as MQTT (Message Queuing Telemetry Transport) and CoAP (Constrained Application Protocol) are widely used in IoMT to support secure and efficient data transmission. This paper evaluates the performance of MQTT and CoAP in terms of latency, efficiency, and energy consumption under varying packet sizes, considering the impact of security mechanisms such as Transport Layer Security (TLS) for MQTT and Datagram Transport Layer Security (DTLS) for CoAP. The results show that CoAP, with its lightweight UDP-based design, achieves 20 to 35% lower latency and 15 to 25% higher efficiency compared toMQTTin packet sizes ranging from 50 to 500 bytes. When security layers are added, CoAP-DTLS outperforms MQTT-TLS by consuming 20–30% less energy while maintaining comparable latency and efficiency. These findings suggest that CoAP-DTLS is the preferred protocol for energy-constrained IoMT devices, ensuring real-time performance and extended battery life while supporting robust security.
- Open Access
- 2 April 2026
- Published by Elsevier in SSRN Electronic Journal
Comparative Study of CoAP and QUIC-Lite: Toward Secure and Energy-Aware IoT Communication
The growing demand for secure and efficient communication in the Internet of Things (IoT)has highlighted the limitations of existing protocols when applied to constrained environ-ments. This paper explores the trade-offs between two protocols: QUIC, a modern, securetransport protocol originally designed for the web, and CoAP, a lightweight applicationlayer protocol tailored for low-power devices. While QUIC offers advanced features like mul-tiplexed streams and encryption by default, its complexity can overwhelm resource-limitedsystems. In contrast, CoAP is simpler but lacks robust security and transport layer reliabil-ity.To bridge this gap, we introduce QUIC-Lite, a streamlined transport layer protocol thatborrows the core strengths of QUIC secure handshake, reliable delivery, stream multiplexingwhile adopting the simplicity and minimal footprint of CoAP. QUIC-Lite is designed witha reduced handshake mechanism based on pre-shared keys and implements adaptive flowcontrol without introducing excessive overhead.A comparative benchmark between CoAP and QUIC-Lite was conducted through net-work simulation, focusing on four key dimensions: energy consumption, throughput, pro-cessing time, and security resilience. Results show that QUIC-Lite achieves faster sessionestablishment (2–3 RTTs), higher throughput ( 26.7 kbps), and improved energy efficiency(up to 40 bytes/mJ) while maintaining stable power consumption across CPU, TX, andRX. Processing time remains low due to protocol simplifications, and security remains ro-bust, with confidentiality and integrity preserved without relying on heavy cryptographicoperations.This work demonstrates that it is feasible to bring modern transport features to con-strained devices by rethinking protocol design. QUIC-Lite offers a balanced alternative,combining the strengths of QUIC and CoAP, and opens the door to scalable, secure, andenergy-aware communication in future IoT deployments.
- 2 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessComparative Study of CoAP and QUIC-Lite: Toward Secure and Energy-Aware IoT Communication
The growing demand for secure and efficient communication in the Internet of Things (IoT)has highlighted the limitations of existing protocols when applied to constrained environ-ments. This paper explores the trade-offs between two protocols: QUIC, a modern, securetransport protocol originally designed for the web, and CoAP, a lightweight applicationlayer protocol tailored for low-power devices. While QUIC offers advanced features like mul-tiplexed streams and encryption by default, its complexity can overwhelm resource-limitedsystems. In contrast, CoAP is simpler but lacks robust security and transport layer reliabil-ity.To bridge this gap, we introduce QUIC-Lite, a streamlined transport layer protocol thatborrows the core strengths of QUIC secure handshake, reliable delivery, stream multiplexingwhile adopting the simplicity and minimal footprint of CoAP. QUIC-Lite is designed witha reduced handshake mechanism based on pre-shared keys and implements adaptive flowcontrol without introducing excessive overhead.A comparative benchmark between CoAP and QUIC-Lite was conducted through net-work simulation, focusing on four key dimensions: energy consumption, throughput, pro-cessing time, and security resilience. Results show that QUIC-Lite achieves faster sessionestablishment (2–3 RTTs), higher throughput ( 26.7 kbps), and improved energy efficiency(up to 40 bytes/mJ) while maintaining stable power consumption across CPU, TX, andRX. Processing time remains low due to protocol simplifications, and security remains ro-bust, with confidentiality and integrity preserved without relying on heavy cryptographicoperations.This work demonstrates that it is feasible to bring modern transport features to con-strained devices by rethinking protocol design. QUIC-Lite offers a balanced alternative,combining the strengths of QUIC and CoAP, and opens the door to scalable, secure, andenergy-aware communication in future IoT deployments.
- 25 January 2026
- Published by Springer Nature
- p. 81-93
AI-Driven Cyberattack Detection and Prevention: A Systematic Review of Emerging Trends and Future Challenges
- 25 January 2026
- Published by Springer Nature
- p. 81-93
AI-Driven Cyberattack Detection and Prevention: A Systematic Review of Emerging Trends and Future Challenges
- Open Access
- 15 January 2026
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-4/W, 7-11
Advanced Drought Prediction Using Hybrid Deep Learning Models: A Case Study of the High Atlas and Anti-Atlas Mountains
Morocco’s High Atlas and Anti-Atlas mountains have faced escalating drought severity in recent years, jeopardizing water security and rural livelihoods. Conventional drought monitoring often underperforms in these regions due to sparse meteorological stations and rugged terrain. This study develops a hybrid deep learning framework for operational SPI drought prediction at 5 km resolution, synthesizing remote sensing and climate variables (SPI, NDVI, soil moisture, precipitation, temperature) from 1990–2024. 128 engineered features—rolling statistics, seasonality, lag dependencies, and cross-variable interactions—enhance learning. We benchmark three recurrent neural network types (LSTM, Bi-LSTM, GRU), validated with held-out data (2021–2024). The GRU model achieved the highest predictive skill, reaching 91.89% accuracy within a ±0.2 SPI threshold and outperforming baselines (Random Forest, ARIMA). Our results demonstrate the value of advanced feature engineering and deep sequence learning for month-ahead drought early warning in semi-arid North Africa.
- 15 January 2026
- Published by Copernicus GmbH in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
- Vol. XLVIII-4/W, 7-11
Open AccessAdvanced Drought Prediction Using Hybrid Deep Learning Models: A Case Study of the High Atlas and Anti-Atlas Mountains
Morocco’s High Atlas and Anti-Atlas mountains have faced escalating drought severity in recent years, jeopardizing water security and rural livelihoods. Conventional drought monitoring often underperforms in these regions due to sparse meteorological stations and rugged terrain. This study develops a hybrid deep learning framework for operational SPI drought prediction at 5 km resolution, synthesizing remote sensing and climate variables (SPI, NDVI, soil moisture, precipitation, temperature) from 1990–2024. 128 engineered features—rolling statistics, seasonality, lag dependencies, and cross-variable interactions—enhance learning. We benchmark three recurrent neural network types (LSTM, Bi-LSTM, GRU), validated with held-out data (2021–2024). The GRU model achieved the highest predictive skill, reaching 91.89% accuracy within a ±0.2 SPI threshold and outperforming baselines (Random Forest, ARIMA). Our results demonstrate the value of advanced feature engineering and deep sequence learning for month-ahead drought early warning in semi-arid North Africa.
- 30 August 2025
- Published by Springer Nature
- p. 583-594
Enhancing Maternal Health Risk Prediction Through Advanced Boosting Methods
- 30 August 2025
- Published by Springer Nature
- p. 583-594
Enhancing Maternal Health Risk Prediction Through Advanced Boosting Methods
- Open Access
- 20 August 2025
- Published by Elsevier in Procedia Computer Science
- Vol. 265, 236-243
A New Version of RPL Protocol to Maximize Lifetime in MIoT Networks
- 20 August 2025
- Published by Elsevier in Procedia Computer Science
- Vol. 265, 236-243
Open AccessA New Version of RPL Protocol to Maximize Lifetime in MIoT Networks
- 6 November 2024
- Published by Taylor & Francis
- p. 30-51
Challenges and countermeasures for using machine learning and artificial intelligence in blockchain and IoT applications
The integration of machine learning (ML) and artificial intelligence (AI) with blockchain and Internet of Things (IoT) technologies presents both significant opportunities and formidable challenges. This chapter explores the primary obstacles and proposes effective countermeasures in the convergence of these cutting-edge domains. Key challenges include scalability issues, data privacy concerns, computational resource constraints, and the complexities of ensuring interoperability among diverse systems. To address these, we examine advanced ML algorithms and AI techniques tailored for optimizing blockchain efficiency, enhancing IoT data security, and improving overall system robustness. Furthermore, we discuss novel frameworks and architectures that leverage AI-driven decision-making for dynamic adaptation in IoT networks and secure data transactions in blockchain platforms. Through comprehensive analysis and case studies, this chapter provides valuable insights and practical solutions aimed at overcoming the current limitations, thereby paving the way for more resilient and intelligent blockchain and IoT ecosystems.
- 6 November 2024
- Published by Taylor & Francis
- p. 30-51
Challenges and countermeasures for using machine learning and artificial intelligence in blockchain and IoT applications
The integration of machine learning (ML) and artificial intelligence (AI) with blockchain and Internet of Things (IoT) technologies presents both significant opportunities and formidable challenges. This chapter explores the primary obstacles and proposes effective countermeasures in the convergence of these cutting-edge domains. Key challenges include scalability issues, data privacy concerns, computational resource constraints, and the complexities of ensuring interoperability among diverse systems. To address these, we examine advanced ML algorithms and AI techniques tailored for optimizing blockchain efficiency, enhancing IoT data security, and improving overall system robustness. Furthermore, we discuss novel frameworks and architectures that leverage AI-driven decision-making for dynamic adaptation in IoT networks and secure data transactions in blockchain platforms. Through comprehensive analysis and case studies, this chapter provides valuable insights and practical solutions aimed at overcoming the current limitations, thereby paving the way for more resilient and intelligent blockchain and IoT ecosystems.
- Open Access
- 22 May 2024
- Published by International Association of Online Engineering (IAOE) in International Journal of Interactive Mobile Technologies (iJIM)
- Vol. 18 (10) , 90-105
Optimal Method Combining Blockchain and Multi-Agent System to Ensure Data Integrity and Deduplication in the Cloud Environment
Cloud computing (CC) refers to the transmission, storage, and processing of any type of information at a location that is not owned or controlled by the information owner. This information can be stored and accessed anytime, from anywhere, and using any device. However, several challenges threaten the deployment of cloud computing technology within organizations, including user privacy, data integrity, and managing the retention of the vast amount of data that will be stored and analyzed in this environment. In recent years, various data integrity checking systems and data storage mechanisms have been proposed to address these diverse challenges. Nevertheless, some protocols do not guarantee data confidentiality against auditors during the auditing process, while others do not provide batch auditing. Additionally, the majority of these protocols do not support the deduplication technique. Therefore, in this research work, the security features of blockchain technology and those of multi-agent systems (MAS) have prompted us to propose models that enable both data deduplication, saving storage space, and data integrity auditing, efficiently verifying the accuracy of data outsourced to CC environments. Implementation findings indicate that the suggested system is highly feasible and applicable in real-world scenarios.
- 22 May 2024
- Published by International Association of Online Engineering (IAOE) in International Journal of Interactive Mobile Technologies (iJIM)
- Vol. 18 (10) , 90-105
Open AccessOptimal Method Combining Blockchain and Multi-Agent System to Ensure Data Integrity and Deduplication in the Cloud Environment
Cloud computing (CC) refers to the transmission, storage, and processing of any type of information at a location that is not owned or controlled by the information owner. This information can be stored and accessed anytime, from anywhere, and using any device. However, several challenges threaten the deployment of cloud computing technology within organizations, including user privacy, data integrity, and managing the retention of the vast amount of data that will be stored and analyzed in this environment. In recent years, various data integrity checking systems and data storage mechanisms have been proposed to address these diverse challenges. Nevertheless, some protocols do not guarantee data confidentiality against auditors during the auditing process, while others do not provide batch auditing. Additionally, the majority of these protocols do not support the deduplication technique. Therefore, in this research work, the security features of blockchain technology and those of multi-agent systems (MAS) have prompted us to propose models that enable both data deduplication, saving storage space, and data integrity auditing, efficiently verifying the accuracy of data outsourced to CC environments. Implementation findings indicate that the suggested system is highly feasible and applicable in real-world scenarios.