Slide
Centre Interdisciplinaire
de Recherche et d’Innovation
en Cybersécurité et Société
de Recherche et d’Innovation
en Cybersécurité et Société
1.
Allaoui, M. L.; Allili, M. S.
LG-TargetSeg: Deep semantic fusion via language-conditioned selective scanning for targeted medical image segmentation Journal Article
In: Information Fusion, vol. 139, 2027, ISSN: 15662535 (ISSN).
Abstract | Links | BibTeX | Tags: Computerized tomography, decoding, Deep vision-language fusion, Encoder-decoder, Gaussians, Medical image processing, Medical image segmentation, Modeling languages, Natural language processing systems, Reinforced gaussian dice loss, Selective scanning, Semantic fusion, Semantic segmentation, Semantics, Shallow levels, Signal encoding, State space methods, State-space modeling, State-space models, Training time
@article{allaouiLGTargetSegDeepSemantic2027,
title = {LG-TargetSeg: Deep semantic fusion via language-conditioned selective scanning for targeted medical image segmentation},
author = {M. L. Allaoui and M. S. Allili},
url = {https://www.scopus.com/pages/publications/105051242411?origin=resultslist},
doi = {10.1016/j.inffus.2026.104797},
issn = {15662535 (ISSN)},
year = {2027},
date = {2027-01-01},
journal = {Information Fusion},
volume = {139},
publisher = {Elsevier B.V.},
abstract = {Clinical language defines the semantic intent of targeted medical image segmentation, yet most vision–language methods incorporate it only at a shallow level or through training-time conditioning, leaving spatial reasoning largely language-blind. We propose LG-TargetSeg (Language-Guided Targeted Segmentation), a cross-modal framework that embeds clinical language directly into the selective scan of a Mamba-based encoder-decoder. Natural-language prompts, encoded by a frozen MedSigLIP foundation model, shape selective scanning throughout the encoder–decoder and additionally adapt the bottleneck state transition, alongside gated modulation and cross-attention, enabling semantic guidance to continuously shape feature propagation toward the target anatomy. A Text-Driven Selective Scan 2D (TD-SS2D) module achieves prompt-aware multi-scale fusion with linear complexity in the number of spatial tokens. To improve robustness to boundary ambiguity and class imbalance, we introduce the Reinforced Gaussian Dice Loss (RGDL) with learnable weighting parameters. Evaluated under both task-specific and universal generalization settings, LG-TargetSeg achieves state-of-the-art performance. Moreover, its prompt-driven selectivity remains reliable under genuine multi-target competition: on abdominal CT images containing up to 16 annotated organs, changing only the text prompt on a frozen model produces organ masks with less than 1% pairwise overlap. Additional materials and code are available at: https://github.com/MOHAMEDLamine77/LG-TargetSeg . © 2026 Elsevier B.V.},
keywords = {Computerized tomography, decoding, Deep vision-language fusion, Encoder-decoder, Gaussians, Medical image processing, Medical image segmentation, Modeling languages, Natural language processing systems, Reinforced gaussian dice loss, Selective scanning, Semantic fusion, Semantic segmentation, Semantics, Shallow levels, Signal encoding, State space methods, State-space modeling, State-space models, Training time},
pubstate = {published},
tppubtype = {article}
}
Clinical language defines the semantic intent of targeted medical image segmentation, yet most vision–language methods incorporate it only at a shallow level or through training-time conditioning, leaving spatial reasoning largely language-blind. We propose LG-TargetSeg (Language-Guided Targeted Segmentation), a cross-modal framework that embeds clinical language directly into the selective scan of a Mamba-based encoder-decoder. Natural-language prompts, encoded by a frozen MedSigLIP foundation model, shape selective scanning throughout the encoder–decoder and additionally adapt the bottleneck state transition, alongside gated modulation and cross-attention, enabling semantic guidance to continuously shape feature propagation toward the target anatomy. A Text-Driven Selective Scan 2D (TD-SS2D) module achieves prompt-aware multi-scale fusion with linear complexity in the number of spatial tokens. To improve robustness to boundary ambiguity and class imbalance, we introduce the Reinforced Gaussian Dice Loss (RGDL) with learnable weighting parameters. Evaluated under both task-specific and universal generalization settings, LG-TargetSeg achieves state-of-the-art performance. Moreover, its prompt-driven selectivity remains reliable under genuine multi-target competition: on abdominal CT images containing up to 16 annotated organs, changing only the text prompt on a frozen model produces organ masks with less than 1% pairwise overlap. Additional materials and code are available at: https://github.com/MOHAMEDLamine77/LG-TargetSeg . © 2026 Elsevier B.V.
2.
Laamari, A.; Moudoud, H.; Houda, Z. A. El
Lightweight LLM Adaptation for Intrusion Detection via Token-Efficient Flow Representation Proceedings Article
In: IEEE Conf. Artif. Intell., CAI, pp. 2122–2127, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 979-833156039-3 (ISBN), (Journal Abbreviation: IEEE Conf. Artif. Intell., CAI).
Abstract | Links | BibTeX | Tags: Classification (of information), Data flow analysis, Decoder-only large language model, Decoder-only LLMs, decoding, Flow classification, Intrusion Detection, Intrusion-Detection, Language model, Large language model, LLMs, LoRA, Low-rank adaptation, Network Flow Classification, Network intrusion, Network security, Networks flows, Qwen2.5, Signal encoding, T5-Small, Token-oriented object notation, Tokenization, TOON
@inproceedings{laamariLightweightLLMAdaptation2026,
title = {Lightweight LLM Adaptation for Intrusion Detection via Token-Efficient Flow Representation},
author = {A. Laamari and H. Moudoud and Z. A. El Houda},
url = {https://www.scopus.com/pages/publications/105042133342?origin=resultslist},
doi = {10.1109/CAI68641.2026.11536533},
isbn = {979-833156039-3 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {IEEE Conf. Artif. Intell., CAI},
pages = {2122–2127},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Large language models (LLMs) are emerging as a promising approach for intrusion detection using structured network flow data. However, their practical deployment is constrained by context window limitations and the excessive token overhead introduced by conventional tabular serialization formats such as JSON. Verbose data representations inflate sequence lengths, often exceeding model input limits and causing feature truncation. Additionally, it remains unclear which LLM architecture is more suitable for structured intrusion detection tasks under limited training resources. To tackle this issue, we propose a novel representation-aware intrusion detection framework based on Token-Oriented Object Notation (TOON), a compact serialization format that maximizes token efficiency while preserving schema structure. Also, we integrate a Low-Rank Adaptation (LoRA) scheme to enable parameter-efficient fine-tuning. Finally, we evaluate the proposed framework as an encoder-decoder model (T5-Small) with a decoder-only model (Qwen2.5) on three benchmark datasets, including NSL-KDD, UNSW-NB15, and CIC-IDS2018 for binary and multi-class classification scenarios. The results show that our tokenizer-aligned, representation-aware preprocessing combined with lightweight encoder-decoder adaptation provides a practical and resource-efficient foundation for LLM-based intrusion detection. © 2026 IEEE.},
note = {Journal Abbreviation: IEEE Conf. Artif. Intell., CAI},
keywords = {Classification (of information), Data flow analysis, Decoder-only large language model, Decoder-only LLMs, decoding, Flow classification, Intrusion Detection, Intrusion-Detection, Language model, Large language model, LLMs, LoRA, Low-rank adaptation, Network Flow Classification, Network intrusion, Network security, Networks flows, Qwen2.5, Signal encoding, T5-Small, Token-oriented object notation, Tokenization, TOON},
pubstate = {published},
tppubtype = {inproceedings}
}
Large language models (LLMs) are emerging as a promising approach for intrusion detection using structured network flow data. However, their practical deployment is constrained by context window limitations and the excessive token overhead introduced by conventional tabular serialization formats such as JSON. Verbose data representations inflate sequence lengths, often exceeding model input limits and causing feature truncation. Additionally, it remains unclear which LLM architecture is more suitable for structured intrusion detection tasks under limited training resources. To tackle this issue, we propose a novel representation-aware intrusion detection framework based on Token-Oriented Object Notation (TOON), a compact serialization format that maximizes token efficiency while preserving schema structure. Also, we integrate a Low-Rank Adaptation (LoRA) scheme to enable parameter-efficient fine-tuning. Finally, we evaluate the proposed framework as an encoder-decoder model (T5-Small) with a decoder-only model (Qwen2.5) on three benchmark datasets, including NSL-KDD, UNSW-NB15, and CIC-IDS2018 for binary and multi-class classification scenarios. The results show that our tokenizer-aligned, representation-aware preprocessing combined with lightweight encoder-decoder adaptation provides a practical and resource-efficient foundation for LLM-based intrusion detection. © 2026 IEEE.
3.
Mehrban, A.; Moudoud, H.; Brik, B.; Houda, Z. A. El
Enhancing Spectral Efficiency and Resilience in SAGIN-ITS: An IRS-Assisted Semantic Offloading Framework Journal Article
In: IEEE Internet of Things Magazine, vol. 9, no. 4, pp. 30–36, 2026, ISSN: 25763180 (ISSN).
Abstract | Links | BibTeX | Tags: Air grounds, Bandwidth, Computation offloading, Federated learning, federated learning (FL), IEEE Standards, Integrated networks, Intelligent reflecting surface, Intelligent reflecting surfaces (IRS), Intelligent transportation system, Intelligent transportation systems, intelligent transportation systems (ITS), Intelligent vehicle highway systems, Learning systems, Reflecting surface, Semantic offloading, Semantics, Sensor networks, Sensors data, Signal encoding, Space-air-ground integrated network, space-air-ground integrated networks (SAGIN), Traffic congestion, Transmissions
@article{mehrbanEnhancingSpectralEfficiency2026,
title = {Enhancing Spectral Efficiency and Resilience in SAGIN-ITS: An IRS-Assisted Semantic Offloading Framework},
author = {A. Mehrban and H. Moudoud and B. Brik and Z. A. El Houda},
url = {https://www.scopus.com/pages/publications/105036681552?origin=resultslist},
doi = {10.1109/MIOT.2026.3676349},
issn = {25763180 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {IEEE Internet of Things Magazine},
volume = {9},
number = {4},
pages = {30–36},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {In dense Vehicle-to-Everything (V2X) scenarios within Space-Air-Ground Integrated Networks (SAGIN), connected autonomous cars (CAVs) generate a large volume of sensor data, which leads to significant bandwidth congestion. There is a risk to safety because of the poor Packet Success Rate (PSR) and dangerous transmission delays caused by this congestion. To address this issue, we introduce SemantIRS, a framework that integrates semantic offloading with Intelligent Reflecting Surface (IRS)-assisted links to achieve high network resilience, when compute resources permit and lower latency. SemantIRS deploys lightweight Large Language Models (LLMs) as encoders to compress massive sensor data at the edge of the SAGIN network before transmission through IRS-assisted links. Intent-Based Networking (IBN) and Federated Learning (FL) are also integrated to automate resilience policies and preserve distributed offloading privacy throughout SAGIN layers, respectively. We conduct our experiments using a 12-state hardware benchmark of three LLM encoders (Phi-2, Llama-2-7B-Chat, and Mistral-7B-Instruct-v0.2) on four compute tiers (TPU v6e-1, A100, L4, and Intel Xeon CPU). The measured inference latencies are then fed into an IEEE 802.11p airtime simulator. Our obtained results show that, in congested network conditions, SemantIRS improves the packet success rate from 1.7% to 66.4% while reducing transmission energy by 90%. © 2026 IEEE.},
keywords = {Air grounds, Bandwidth, Computation offloading, Federated learning, federated learning (FL), IEEE Standards, Integrated networks, Intelligent reflecting surface, Intelligent reflecting surfaces (IRS), Intelligent transportation system, Intelligent transportation systems, intelligent transportation systems (ITS), Intelligent vehicle highway systems, Learning systems, Reflecting surface, Semantic offloading, Semantics, Sensor networks, Sensors data, Signal encoding, Space-air-ground integrated network, space-air-ground integrated networks (SAGIN), Traffic congestion, Transmissions},
pubstate = {published},
tppubtype = {article}
}
In dense Vehicle-to-Everything (V2X) scenarios within Space-Air-Ground Integrated Networks (SAGIN), connected autonomous cars (CAVs) generate a large volume of sensor data, which leads to significant bandwidth congestion. There is a risk to safety because of the poor Packet Success Rate (PSR) and dangerous transmission delays caused by this congestion. To address this issue, we introduce SemantIRS, a framework that integrates semantic offloading with Intelligent Reflecting Surface (IRS)-assisted links to achieve high network resilience, when compute resources permit and lower latency. SemantIRS deploys lightweight Large Language Models (LLMs) as encoders to compress massive sensor data at the edge of the SAGIN network before transmission through IRS-assisted links. Intent-Based Networking (IBN) and Federated Learning (FL) are also integrated to automate resilience policies and preserve distributed offloading privacy throughout SAGIN layers, respectively. We conduct our experiments using a 12-state hardware benchmark of three LLM encoders (Phi-2, Llama-2-7B-Chat, and Mistral-7B-Instruct-v0.2) on four compute tiers (TPU v6e-1, A100, L4, and Intel Xeon CPU). The measured inference latencies are then fed into an IEEE 802.11p airtime simulator. Our obtained results show that, in congested network conditions, SemantIRS improves the packet success rate from 1.7% to 66.4% while reducing transmission energy by 90%. © 2026 IEEE.
4.
Messaoudi, H.; Belaid, A.; Allaoui, M. L.; Zetout, A.; Allili, M. S.; Tliba, S.; Salem, D. Ben; Conze, P. -H.
Efficient Embedding Network for 3D Brain Tumor Segmentation Journal Article
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 12658 LNCS, pp. 252–262, 2021, ISSN: 03029743, (ISBN: 9783030720834).
Abstract | Links | BibTeX | Tags: 3D medical image processing, Brain, Brain tumor segmentation, Classification networks, Convolutional neural networks, Deep learning, Embedding network, Image segmentation, Large dataset, Large datasets, Medical imaging, Natural images, Net networks, Semantic segmentation, Semantics, Signal encoding, Tumors
@article{messaoudi_efficient_2021,
title = {Efficient Embedding Network for 3D Brain Tumor Segmentation},
author = {H. Messaoudi and A. Belaid and M. L. Allaoui and A. Zetout and M. S. Allili and S. Tliba and D. Ben Salem and P. -H. Conze},
editor = {Bakas S. Crimi A.},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85107387134&doi=10.1007%2f978-3-030-72084-1_23&partnerID=40&md5=b3aa3516b0465a1bf5611db4727d95f1},
doi = {10.1007/978-3-030-72084-1_23},
issn = {03029743},
year = {2021},
date = {2021-01-01},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
volume = {12658 LNCS},
pages = {252–262},
publisher = {Springer Science and Business Media Deutschland GmbH},
abstract = {3D medical image processing with deep learning greatly suffers from a lack of data. Thus, studies carried out in this field are limited compared to works related to 2D natural image analysis, where very large datasets exist. As a result, powerful and efficient 2D convolutional neural networks have been developed and trained. In this paper, we investigate a way to transfer the performance of a two-dimensional classification network for the purpose of three-dimensional semantic segmentation of brain tumors. We propose an asymmetric U-Net network by incorporating the EfficientNet model as part of the encoding branch. As the input data is in 3D, the first layers of the encoder are devoted to the reduction of the third dimension in order to fit the input of the EfficientNet network. Experimental results on validation and test data from the BraTS 2020 challenge demonstrate that the proposed method achieve promising performance. © 2021, Springer Nature Switzerland AG.},
note = {ISBN: 9783030720834},
keywords = {3D medical image processing, Brain, Brain tumor segmentation, Classification networks, Convolutional neural networks, Deep learning, Embedding network, Image segmentation, Large dataset, Large datasets, Medical imaging, Natural images, Net networks, Semantic segmentation, Semantics, Signal encoding, Tumors},
pubstate = {published},
tppubtype = {article}
}
3D medical image processing with deep learning greatly suffers from a lack of data. Thus, studies carried out in this field are limited compared to works related to 2D natural image analysis, where very large datasets exist. As a result, powerful and efficient 2D convolutional neural networks have been developed and trained. In this paper, we investigate a way to transfer the performance of a two-dimensional classification network for the purpose of three-dimensional semantic segmentation of brain tumors. We propose an asymmetric U-Net network by incorporating the EfficientNet model as part of the encoding branch. As the input data is in 3D, the first layers of the encoder are devoted to the reduction of the third dimension in order to fit the input of the EfficientNet network. Experimental results on validation and test data from the BraTS 2020 challenge demonstrate that the proposed method achieve promising performance. © 2021, Springer Nature Switzerland AG.