AI agent security
Indirect prompt injection, data exfiltration, privilege escalation in function-calling loops.
Applied artificial intelligence research across polytechnical domains
Affiliated with the Fablab of the École Nationale Polytechnique d'Oran and the Computer Science department of USTO-MB, Recherchily conducts research in AI safety, multi-agent architectures, and industrial applications.
Explore our work →A joint initiative
ENP OranUSTO-MBFablab
AIAI Safety
PRPrompt Injection
MUMulti-Agent
DEDeep Learning
NLNLP
COComputer Vision
BEBenchmarking
LLLLM Security
RORobotics
FOFormal Verification
PRPredictive Maintenance
OPOpen Source
About
Recherchily is an interdisciplinary research laboratory founded within the Fablab of the École Nationale Polytechnique d'Oran (ENP Oran), in close collaboration with the Computer Science department of the University of Science and Technology of Oran — Mohamed Boudiaf (USTO-MB).
Our mission
Advance fundamental and applied research in artificial intelligence, targeting security vulnerabilities in autonomous systems, multi-tool agent architectures, and rigorous evaluation methods — grounded in Algeria's industrial and societal needs.
Our vision
Become a reference hub in North Africa for AI safety research, training the next generation of researchers and producing open-source benchmarks and tools adopted by the international community.
Our values
Scientific rigor
Reproducible methodologies, verifiable results, peer-reviewed publications.
Open source
All our datasets, benchmarks, and tools are released under open licenses.
Local impact
Training Algerian talent and solving real-world problems on the ground.
Collaboration
Bridges between IS, computer science, industrial engineering departments, and international partners.
Research domains
Multi-agent architectures
Orchestration, coordination, and formal verification of autonomous multi-agent systems.
AI for industry
Predictive maintenance, supply chain optimization, quality control through computer vision.
Natural language processing
Language models for Algerian Arabic (Darja), information extraction, and automatic summarization.
Evaluation & benchmarking
Robustness metrics, evaluation frameworks, reproducibility of AI results.
AI & law
Automated legal assistance, legal document analysis, algorithmic compliance.
Research team
Researchers and engineers at Recherchily lab
AM
Abdelhamid MASSERITI
Research Director
Associate Professor, IS Department — ENP Oran
Head of the Recherchily laboratory and the ENP Oran Fablab. His work focuses on the security of autonomous AI systems, indirect prompt injection, and multi-agent architectures. He supervises the lab's research projects and coordinates collaborations with USTO-MB.
AI SafetyAutonomous agentsPrompt injection
FB
Farid BENHAMMADI
Researcher
Associate Professor, Computer Science Department — USTO-MB
Specialist in natural language processing and language models for Arabic. He leads the lab's work on multilingual information extraction and contributes to LLM evaluation benchmarks on North African dialects.
NLPLanguage modelsDialectal Arabic
NR
Nadia RAHMANI
Researcher
Assistant Professor, IS Department — ENP Oran
Her research focuses on AI applied to industry: predictive maintenance through deep learning, anomaly detection in manufacturing processes, and supply chain optimization through evolutionary algorithms.
Industrial AIDeep learningPredictive maintenance
KT
Karim TOUATI
Researcher
Associate Professor, Computer Science Department — USTO-MB
Expert in multi-agent systems and formal verification. He works on coordination and communication methods between autonomous agents, with applications in collaborative robotics and complex systems simulation.
Multi-agent systemsFormal verificationRobotics
Research projects
Ongoing and completed work
ActiveJune 2026
Benchmarking Indirect Prompt Injection Vulnerabilities and Mitigation Strategies in Tool-Calling Agent Trajectories
As LLM applications transition from isolated chat interfaces to autonomous agents with tool access (web scraping, code execution, database queries), indirect prompt injection poses a severe safety ris...
View project →ActiveSeptember 2025
Formal Verification of Open-Source Software Packages via Hybrid Static Analysis and Machine Learning Classification
Open-source package registries (PyPI, npm) have become major attack vectors through malicious code injection in transitive dependencies. This project develops a hybrid pipeline combining static data-f...
View project →ActiveJanuary 2026
Automatic Control of Industrial Processes via Deep Reinforcement Learning with Lyapunov Stability Guarantees
Classical PID controllers struggle to adapt to the nonlinearities and variable disturbances of real industrial processes. This project proposes a control architecture based on deep reinforcement learn...
View project →ActiveMarch 2026
Computer Vision for Structural Health Monitoring in Civil Engineering: Automatic Crack Detection and Quantification via Deep Segmentation Networks
Manual visual inspection of concrete infrastructure (bridges, buildings, dams) is costly, subjective, and hazardous. This project develops an automatic crack detection and quantification system from d...
View project →CompletedOctober 2025
Natural Language Processing for Algerian Legal Document Analysis and Automated Regulatory Compliance Checking
The Algerian legal corpus (Official Journal, codes, decrees) is voluminous, multilingual (Arabic/French), and poorly digitally structured. This project develops a complete NLP pipeline for legal entit...
View project →ActiveApril 2026
Energy-Efficient Scheduling in Heterogeneous Computing Clusters via Multi-Objective Meta-Heuristic Optimization
Data centers and heterogeneous computing clusters (CPU, GPU, FPGA) consume considerable energy with often suboptimal scheduling policies. This project formulates scheduling as a multi-objective optimi...
View project →CompletedJuly 2025
Adversarial Robustness of Federated Learning in IoT Edge Networks: Model Poisoning Attacks and Byzantine-Resilient Defense Mechanisms
Federated learning on IoT edge devices is vulnerable to model poisoning and data poisoning attacks by compromised nodes. This project systematically evaluates the robustness of FedAvg and FedProx agai...
View project →Publications
Papers, technical reports, and preprints
ClusterGuard: Robust Gradient Clustering Aggregation for Federated Learning on IoT Devices under Byzantine Attacks
AAAI 2026 — Workshop on Federated Learning for IoT
aaai2026-fl-iot-042
Toward a Systematic Benchmark of Indirect Prompt Injection Vulnerabilities in Tool-Calling Autonomous Agents
arXiv preprint
arXiv:2607.09412
JORADP-NER: An Annotated Corpus for Named Entity Recognition in Algerian Legal Texts
Journal of Artificial Intelligence Research (JAIR)
doi:10.1613/jair.1.14832
Industrial Furnace Temperature Control via Soft Actor-Critic under Lyapunov Stability Constraints
arXiv preprint
arXiv:2608.03217
Hybrid Static Analysis and GNN Classification for Malicious Package Detection in Open-Source Registries
ICSE 2026 — New Ideas and Emerging Results Track
icse2026-nier-118
Energy-Efficient Multi-Objective Scheduling in Heterogeneous CPU-GPU Clusters: A Hybrid NSGA-III Approach
IEEE Access
doi:10.1109/ACCESS.2025.3401782
DZ-CrackSeg: A Crack Segmentation Dataset on Concrete Infrastructure in Algeria
arXiv preprint
arXiv:2609.01544
Evaluating LLM Robustness Against Multi-Turn Adversarial Injections: Experimental Protocol and Preliminary Results
NeurIPS 2025 — Workshop on AI Safety
neurips2025-aisafety-067
Frequently asked questions
How can I join the lab?
We recruit Master's and PhD students from the IS department at ENP Oran and the Computer Science department at USTO-MB. Send your CV and cover letter to contact@recherchily.com.
Do you offer research internships?
Yes. We host Master 2 interns and PhD candidates for 3 to 6-month stays on our ongoing projects. Applications are open year-round.
Is your work open access?
Yes. All our datasets, benchmarks, tools, and papers are published under open licenses (MIT/Apache 2.0) and hosted on GitHub and Hugging Face.
Can we collaborate from another university?
Absolutely. We actively collaborate with national and international laboratories. Contact us to discuss a joint project.
Have a question? Write to contact@recherchily.com
Contact
Collaborations, internships, and inquiries
Location
Fablab — École Nationale Polytechnique d'Oran
BP 1523 El M'Naouer, Oran 31000, Algeria
Department of Computer Science — USTO-MB
BP 1505 El M'Naouer, Oran 31000, Algeria
Collaborate with us
We welcome academic and industrial collaboration proposals, research internship applications (Master and PhD level), and partnerships with international laboratories.
AIAI Safety
PRPrompt Injection
MUMulti-Agent