Department of Computer Science · Gulf University for Science and Technology
  1. Home
  2. Research

Research

The group studies resource allocation, service orchestration and security in edge computing and IoT systems. This page describes the research areas, funded projects, patents and methods.

Areas

Research areas

01

Edge computing and scheduling

Multi-access edge computing (MEC) can serve latency-sensitive applications only if tasks are placed and scheduled appropriately. This work develops schedulers for heterogeneous machine-learning pipelines subject to timing, synchronization and communication constraints, and fault-tolerant execution based on virtual task duplication.

MEC5GML pipelinesFault tolerance
Current work
  • Ahead-of-time scheduling of ML pipelines on edge systems
  • Fault-tolerant edge execution (patent application with Ericsson)
  • Task scheduling and synchronization for 5G MEC (patent application with Ericsson)
02

IoT orchestration and UAV systems

Mobile ad hoc networks (MANETs) have no fixed infrastructure, so services must be composed from devices whose position and reliability change over time. This work composes IoT services that account for spatial context and sensing reliability, and coordinates UAV fleets to collect data from distributed sensors.

MANETIoTUAVsService composition
Current work
  • Spatial context-aware service composition for MANET IoT
  • Joint sensing reliability and resource scheduling
  • Collaborative UAV networks for IoT data acquisition

Project page

03

Autonomous IoT security

Most IoT security mechanisms protect the confidentiality and integrity of data rather than the behavior of the device. This work models each device's legitimate behavior as an automaton, updates the model at runtime and scores states that are becoming vulnerable, so that threats are detected before they disrupt service. A reinforcement-learning controller balances security against availability.

Automata theoryIoT securityReinforcement learningCritical infrastructure
Current work
  • Automaton-based online vulnerability analysis and detection
  • RL-based adaptive control of the security–availability trade-off
  • Laboratory testbed of programmable IoT nodes, with the University of Ottawa

Project page

04

Internet of Vehicles and intelligent transportation

Connected and autonomous vehicles generate and consume large volumes of time-sensitive data. This work optimizes workload scheduling in vehicular fog networks and information freshness (Age of Information) for cooperative driving, including RIS-assisted links. It also addresses EV charging and routing, and the scheduling of urban patrol vehicles.

V2XAge of InformationRISElectric vehicles
Current work
  • Age of Information in MEC-enabled cooperative driving
  • Workload scheduling in vehicular edge and fog networks
  • EV charging, routing and V2V energy transfer
  • Vehicle patrol scheduling for urban safety
05

AI for health

This work applies graph neural networks to multimodal clinical signals. Gait dynamics and EEG activity are represented as graphs to support earlier and more accurate detection of Parkinson's disease and dementia.

Graph neural networksEEGGait analysis
Current work
  • GNN-based diagnosis from gait and EEG data
  • In collaboration with the University of Southern California
Funding

Funded projects

Supported by the Kuwait Foundation for the Advancement of Sciences (KFAS) and the Gulf University for Science and Technology.

Kuwait Foundation for the Advancement of Sciences (KFAS)
KFAS Applied Research Grant 2026 – 2028 Sep 2026 – Sep 2028
New

Towards Autonomous IoT Security: Automaton-Based Online Vulnerability Analysis and Detection

Under Kuwait Vision 2035, IoT devices are being deployed in healthcare, energy, education and oil production, which makes threats to device functionality as important as threats to data. The project assigns each IoT device an automaton describing its legitimate behavior. The automaton is relabeled at runtime with vulnerability scores derived from observations of the environment and the system, so that threats can be detected and contained before a service is disrupted or a device is compromised. A reinforcement-learning agent sets how restrictive the system is, trading security against availability.

Expected outputs are a prototype validated on a laboratory testbed, an open reference implementation and datasets.

Principal Investigator: Dr. Ebrahim Sarkhouh · Collaborators: Dr. Paria Shirani (University of Ottawa), Dr. Reem Albaghli (Kuwait University) · Junior Researcher: Lina Altinawi

Project page
KFAS Applied Research Grant 2026 – 2027 Sep 2026 – Sep 2027
New

Spatial Context-Aware Service Composition for MANET IoT Applications

This project develops a context-aware orchestration framework that combines perception, actuation and edge-computing components into IoT applications running over mobile ad hoc networks. It has three components:

  • a probabilistic spatio-temporal perception model that jointly assesses the sensing reliability of different devices;
  • machine-learning predictors of changes in wireless conditions, edge-server load and device availability;
  • optimization and metaheuristic service-composition algorithms (column generation, ACO and PSO) that divide each request into finer sub-requests to adapt in real time.

The prototype will be evaluated in simulated environments representative of industrial fields and healthcare infrastructure in Kuwait.

Principal Investigator: Dr. Ebrahim Sarkhouh · International Collaborator: Prof. Muthucumaru Maheswaran (McGill University) · Junior Researchers: Sulaiman Althuwaikh, Mohammad Kamali

Project page
GUST Seed Grant 2025 – 2026 Jun 2025 – Jun 2026
Seed grant

A Framework for Adaptive UAV-Based Data Collection and Resource Allocation in Dynamic Distributed Systems

An adaptive framework in which UAVs collect data from distributed sensors, with flight paths, communication and resource allocation optimized jointly as network conditions change.

Principal Investigator: Dr. Ebrahim Sarkhouh · International Co-Investigator: Prof. Muthucumaru Maheswaran (McGill University)

GUST Seed Grant 2025 – 2026 Jun 2025 – Jun 2026
Seed grant

Diagnosing Parkinson's Disease and Dementia Using Gait and EEG Data with Graph Neural Networks

Graph neural networks applied to gait and EEG recordings to support earlier and more accurate diagnosis of neurodegenerative disease.

Principal Investigator: Dr. Ebrahim Sarkhouh · International Co-Investigator: Dr. Luciano Nocera (University of Southern California)

Intellectual property

Patents

PCT/IB2024/052853

Fault-Tolerant Methods for Edge Task Execution via Virtual Duplication

E. Sarkhouh, M. Maheswaran, E. Fapi Thepie, Z. Zhu. Filed March 2024, with Ericsson.

PCT/IB2022/062009

Systems, Methods and Means for Task Scheduling and Synchronization in 5G Mobile Edge Computing Applications

E. Sarkhouh, M. Maheswaran, E. Fapi Thepie, Z. Zhu. Filed December 2022, with Ericsson.

Approach

Methods

Discrete optimization

Integer linear programming, column generation and Dantzig–Wolfe decomposition, Benders decomposition, Lagrangian relaxation and dynamic programming.

Heuristics and algorithms

Greedy and approximation algorithms, ant colony and particle swarm optimization, genetic algorithms, simulated annealing, automata-based modeling and event-driven stochastic simulation.

Machine learning

Deep learning, graph and sequential neural networks, reinforcement learning, and classical models such as support vector machines, ensembles and random forests.