Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : LISIC – Université du Littoral
Durée : 1 year
Contact : esteban.bautista-ruiz@lip6.fr
Date limite de publication : 2025-10-31

Contexte :
The Industrial Internet of Things (IIoT) is a rapidly evolving paradigm in which industrial sensors, machines, and other instruments are connected to the internet. The data provided by these systems are of great value for automated surveillance and detecting anomalies in them is therefore a task of utmost importance.

Temporal graphs are a very effective model for IIoT data, where nodes represent devices; node features represent their measurements; and time-varying weighted edges capture various types of information. In the initial phase of this project, we have already developed powerful auto-encoder models capable of detecting both temporal and structural anomalies in temporal graphs. Yet, while our models allow us to detect the presence of an anomaly, they are still unable to localise it.

The main goal of this postdoctoral project is therefore to fundamentally extend our temporal graph auto-encoder models beyond anomaly detection to also perform anomaly localization.

Sujet :
The recruited postdoc will have two main goals:

1. Extension of Auto-Encoders for Anomaly Localization.

Our models currently receive a temporal graph as input and produce an anomaly score that is proportional to the severity of its abnormal events. Even though they can be used to localize anomalies by searching for the subset of maximal anomaly score, such an approach is impractical due to the exponential number of subsets to test. We aim to prune-down the search space by properly structuring the auto-encoder’s latent representation. This will require substantial architectural innovation to disentangle the auto-encoder’s structured representation and achieve fine-grained anomaly localization.

2. Application to IIoT logs and measurements.

We aim to evaluate the methodological developments above in real-world IIoT dataset that contain various types of attacks (structural anomalies) and measurement faults (feature anomalies). We aim to explore the advantages of our approach in real-world, potentially on-line, scenarios, such as machine health monitoring, transportation network monitoring, or other use cases that may arise from collaborations with local companies.

Profil du candidat :
We look for highly motivated candidates with relevant experience in anomaly detection, graph machine learning, and/or deep learning. Experience in Python programming, cybersecurity and/or streaming algorithms is a plus. Ideal candidates will have a publication record in selective AI conferences.

Interested candidates are invited to send a cover letter, a detailed CV (with a publication list and the contact details of two references), and their PhD manuscript or a recent paper to:

– esteban.bautista@univ-littoral.fr
– claire.guilloteau@univ-littoral.fr

Applications will be reviewed on a rolling basis until the position is filled.

Formation et compétences requises :

Adresse d’emploi :
Saint-Omer, France

Document attaché : 202509011756_Postdoc___Temporal_Graph_Auto_Encoders____Job_Offer.pdf