Login| KEYWORD |
Generalizable language models for network measurement and cybersecurity
Parole chiave AI, CYBERSECURITY, FIREWALL, LLM, MACHINE LEARNING
Riferimenti MARCO MELLIA, LUCA VASSIO
Riferimenti esterni Idilio Drago - Unito - idilio.drago@unito.it
Gruppi di ricerca DATABASE AND DATA MINING GROUP - DBDM, SmartData@PoliTO, Telecommunication Networks Group
Descrizione Firewall/IPS and EDR and Cloud security services analyze huge amounts of structured data to detect and classify threats, mainly based on human-written rules.
The thesis goal is to understand if lightweight and generalizable language models can extract insights from raw data. A key objective is to ensure generalization abilities beyond syntactic heuristics. A possible solution is to create multi-modal embeddings to conceptually constrain the embeddings towards the right task.
Thesis Goal
- Propose techniques based on language models for identify network traffic threats
- Ensure generalization beyond simple rule by crafting proper training and validation data
- Use techniques based on multi-modal embeddings (similar to OpenAI CLIP)
Conoscenze richieste - Good programming skills (such as Python and Spark)
- Machine Learning knowledge (such as Torch, Tensorflow)
- Basics of NLP
- Basics of Networking and security
Note A GPA of at least 27/30 is preferred.
Possible graduation prize of 2000 euros.
Scadenza validita proposta 14/01/2026
PROPONI LA TUA CANDIDATURA