👋 Hi, I’m Matteo Boffa, a PhD student at Politecnico di Torino and a member of the SmartData group.
As a Machine and Deep Learning enthusiast, in my research, I investigate the adoption of Natural Language Processing (NLP) in cybersecurity.
You can find more details in the following papers:
- Towards NLP-based processing of honeypot logs: Leveraging NLP-based techniques to represent attacking sessions. Instrumental tool for the security experts. The paper was presented at WTMC, a workshop co-located with the IEEE European Symposium on Security and Privacy.
- On using pretext tasks to learn representations from network logs: Assessing network logs’ learned representation through auxiliary (pretext) tasks. Unsupervised application on logs collected from SSH honeypots. The paper was presented at NativeNI, a workshop co-located with ACM CoNEXT 2022.
- LogPrécis: Unleashing language models for automated malicious log analysis: Leveraging Language Models (e.g., BERT, CodeBERT, GPT) for a tool that supports the analysts better understanding attacks, detecting novelty, linking similar attacks and tracking families and mutations. Named Entity Recognition and Entity Extraction. Published on Computer & Security (COSE).
I have also worked on optimization problems, data mining, and analysis in my studies. For more information, please refer to:
- Neural combinatorial optimization beyond the TSP: Existing architectures under-represent graph structure: Graph Neural Network applied to network combinatorial problems (power and channel allocation) on a Reinforcement Learning framework. Results in the field of Representation Learning. Presented at GCLR, a workshop co-located with AAAI
- The polynomial robust knapsack problem: Operational research. Exact solution, Machine Learning and Genetic algorithms heuristics. Solves a portfolio optimization. > 5% margin from optimum. Published on EJOR - European Journal of Operational Research (Elsevier).
- Automatic identification of urban functions via social mining: Exploiting geo-referenced data from Social Networks for automatic characterization of urban functions. Data mining, clustering algorithms and transfer learning (e.g., InceptionV3). Published on Cities - The International Journal of Urban Policy and Planning (Elsevier).
Don't hesitate to contact me on my personal email or linkedin profile!