Resume | Lorenzo Pappone

2021 - now
Ph.D. Student

Saint Louis University, St. Louis, USA, MO
- Machine Learning for Networks
2021
M.S. in Computer Engineering

University of Naples Federico II, Naples, Italy
- Highlights
  - Machine Learning
  - Deep Learning
  - Network Security
  - Internet Performance Analysis
2018
B.S. in Computer Engineering

University of Naples Federico II, Naples, Italy
- Highlights
  - Algorithms
  - Computer Architectures
  - Database
  - Operative Systems

Aug 2021 - now
Graduate Research Assistant

Saint Louis University, St. Louis, MO, USA
- Conduct research on Deep Reinforcement Learning for context-aware end-to-end congestion control.
- Analysis and development of AI-based computer vision algorithms for network telemetry.
May-Sep 2023
Visiting Researcher

Boston University, Boston, MA, USA
- Collaborated on a CNN-based Transfer-Learning model for network intrusion detection using Generative Adversarial Networks and XAI.
May-Aug 2022
Visiting Researcher

KTH Royal Institute of Technology, Stockholm, Sweden
- Collaborated on deep learning-based anomaly detection models for congestion control to enhance network performance and user experience.
Apr-Aug 2021
Big Data Engineer

Almaviva DigitalTec, Naples, Italy
- Design and development of a back-end Spark job for a big data management platform. The back-end job supports SQL-like operations over geo-spatial data.
Sep '20 - Mar '21
Graduate Research Assistant

University of Naples Federico II, Naples, Italy
- Analysis and implementation of Deep Neural Networks (CNN, GRU, LSTM) to predict mobile-app coarse-grained network traffic in short-term time windows.

Software Engineering
- Leading the design and the development of a NLP-based web-application for context-aware spell-checking based on Flask and Javascript.
Internet Performance Analysis
- Conduct research for mobile-network traffic classification using a distributed deep learning framework (Horovod) on a Spark cluster.
Machine Learning
- Design and implementation of a predictor for a solar panel energy production forecasting on the basis of weather conditions, training data using Random Forest algorithm.
Network Security
- Generation and propagation of a Trojan malware using Metasploit and SEToolkit tools on a Kali Linux OS. The malware could avoid AV static and dynamic analysis and tested on a Windows OS, after being embedded in a Windows installer using the NSIS tool.
Datacenter Networks
- Implementation of the ECMP protocol using software defined networking on a leaf-spine topology, built using the Mininet tool. The control plane was deployed using a Ryu controller and the OpenFlow protocol.
Computer Systems
- Implementation and of a client-server application for workload analysis and load balancing of incoming HTTP requests. A server pool was deployed using Microsoft Azure. The communication has been tested using the Jmeter tool and each virtual machine has been provided with an Apache web server.

Transfer learning via Generative Adversarial Network for Intrusion Detection
- A meta-learning based approach to train Wasserstain Generative Adversarial Networks (WGANs) on few-shot attack detection dataset.
Network traffic matrix estimation using AI-based computer vision techniques
- Development an AI system based on computer vision algorithms (i.e., super-resolution) that can accurately estimate fine-grained network traffic matrix from aggregated measurements, which can provide valuable insights into network performance and help identify potential bottlenecks or security threats.
Routing using Deep Reinforcement Learning and Graph Convolutional Networks.
- Design, implementation, and evaluation of innovative approaches for network routing based on Deep Reinforcement Learning. Single-Agent Graph Convolutional Network could help to minimize retraining needs, whereas Multi-agent Deep Q-Networks could provide scalable alternatives where multiple federated domain routing agents need to cooperate, e.g., in eBGP or other Wide Area Networks under a single domain
Mobile Traffic Prediction using Multitask Deep Learning.
- Comparison and discussion of advanced DL-based prediction models, such as Convolutional Neural Networks, Gated Recurrent Unit, Long Short-Term Memory and their performance to predict highly dynamic coarse-grained encrypted mobile-app traffic using flow-level statistics.