Bloom Filter Data Summarization: An Efficient Big-Data Approach to Discover Network-Flow Patterns based on Multidimensional Convolutional Neural Networks
Date : 23rd October 2023 at 10 AM
Place : Espace de convivialité du L2TI (2nd floor, Building E), 99 avenue Jean-Baptiste Clément, 93430 Villetaneuse (Trameway 11 or Trameway 8 – “Villetaneuse Université” station)
Abstract :
In this talk, we propose a data summarization technique based on Bloom Filter to aggregate network flow in a two-dimensional sliding window bitmap. After generating the summaries, we apply a deep learning approach that introduces convolutional neural network layers to segment the bitmap. We evaluate the proposal on a real dataset from a Brazilian broadband access provider, which accounts for access data of 373 residential users for one week. In the evaluated use case, the results demonstrate the efficiency of the proposed summarization and the high precision of the deep learning model in incrementally classifying the threat risk in the network.
Bio:
Diogo Menezes Ferrazani Mattos is a professor at Universidade Federal Fluminense (Niterói, Brazil). He received his degrees of D.Sc. and M.Sc. in Electrical Engineering from Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil, in 2017 and 2012. He received a Computer and Information Engineer degree from Universidade Federal do Rio de Janeiro, in 2010, awarded with Magna Cum Laude. Between 2015 and 2016, he had a sandwich scholarship to work on his Ph.D. thesis on the LIP6 (Laboratoire d’Informatique de Paris 6) at Sorbonne Université (Campus Pierre et Marie Curie), Paris, France. In 2017, he was a postdoctoral fellow in the Electrical Engineering Program at Universidade Federal do Rio de Janeiro. His publications and research interests cover network security, next-generation networks, virtualization, software-defined networking, and the Internet of the future. He coordinates research projects focused on providing secure primitives based on machine learning for next-generation networks and retrieving knowledge from social networks.