# Wolfram Summer SchoolFounded in 2003

17th Annual Wolfram Summer School, held at Bentley University June 23–July 12, 2019

Alumni

# Gerardo Lamastra

Class of 2005

## Bio

Gerardo Lamastra is Italian and 34 years old. He received his laurea degree in computer engineering in 1996 from the University of Pisa, and his Ph.D. in real-time computer systems from the Sant'Anna School of Advanced Studies in Pisa. He currently works for the R&D department of the main Italian Telecom Operator, in the field of internet security, specifically in intrusion detection and computer forensics.

## Project: Modeling Internet Malware Diffusion: An NKS Approach

Understanding and forecasting the dynamic of malware diffusion in the Internet is a challenging task. Most of the existing literature on this topic is based on traditional epidemiology, with several interesting contributions derived from complex networks analysis. This work is an attempt to build an elementary model for virus spreading, based on NKS.

A simple model for malware diffusion, based on a simple interleaved evolution of a general mobile automaton and an elementary CA, is analyzed to identify the virus-like diffusion patterns. There was also an interesting example found during experiments. The model offers a new way to analyze the virus diffusion phenomenon and provides an example that loosely resembles the evolution of a slow-spreading virus; more experiments are needed to ascertain the validity of the model.

## Favorite Four-Color, Nearest-Neighbor, Totalistic Rule

Rule chosen: 43245

I like this rule because there is symmetry and randomness at the same time.

However, doing a "visual search" on this set of rules is hard, even using Mathematica. An alternative approach is to try some compression scheme on the corresponding image and retain only an image that does not compress too much, because this would somehow identify some complex behavior. Then, we would only look at those elements. I started to experiment with some encoding from the notes section. The idea was to look for "interesting" patterns as those that would not compress well; however, this approach is somehow limited. It is slow, and it may skip interesting rules.

The ultimate approach would be to set up a Kohonen neural network, train it with some sort of good/evil feedback from the operator over a limited set of interesting cases, and the let the KNN do the search.