@masterthesis{Frazier2024,
       author = "Frazier, James Gunder",
       title =  {{Applications of Artificial Intelligence to Information Privacy}},	
       type = {Bachelor's Thesis},

       howpublished = "\url{https://ir.hamilton.edu/do/8817f607-6726-4b4a-90d5-bba0f4ccb7b2}",
       institution = {Hamilton College},
       year = 2024,
       month = may,
       school = {Computer Science},
}

Creator

Date

May 2024

Description

Software De!ned Perimeter (SDP) is a zero-trust network-isolation defense technique which aims to limit security risks by giving dy- namic account type assignments to network users. Despite SDP being proven as an e"ective defense strategy in various domains, it has yet to see wide-spread use due to its drawbacks. One of SDP’s most pressing issues is the need for an expert to manually con!gure it for each unique application. Here we describe a novel system for designing SDP networks called SDPush which can automatically design and analyze possible con!gurations for a given network with user-speci!cations. Since there is not a systematic approach for account type design and assignment, we develop a two-step optimization system consisting of a bitstring genetic algorithm and a genetic programming sub-system for designing and evaluating SDP networks respectively. In order to evolve an SDP con!guration exhibiting the user-speci!ed characteristics while also minimizing security risk, we implement our system to support multi-objective search spaces by providing the system’s training set with di"erent cases aimed at evaluating di"erent aspects of the network con!g- uration. We present initial results of experiments on networks of varying size and characteristic requirements.

Type of Work

• Thesis

Institution

Hamilton College

Area of Study

• Computer Science

Degree

Bachelor of Arts

Advisor

Part of Collection

Senior Theses

Rights Statement

The copyright for this work is held by its creators and Hamilton College. More information