On assessing the disclosure risk of controlled adjustment methods for statistical tabular data

Minimum distance controlled tabular adjustment is a recent perturbative approach for statistical disclosure control in tabular data. Given a table to be protected, it looks for the closest safe table, using some particular distance. Con trolled adjustment is known to provide high data utility. Howev...

Descripción completa

Detalles Bibliográficos
Autor: Castro Pérez, Jordi|||0000-0003-3573-4568
Tipo de recurso: informe técnico
Fecha de publicación:2012
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/17954
Acceso en línea:https://hdl.handle.net/2117/17954
Access Level:acceso abierto
Palabra clave:Programming (Mathematics)
Programació (Matemàtica)
Classificació AMS::90 Operations research, mathematical programming::90C Mathematical programming
Àrees temàtiques de la UPC::Matemàtiques i estadística::Investigació operativa::Optimització
Descripción
Sumario:Minimum distance controlled tabular adjustment is a recent perturbative approach for statistical disclosure control in tabular data. Given a table to be protected, it looks for the closest safe table, using some particular distance. Con trolled adjustment is known to provide high data utility. However, the disclosure risk has only been partially analyzed using theoretical results from optimization. This work ext ends these previous results, providing both a more detailed theoretical analysis, and an extensive empirical assess- ment of the disclosure risk of the method. A set of 25 instance s from the literature and four different attacker scenarios are considered, with sever al random replications for each scenario, both for L 1 and L 2 distances. This amounts to the solution of more than 2000 optimization problems. The analysis of the results shows th at the approach has low dis- closure risk when the attacker has no good information on the bounds of the optimization problem. On the other hand, when the attacker has good estima tes of the bounds, and the only uncertainty is in the objective function (which is a very strong assumption), the disclosure risk of controlled adjustment is high and it s hould be avoided.