UNIVERSITY OF TARGU MURES
Linear Programming (LP) is one of the most versatile polynomial-time solvable optimization problems. In this paper the authors demonstrate a number of attacks against proposed protocols for privacy-preserving linear programming, based on publishing and solving a transformed version of the problem instance. Their attacks exploit the geometric structure of the problem, which has mostly been overlooked in the previous analyzes and is largely preserved by the proposed transformations. The attacks are efficient in practice and cast serious doubt to the viability of transformation-based approaches in general.