Security Reasoning for Distributed Systems with Uncertainty (CMU/Cornell Collaborative Proposal) - July 2014

Public Audience
Purpose: To highlight progress. Information is generally at a higher level which is accessible to the interested public.

PI(s): Andre Platzer
Co-PI(s):
Researchers: Dexter Kozen (Cornell)

HARD PROBLEM(S) ADDRESSED
This refers to Hard Problems, released November 2012.

Scalability & composability and resilient architecture. We are interested in answering the question "Is my system sufficiently robust to both stochastic failures and deliberate attack"?

PUBLICATIONS
Report papers written as a results of this research. If accepted by or submitted to a journal, which journal. If presented at a conference, which conference.

ACCOMPLISHMENT HIGHLIGHTS

Studied approximation procedures and error bounds for a class of linear complementarity problems (LCPs). Complementarity problems unify problems from many different domains including convex optimization, machine learning, game theory, and planning.

LCPs can find Nash equilibria in sequential and simultaneous two-player games with multiple rounds of interaction and uncertainty. LCPs can also find optimal policies of MDPs (Markov Decision Processes).
Our new approximation techniques and bounds are based on learning theory such as Rademacher complexity and Fourier basis representations.