Anonymous On-line Communication Between Program Analyses

TL;DR

This paper introduces a lightweight, anonymous client-server communication model for program analyses, enabling spontaneous information exchange that improves analysis results without central control.

Contribution

It presents a novel, decentralized communication framework for program analyses that enhances cooperation and performance through spontaneous, anonymous information sharing.

Findings

Model performs well across multiple analysis combinations

Communication improves analysis results significantly

Effective on over 400 benchmarks

Abstract

We propose a light-weight client-server model of communication between program analyses. Clients are individual analyses and the server mediates their communication. A client cannot see properties of any other and the communication is anonymous. There is no central algorithm standing above clients which would tell them when to communicate what information. Clients communicate with others spontaneously, according to their actual personal needs. The model is based on our observation that a piece of information provided to an analysis at a right place may (substantially) improve its result. We evaluated the proposed communication model for all possible combinations of three clients on more than 400 benchmarks and the results show that the communication model performs well in practice.