ML and Network Traces to M.A.R.S

Waldemar Cruz, Laurent D. Michel, Benjamin Drozdenko, Steven Roodbeen

Abstract: MARS is a Microservice Architecture Recovery Solution that uses Machine Learning and lightweight Network Traces to recover the architecture of applications in order to deploy network security policies and protect the organization against complex threats that may exploit several vulnerabilities to breach an application and either exfiltrate sensitive data or carry out denial of service attacks. The adoption of such security policies is often hindered by the lack of suitable documentation. This paper describes a novel methodology that uses machine learning on captured network traces to recover application architectures.