Stealth Attacks - Detection and Investigation presented at BlackHatAbuDhabi 2011

by Ryan Jones, Thomas Mackenzie,

URL : https://media.blackhat.com/bh-ad-11/Mackenzie/bh-ad-11-MacJon-StealthAttacks_Slides.pdf

Summary : Meticulous attackers can subvert audit controls to the point where a compromise is almost undetectable. We look at the tools and techniques which can be used by attackers to minimise evidence left behind and propose a novel strategy for managing this issue.
Fully identifying the method and impact of a data compromise is heavily reliant on the forensic information available to investigators. Commonly this is dependent on having logs for the compromised period. However, in the cases where an attacker has taken steps to reduce their footprint on the system, investigations can be more challenging.
We explore the various evidential sources which are commonly used to identify the extent and method of a web application compromise. We then discuss an attack which, due to its nature, is more complicated to identify and understand. The presentation will draw together the techniques used in investigating a data compromise and create an attack which is designed to completely compromise the web server while leaving the least amount of evidence on the system.
Incident readiness specialists can often recommend that verbose logging is put in place. Logging such as full http request and response logging fits the bill for the investigator but by their nature these logs have serious drawbacks for the day to day management of the server; large storage requirements, incidental storage of sensitive data and performance issues are common problems.
We suggest a new approach, restricting access or logging anomalies at the framework level. By blending the information gained at the framework level with automated application profiling techniques we can create heavily targeted logs bespoke to the specific application. This can be implemented for all applications regardless of whether source code is available. This method gives us the best chance of keeping logging to an absolute minimum whilst ensuring that techniques used to minimise forensic evidence left by an attack are unsuccessful.