Detecting Data Exfiltration by Integrating Information Across Layers

Abstract

Data exfiltration is the unauthorized leakage of confidential data from a system. Unlike intrusions that seek to overtly disable or damage a system, it is particularly hard to detect because it uses a variety of low/slow vectors and advanced persistent threats (APTs). It is often assisted (intentionally or not) by an insider who might be an employee who downloads a trojan or uses a hardware component that has been tampered with or acquired from an unreliable source. Conventional scan and test based detection approaches work poorly, especially for hardware with embedded trojans. We describe a framework to detect potential exfiltration events that actively monitors of a set of key parameters that cover the entire stack, from hardware to the application layer. An attack alert is generated only if several monitors detect suspicious activity within a short temporal window. The cross-layer monitoring and integration helps ensure accurate alerts with fewer false positives and makes designing a successful attack more difficult.