Author(s): Yonghwi Kwon, Brendan Saltaformaggio, I Luk Kim, Kyu Hyung Lee, Xiangyu Zhang, Dongyan Xu

Download: Paper (PDF)

Date: 27 Feb 2017

Document Type: Reports

Additional Documents: Slides Video

Associated Event: NDSS Symposium 2017

Abstract:

Malicious payload injection attacks have been a serious threat to software for decades. Unfortunately, protection against these attacks remains challenging due to the ever increasing diversity and sophistication of payload injection and triggering mechanisms used by adversaries. In this paper, we develop A2C, a system that provides general protection against payload injection attacks. A2C is based on the observation that payloads are highly fragile and thus any mutation would likely break their functionalities. Therefore, A2C mutates inputs from untrusted sources. Malicious payloads that reside in these inputs are hence mutated and broken. To assure that the program continues to function correctly when benign inputs are provided, A2C divides the state space into exploitable and post-exploitable sub-spaces, where the latter is much larger than the former, and decodes the mutated values only when they are transmitted from the former to the latter. A2C does not rely on any knowledge of malicious payloads or their injection and triggering mechanisms. Hence, its protection is general. We evaluate A2C with 30 realworld applications, including apache on a real-world work-load, and our results show that A2C effectively prevents a variety of payload injection attacks on these programs with reasonably low overhead (6.94%).