UC Riverside

Keeping users sensitive information secure and private in todays network is challenging. Networks are large, complicated distributed systems and are subject to a wide variety of attacks, such as eavesdropping, identity spoofing, hijacking, etc. What is worse, encrypting data is often not enough in light of advanced threats such as side channel attacks, which enable malicious attackers to infer sensitive data from insignificant network information unexpectedly. For this purpose, we pro- pose series of techniques to prevent such information leakage at different layers in network stacks, and raise awareness of its severity. More specifically, 1) we propose a practical physical (PHY) layer security framework FOG, for effective packet header obfuscation using MIMO, to keep eavesdroppers from receiving any meaningful packet information; 2) we identify and fix a subtle yet serious pure off-path side channel vulnerability (CVE-2016-5696) introduced in both TCP specification and its implementation in Linux kernel, which prevents malicious attackers from exploiting it to indicate arbitrary connections state, reset the connection or even further hijack the connection; 3) we propose a principled TCP side channel vulnerability discovery solution based on model checking and program analysis, and automatically identify 12 new side channel vulnerabilities (and 3 old ones) from TCP implementation in Linux and FreeBSD kernel code. The ultimate goal is to help guide the future design and implementation of network stacks.

Keeping users’ sensitive information secure and private in today’s network is challenging. Network nowadays are subject to a wide variety of attacks, such as eavesdropping, identity spoofing, denial of service, etc. What is worse, encrypting sensitive data is often not enough in light of advanced threats such as side channel attacks, which enable malicious attackers to infer sensitive data from “insignificant” network information unexpectedly. For this purpose, we propose series of techniques to prevent such information leakage at different layers in network stack, and raise awareness of its severity. In our first work, we propose a practical physical (PHY) layer security framework FOG, for effective packet header obfuscation using MIMO, to prevent eavesdroppers from receiving any packet headers to profile users. Secondly, we identify and fix a subtle yet serious pure off-path side channel vulnerability (CVE-2016-5696) introduced in both TCP specification and its implementation in Linux kernel. This vulnerability allows malicious attackers to indicate arbitrary TCP connection’s state, reset the connection or even further hijack the connection. Motivated by the fact that most previous TCP side channel vulnerabilities are manually identified, in our last work, we propose a principled TCP side channel vulnerability discovery solution based on model checking and program analysis. It automatically identifies 12 new side channel vulnerabilities (and 3 old ones) from TCP implementation in Linux and FreeBSD kernel code. The ultimate goal of my research is to help guide the future design and implementation of network stacks.