Date of Award


Document Type

Open Access Dissertation


Computer Science and Engineering

First Advisor

Wenyuan Xu


In the past few decades, network architectures and protocols are often designed to achieve a high throughput and a low latency. Security was rarely considered during the initial design phases. As a result, many network systems are insecure by design. Once they are widely deployed, the inherent vulnerabilities may be difficult to eliminate due to the prohibitive update cost. In this dissertation, we examine such types of vulnerabilities in various networks and design end-to-end-based solutions that allow end systems to address such loopholes.

The end-to-end argument was originally proposed to let end hosts implement application-specific functions rather than letting intermediate network nodes (i.e., routers) perform unneeded functions. In this dissertation, we apply the end-to-end principle to address three problems in wireless networks that are caused by design flaw with following reasons: either because integrating solutions into a large number of already deployed intermediate nodes is not a viable option or because end hosts are in a better position to cope with the problems. First, we study the problem of jamming in a multihop wireless network. Jamming attacks are possible because wireless networks communicate over a shared medium. It is easy to launch a jamming attack but is difficult to defend against it. To ensure the end-to-end packet delivery, we propose a jamming-resilient multipath routing algorithm that maximizes end-to-end availability based on the availability history between sources and destinations. Second, we investigate caller ID spoofing attacks in telephone networks in which an attacker can send a fake caller ID to a callee rather than her real one to impersonate as someone else. Such attacks are possible because there is no caller ID authentication mechanism in operator interconnection protocols. Modifying current protocols to verify caller ID between operators may be infeasible due to the scale of deployed systems. So, we propose two schemes to detect caller ID spoofing attacks based on end-to-end verification. Finally, we examine evil twin access point attacks in wireless hotspots. In such attacks, an adversary sets up a phishing access point that has the same Service Set IDentification (SSID) as the legitimate ones in the hotspot. Such attacks are easy to launch because of how 802.11 standards are designed. Existing solutions take away convenience from the user while providing security. Our aim is to detect evil twin access point attacks in wireless hotspots without modifying how access point works in hotspots and without additional infrastructure support. We propose an end-to-end-based mechanism that can effectively detect evil twin access point attacks in wireless hotspots.