Topics in System Security
17 students in 8 teams
This module aims to prepare graduate and senior undergraduate students for research and development in system security by investigating security and privacy issues in network and distributed systems. At the end of the module, students will be able to: (1) explain the security challenges and opportunities of various emerging network and distributed systems. (2) critique state-of-the-art attack/defense mechanisms and identify possible gaps that could be addressed by future work. (3) find interesting research problems and propose novel solutions.
There have been multiple reports that the traffic created by Internet-of-Things (IoT) devices leak significant privacy sensitive information, like sleeping patterns, personal preferences, etc, and these leakage has become increasingly serious as more and more IoT devices steps into our daily life. In this project, we are focusing on one of the popular IoT devices, namely, Philips Hue (one kind of smart light bulbs). Through studying the working principles as well as analyzing the traffic patterns, we are trying to exploit vulnerabilities of the smart bulb system through potential attacks.
Distributed reflective denial-of-service (DRDoS) attack is one of the special forms of denial-of-service attacks. The attackers send a small amount of requests to the public server running vulnerable protocols so that the public server will flood the target victim with large-volume traffic. To resolve this attack, the first step is to find the vulnerable protocols and servers running that protocols, which can only be achieved manually or semi-manually in current solutions. In this project, we first propose an automatic tool to identify the vulnerable protocols and servers (a.k.a the amplifiers) and measure the amplification factors by analyzing the traffic flow. This will greatly release the burden of the manual work and helps on ranking the severity of DRDoS attacks since the amplification factor is an essential component in threat evaluation.
With the growing number of network-connected devices owned by users, the Internet of Things (IoT) is set to become a indispensable part of our lives. To manage such a large number of devices, a central management solution is required. Cloud based solutions are a popular candidate but they give rise to many privacy concerns for the user. In this project, we study the cloud based management of IoT devices where the cloud service is not trustworthy and identify related privacy concerns. Further, we design an alternative Cloud Based Service (CBS) that addresses those privacy concerns. The proposed CBS uses three approaches to achieve privacy, namely distribution of trust over multiple parties, policy based solutions and technical solutions (for ex. cryptography). Finally, we derive important insights from our proposed model and discuss them in detail.
Web search engines such as Google are designed to look for websites. Shodan is the search engine for internet-connected devices. The users can look for physical devices (servers, webcams or routers) or services (apache server, SMTP). At first sight, Shodan seems to be a hacker tool to find vulnerable devices. One part of our work focuses on understanding the sometime obscure mechanisms of Shodan. Another consists in using the developer API to study global characteristics of vulnerable protocols as well as to target specific vulnerable protocols from a hacker’s point of view.
This project aims to measure the censorship in some countries by analyzing DNS query records in top-level domain name server (.com/.net name server). We measure the "popularity" of top 50 websites on Alexa.com in different countries, and design proper measurement to infer potential censorship level in these countries. The result of this experiment can be instructive to a thorough examine on the full DNS query record of the entire Internet.
Massive scale distributed denial of service attacks are are able to completely overwhelm the resources of a single autonomous system. To protect against this, collaborative defense mechanisms that enable multiple autonomous systems to reroute legitimate traffic and filter out malicious traffic have been proposed. Unfortunately, it is possible for a malicious autonomous system to claim an fictitious attack in order to trick his collaborators into re-routing or filtering traffic not related to a true attack. In this paper, we present a method to verify whether connections passing through an autonomous system are truly being degraded. In the method we propose, we prevent a malicious autonomous system from raising an alarm without actually causing network degradation.