• About Us
• CSEAdmin
• Academics
  • Undergraduate
  • Graduate
• Admissions
• Advising
  • Undergraduate
  • Graduate
• People
  • Faculty
  • Staff
  • Student
• Research
• News/Events
• Contact Us

• News & Events Home
• High School Robot Programming Contest
• ACM Intercollegiate Computer Programming Contest>
• Seminars
• Distinguished Lecture Series
  • 2005-2006 Lecture Series
  • 2004-2005 Lecture Series
  • 2003-2004 Lecture Series

Dr. Matthew Wright, Dr. Dongang Liu and students participate in SCISS 2006

Title: Towards Stronger Peer-To-Peer Anonymity

Authors: Arjun Nambiar and Matthew Wright

Abstract: Anonymous communications systems on the Internet provide protection against eavesdroppers and others that seek to link users with their communications. These systems have many important applications in areas such as law enforcement, intelligence gathering, business privacy, anonymous publishing, and personal privacy. Currently deployed systems rely on a relatively small set of advertised servers, or proxies, to forward messages for the user. These systems can suffer from scalability problems, with potentially large bandwidth and system overhead costs, and the servers themselves can be targets of direct attacks.

Peer-to-peer anonymous communications systems, such as Tarzan and MorphMix, have been proposed as a way to alleviate these problems with a large and dynamic set of peers that act as both clients and proxies. This makes direct attacks less effective and increases scalability. Tarzan, however, requires that each peer know the identity of all other peers, which makes it highly vulnerable to intersection attacks and does not well scale beyond 10,000 nodes. MorphMix does not have this requirement, but it requires that users allow other peers to choose the proxies on their path. Attacker-controlled peers will always select other colluding attackers to be the proxy. Although the authors of MorphMix propose a collusion detection scheme, attackers will be able to choose each other as proxies for others far too often while the detection algorithm is still learning. The fundamental problem is one of selecting peers independently at random, to ensure unbiased path selection, while not distributing lists of all the peers in the system to all other peers.

We propose a new peer-to-peer anonymous communications system using distributed hash tables (DHTs). Similar to peer-to-peer file-sharing systems that use DHTs, our system maps each IP address to a point on the ID space using consistent hashing. We further divide the ID space into groups, conceptually organized as a binary tree for purposes of node lookup. Each node has knowledge of all the nodes in its own group, as well as knowing a limited number of nodes in other groups. This knowledge is enough to effectively route lookups throughout the system. Nodes use redundancy and probabilistic checking when performing lookups to prevent malicious nodes from returning false information without detection. We show that our scheme prevents attackers from biasing path selection, while incurring moderate overheads, as long as the fraction of malicious nodes is less than 20%. Additionally, the system prevents attackers from obtaining a snapshot of the entire system until the number of attackers grows too large (e.g. 15% of all nodes for 10,000 peers and 256 groups). The number of groups can be used as a tunable parameter in the system that can be used to balance performance and security.

---

Title: HB+ and an Active Attack

Authors: Bridget Beamon

Abstract: Many low-cost pervasive devices lack the resources to implement secure cryptographic protocols. Most of today's RFID devices fit this category. Existing schemes only protect against passive attacks or are too computationally costly for low-cost RFID tags. We believe that it is important to protect against limited passive attacks without the use of expensive cryptographic techniques. However, inexpensive and secure seem be contradictory terms by today's standards. Research in Human-Computer Authentication protocols holds promise.

Background: Humans can be seen as weak computing devices with similar limitations as RFID devices. Humans don't tend to remember long passwords and can't efficiently compute lengthy calculations. There are other similarities. So, it is reasonable to conclude that secure protocols performed by unaided humans could be adapted for low-cost resource-limited devices. This is the focus of work done by Juels and Weis. They propose HB+, a lightweight symmetric key authentication protocol that improves upon work done by Hopper and Blum. HB+ is based on the LPN problem, known to be NP-Hard. Also important, is the fact that HB+ is easy and inexpensive to implement in hardware. It involves "computing a binary inner product which only requires bitwise AND & XOR operations that can be computed on the fly as bits are received". This is important since stronger cryptographic algorithms such as AES require more resources that many low cost pervasive devices have or ever will have. Juels and Weis thought they had proved that it is secure against passive and active attacks. This proof was later shown to be flawed. A realistic active attack was presented in a paper by Gilber, Robshaw and Sibert. And though the proof Juels and Weis seems sound, it does not consider the fact that each "accept" or "reject" from a valid reader leaks one bit of information about the secret key. An active attack may be successfully launched on HB+ in time linear with key size and number of rounds.

Research: Our ongoing research will implement the HB+ protocol as described by Juels and Weiss. The secret key size, number of rounds and noise probability will be programmable parameters. The active attack discussed by Gilber, Robshaw and Sibert will also be demonstrated as a java simulation. Timing Analysis with varying key sizes and rounds will indicate optimal combinations for security. The speed and accuracy of the HB+ algorithm and its active attack will be demonstrated. One idea of improving the HB+ protocol by reducing the discovery of authentication results to an LPN problem will be explored. Also, to be considered is the practicality of the aforementioned active attacks in specific contexts. HB+ could in fact be secure within certain constraints. In particular, it may be appropriate as-is for many current RFID uses. We will explore constraints for security of the existing HB+ protocol when applied RFID. In all, this research seeks to fill a gap in existing work on secure low cost authentication solutions.

---

Title: Evaluating VoIP Performance over the Tor Anonymous Communications Network

Authors: Nayantara Mallesh and Matthew Wright

Abstract: Anonymous communication has become popular over the past few years due to the rising demand for user privacy. Anonymity solutions like Crowds, Tor, and Hordes allow users to protect their privacy on the public Internet. Users can route their connections via these anonymous networks so that the recipient or any random observer on the Internet cannot associate the connection with the originating user. It seems natural that voice over IP (VoIP) is another application that will see an increased demand for anonymity.

Currently, the popular Tor anonymous communications overlay network supports only TCP connections, but UDP-enabled Tor has been introduced in a recent proposal. Support for UDP is essential if Tor has to support popular applications like VoIP and streaming multimedia. However, VoIP users expect quality of service to be as good as that provided by the circuit-switched telephone network. Dedicated circuit-switched connections ensure that voice calls have minimum delay, negligible jitter, minimum packet loss, and maximum throughput. VoIP attempts to provide comparable quality for voice applications using packets to transport voice data over the Internet. Sending voice packets over the Internet exposes them to packet loss, long latencies, jitter, and low throughput, all of which are detrimental to the quality of voice connections.

Routing a VoIP connection through an anonymizing network can provide privacy but also increases the latency of the voice connection. In this project we aim to measure the performance characteristics of VoIP connections that are sent through a UDP-enabled Tor network and compare it with the performance of a direct VoIP connection and with the performance of a VoIP connection passing through a single hop anonymous connection like Anonymizer. The results of this project will show the network characteristics introduced by an anonymizing network like Tor and whether VoIP calls are feasible over such a network. We plan to use ns2 for simulating the network and data flows, and we will measure performance characteristics such as end-to-end delay, jitter, packet loss, and throughput.