University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Daniel Kaiser

Doctoral Student in the PhD program since 01.03.2012 .

advisors

1. Prof. Dr. Marcel Waldvogel
2. Prof. Dr. Ulrik Brandes

organisational data

Room: V 521
E-mail: daniel.kaiser "at" uni-konstanz.de
Other Resources: Distributed Systems
picture

project description

Today’s networks provide huge amounts of information. A lot of this information is not supposed to be visible to others, however, this information has to be sent in the clear in order for the network to function correctly. This is a huge challenge for privacy. In this work, I am assessing whether all this information actually needs to be shared, or whether less information would also suffice. For example, even though everyone believes that mobile network providers need to always know your exact location for mobile telephony and data to work, we have shown that with just a little more work, this information does not have to be continuously shared. My goal is to extend this result to more network types and protocols, to improve everyone’s privacy and reduce the impact of hacked or malicious nodes on the network security. Multicast DNS Service Discovery (mDNS-SD) is a prevalent technique allowing service distribution and discovery in local networks without configuration (Zeroconf). Among possible application areas are device synchronization, instant messaging, VoiP, file and screen sharing. It is very convenient for users, because they can connect to services (and offer services) when they enter a network without any manual configuration regardless of the network parameters. However, there is a drawback: It requires the public exposure of the offering and searching identities along with information about the offered and requested services. One part of my work is the design of an efficient privacy extension to the mDNS-SD protocol. Since most of the offered services are not meant for the public in the first place, we offer these services only to hosts with a pre-established relationship. Public service instance identifiers are changed to cryptic identifiers which only make sense to known hosts, while all other information that has to be transmitted via multicast is encrypted. These identifiers allow a host to check if a received packet should be processed, using a single hashtable lookup. In order to further enhance privacy, to ease pairing, and to significantly reduce the number of necessary multicast packets, we offer just a single service-offering service using mutlicast, while publishing the actual services using an encrypted unicast connection to hosts who’s network parameters were determined by the service-offering service. The advantage of our solution is easy deployabilty: None of the deeper layer protocols have to be changed, changes are only made in the Zeroconf daemon, and backwards compatibility is granted. To show the feasibility of our mDNS-SD privacy extension we are developing an implementation based on Avahi. Since the Zeroconf daemon is the only software we altered, existing applications using Zeroconf have not to be updated to enable our privacy features. While the above presented privacy extension prevents privacy breaches caused by the application layer protocol mDNS-SD, OSI layer 2 also holds privacy problems, because it uses the MAC address as constant identifier. In a WiFi network everyone can see the MAC addresses of all packets within signal reach when setting the network card to promiscuous mode; even if the WiFi network uses WPA2. Constant identifiers are used for discovery of access points as well as for the following communication. Using the MAC address it is possible to easily correlate packets knowing they where sent by the same device and thereby track that device. To prevent that privacy breach, identifier free WiFi networks were created. They allow to send packets via broadcast, that do not contain a constant identifier like a MAC address, while a host is still being able to tell if a packet was meant for it or not using a constant number of hash operations. The keys are changed for each packet, or at least after a certain time interval, instead of using a constant identifier. Since there are no constant identifiers, packets sent can not be correlated. The discovery part of such a protocol is a challenge, because it is necessary to broadcast uncorrelatable presence information without knowing if trusted parties are in reach to receive these packets. Each packet has to look different without the possibility of transmitting new keys in 1each packet, which can only be done after two entities found each other. Existing proposals for identifier free discovery protocols either rely on time and thus have to deal with time skew and scalability issues because they have to recalculate the whole hashtable, are inefficient or do not provide full unlinkability. The second part of my work is to create an efficient, identifier free discovery method. My solution needs less key material to be send, is not dependent on time, does not need the AP to recalculate the hashtables, granting better scalability, and provides even better unlinkablilty. The main idea is to use simultaneous congruence to construct keys on the cient and modulo as a hash function on the AP in a way that a single key can be recognized by several APs. In order to make each key look different and to prevent correlating keys, a client subdivides its possibly large AP set into subsets and calculates a key for this subset using simultaneous congruence. Since the subsets are chosen randomly before each key creation, this makes keys uncorrelatable without the time skew problem and lets the client decide, when to change keys making it possible to use keys only once; thus granting better unlinkability properties. We are further investigating how to grant efficient private discovery and exploration of nodes in huge peer to peer networks.

publications

The following list of publications covers only those, which are or were published during participation at the Graduiertenkolleg / PhD program.

id should be a number

curriculum vitae

Since 03/2012 Doctoral Student in the PhD Program, University of Konstanz
10/2009 - 12/2011 Master Studies, Information Engineering, University of Konstanz
11/2008 - 12/2011 Student Research Assistant, Algorithmics Group, University of Konstanz
10/2006 - 11/2009 Bachelor Studies, Information Engineering, University of Konstanz