Current Research
In seeking to improve infrastructure systems and make them more usable for various applications, our current research is primarilly focused in the following four areas:
- Optimistic replication and its applications
- Large scale data management
- IT infrastructure protection
- Scalable backup architectures
Below you will find a description of our latest research efforts and links to find out more:
Summary Hash History (SHH) | RepuScore | Privilege Messaging | Honeynet
| Summary Hash History (SHH) |
|
| The unprecedented growth of the world’s first non-profit, open-source encyclopedia has put considerable stress on the Wikipedia foundation, who is constantly looking for
donations to support their rising infrastructure and hosting costs while maintaining adequate quality of service. That the public-owned content depends on a single organization’s
financial fate is a major concern to many. We propose using optimistic replication to ensure that the encyclopedia content is preserved at multiple sites managed by different organizations.
Replicating the Wikipedia database not only requires an efficient update exchange protocol but also needs a mechanism to identify the origin of update pollution or “anonymous
slander” as it is frequently referred to by Wikipedia users. In order to meet these challenges effectively, we introduce the Summary Hash History (SHH) approach. In this approach, each
site maintains a tamper-evident update history to mitigate security challenges and to readily determine the exact set of updates to be transferred during peer-topeer reconciliation between
sites. We first implemented Basic-SHH which confirmed our intuition that SHH can be used for both the tamper-evident history and the efficient update exchange mechanism. However, our
evaluations revealed that Basic-SHH is unable to guarantee convergence among replicas in scenarios involving concurrent updates. Thus, we developed a variant called Associative-SHH that
overcomes Basic-SHH’s limitations by not only providing eventual convergence but also enabling convergence of concurrent updates across partitioned networks. |
| RepuScore |
|
| We propose RepuScore, a collaborative reputation management framework over email infrastrucure, which allows participating organizations to establish sender accountability on the basis of senders' past actions. RepuScore's generalized design can be deployed with any Sender Authentication technique such as SPF, SenderID and DKIM. With RepuScore, participating organizations collect information on sender reputation locally from users or existing spam classification mechanisms and submit it to a central RepuScore authority. The central authority generates a global reputation summary which can be used to enforce sender accountability. We present the algorithms for reputation score calculation and share our findings from experiments based on a RepuScore prototype using
1. our simulation logs and
2. a 20 day log from a non-profit organization with 5 collaborating domains.
RepuScore is capable of thwarting Sybil Attacks where a single attacker subverts the reputation system by creating a large number of pseudonymous entities, using them to gain a disproportionately large influence. A reputation system's vulnerability to a Sybil attack depends on how cheaply identities can be generated, the degree to which the reputation system accepts inputs from entities that do not have a chain of trust linking them to a trusted entity, and whether the reputation system treats all entities identically. |
| Privilege Messaging |
|
| The current email infrastructure is burdened by multiple resource constraints and a plethora of security issues. Apart from the fact that email users are spending more time
and effort sifting through unsolicited emails, more serious problems such as Phishing are on the rise. This can be attributed to a fundamental shortcoming in the current email infrastructure:
a lack of an authorization framework. This allows any user to create content in anyone’s mailbox. In this paper, we revisit the fundamental problem of non-existent authorization and
discuss the design of an effective authorization service overlaying the existing email infrastructure. We propose Privilege Messaging (P-Messaging), a fine-granular authorization framework
that operates on the principle that a sender requires a set of privileges in order to send messages, simultaneously enables the receiver ’s infrastructure server to verify the messages
before accepting it. We present a prototype implementation and discuss its benefits. An automatic classification of email can be effectively performed based on the privilege-tag.
Privilege-tag can provide flexible and fine-granular reputation management than current domain-based solutions. The use of privilege-tag as entry ID in a white-list can be more manageable
than the use of individual email address. Finally, the privilege-tag can be used as an email header, retaining the benefits of currently deployed MTA architecture, namely reliability and
flexibility. |
| Honeynet |
|
| The UNC Charlotte Honeynet project is part of the research effort of the network security group in the Laboratory of Information and Infrastructure Security. We hope to use
the honeynet to gather important data as input to network and security research. This data can then be used as an educational tool to increase the knowledge of all network security students
and professionals. Our honeynet implementation uses an ISP and is not part of the campus network. |