Students: Capts. Anthony Collier & Brian Greunke, USMC

BGP hijacks remain a relevant concern for network professionals. Historical incidents, both malicious and accidental, have caused large scale service loss. Mitigating factors such as route filtering, TCP session encryption, and Secure-BGP, though partially effective, have not prevented the ability for hijacks to occur. Existing tools such as Mininet or GNS3 allow a small scale demonstration of BGP hijacks but cannot reproduce real world scale or even large scale incidents. Our project will focus on a method for automatically implementing hijacks on larger scale emulated networks, with the ability to analyze the results post execution.

Students: Capts. Peter Bose & Brian McCarthy, USMC

Work has been underway at IEEE to develop protocols for vehicular ad-hoc networks (VANET), a crucial technology for the development of an intelligent transportation system. These protocols, termed Wireless Access in Vehicular Environment (WAVE), facilitate communication between vehicles (V2V) and roadside unit (RSU) infrastructure (V2I) in the 5.855 – 5.925 GHz part of the spectrum. WAVE is intended to work at ranges up to 1 km, but some implementations may limit range to as little as 30 m.

Student: Tao-hsiang Chang

Model and evaluate the variants for a cluster network: While doing parallel distributed processing on extremely large dataset (e.g. a cached Internet,) it will be interested what is the best setting for performance when there’s virtually no time constraints (all files are to be processed)? Explicitly, when the processing function is known, would there be a threshold that adding more processors would result in no more improvement in terms of performance? Since DCTCP is optimized for incast & queuing problem, would the performance still be benefit by DCTCP in this case?

Student: LCDR Christopher Wasek, USN

The focus of the project is to analyze the affect network congestion has on the calculation of the clock skew of a virtual machine (VM) instance in the cloud. My thesis involves the calculation of a cloud VM’s clock skew through the collection and measurement of TCP timestamps in order to determine if two VMs are co-located on the same physical server. The process to collect timestamps from all launched cloud instances, as designed, will take many hours. In this sense, it is reasonable to assume that depending on the time of day, the amount of network congestion can vary across the timestamp collection process. Thus, if congestion indeed proves to have a negative impact on the calculation of clock skew, two co-located VMs sampled on opposite ends of the timestamp collection process could easily have significantly different clock skews and register as a false negative co-location.

A virtually simulated cloud environment will be created utilizing the ns-3 discrete event simulation tool. Specific levels of network traffic congestion will be added into the simulation while taking a series of time measurements from a target VM to a data collection VM. The clock skew of the VM will be simulated by injecting the timestamps with a specific skew value. A measurement of time collection will be conducted with no traffic congestion in order to establish an experiment baseline. Various levels of traffic congestion will be injected into the simulation while collecting the times values. The VM’s clock skew will be calculated for each simulation run, checking the value against the baseline measurement.

Student: LCDR Andrew Belding, USN

With IoT device and network development receiving more and more industry attention, 6LoWPAN has been emerging as a standard for low power, mesh networked, device to device communication. With the low power requirements, end to end addressability of IPv6, and resiliency provided by mesh topology, it is an attractive option for many IoT developers. With dynamic sensor networks playing a significant role in military intelligence, much of this capability could potentially translates to the operational arena and can greatly improve tactical situational awareness.

Student: LT Daniel Lukaszewski, USN

Virtual Private Networks (VPN) utilize standard transport protocols (mainly TCP) and may benefit from utilizing Multipath TCP (MPTCP) or Multipath UDP (MPUDP) to transfer data after establishing initial secure tunnel connection. MPTCP/UDP enhance the standard protocols by multiplexing traffic over multiple standard TCP/UDP connections. Multiplexing the data transfer should allow the VPN to provide increased throughput, reduced latency, and greater availability. There has been research to suggest MPTCP in VPNs does not offer any improvements and may negatively impact the data transfer due to multiple layers of congestion control when encapsulating TCP in TCP. MPUDP however has been suggested as a possible solution to the TCP in TCP problem because UDP does not contain congestion control. MPUDP has not yet been tested in a VPN to determine any benefits to it’s use.