As part of a broader organisational restructure, data networking research at Swinburne University of Technology has moved from the Centre for Advanced Internet Architecture (CAIA) to the Internet For Things (I4T) Research Lab.

Although CAIA no longer exists, this website reflects CAIA's activities and outputs between March 2002 and February 2017, and is being maintained as a service to the broader data networking research community.

Advice for people interested in doing a PhD with me

Each year I receive emails from people hoping to become a PhD student under my supervision. Whilst I am interested in finding bright new students who are keen to do data networking research, I only accept people whose interests match my interests. Before you contact me about doing a PhD, please consider the following points.

Ideal student?
You could be my "ideal PhD student" if:
  • You have already experimented with installing Linux or FreeBSD on an old PC, made an IP network (at home or work) and wondered "how does the Internet really work, and how can I make networks and networked applications perform better?"
  • You are aware that "perform better" is a rather imprecise goal without more context, so you seek to understand context.
  • You have a passion for discovering new insights into network behaviours and operations - both theoretical and applied.
  • You are not afraid to build prototypes and testbeds that might violate your assumptions.
  • You have done a 4-year degree in electronic engineering (or similar, such as telecommunications engineering, or strongly systems-focused Computer Science degree) with first class honours (or high 2nd class honours in rare circumstances).
  • You can explain to me why you want a PhD in one of the areas I've listed below under Research Projects.
Please do not contact me if the above description is not you. I do not enjoy writing rejection emails, and you probaby wont enjoy receiving one.

Ideal supervisor?

As noted here, my research interests tend to focus on IP network infrastructure.  In particular the way protocols and systems interact to impact on network performance, scalability, resilience and monitoring. Ultimately I view things through the lens of an end-user's experience of 'the internet'. I am also particularly interested in the interactions between IP networks and multiparty immersive/virtual environnments designed to operate on top of IP networks.

I also take a keen interest in developing the skills of my students in both doing and presenting their research. My approach is heavily directed towards performing experiments to better understand existing systems and test out new ideas.

If this sounds interesting, perhaps I could be your "ideal supervisor".

Research Projects:
I'm currently looking to supervise students in the following areas:

Characterising and modelling the growth of internet routing and addressing
IP address space may be allocated (e.g to ISPs) but that doesn't mean it is utilised. Network address translation (NAT) is increasingly deployed to decouple customer network addressing from limitations of ISP address allocations. The number of ISPs making up the Internet, their business relationships and degree of interconnectedness, continues to change every week. This project will review and improve on existing techniques for measuring (e.g. through probing, monitoring and surveys) and modeling such dynamic attributes of the Internet's structure. Potential benefits include improved predictive models for future address allocation policies, evaluating causes for (and potential mitigtation of) excess BGP update traffic between Autonomous Systems (ASes), estimating the like demand for IPv6 address space if NAT were to be phased out, etc. [STING]

TCP and consumer expectations - what congestion control plays nice with interactive applications?
Transmission control protocol (TCP) deserves significant credit for the internet’s wide-spread utility over the past 25+ years. The relatively modern NewReno variant of TCP balances two key goals: Provide reliable transfer of byte-streams across the IP layer’s unpredictable packet-based service, and minimise congestion inside end hosts and the underlying IP network(s) while maximising performance. Maximisation of TCP performance has been an active and challenging area for academic and industry research into congestion control (CC) techniques. New research is required into an increasingly important new scenario - TCP flows sharing consumer-grade ‘broadband’ links with non-reactive, latency-sensitive UDP-based applications such as Voice over IP (VoIP) and online games. CUBIC (currently the default CC algorithm for Linux) induces more latency (RTT) than NewReno, negatively impacting on VoIP or game traffic sharing a home broadband link. Can we discover and demonstrate an improved TCP variant that co-exists nicely with real-time traffic over home WiFi, ADSL, cable modem and optical fibre links? [NGEN, NewTCP]

Application of IP traffic classification schemes for network resource management and privacy-enhanced traffic analysis
The ability to inspect IP packet payloads has long been used to classify the nature of individual traffic flows. For legal reasons, or due to end-user use of IP layer encryption, IP packet payloads are becoming less useful (or simply unavailble) for classification. In this project we will study and improve upon the use of machine learning to perform classification of IP traffic based on previously-learned statistical patterns. Such newer techniques avoid parsing and interpreting the contents of IP payloads, yet retain the potential to categorise the types of applications that generate different types of IP flows. This has applications as diverse as automated control of network Quality of Service (QoS) features, non-intrusive characterisation of customer traffic profiles for Lawful Interception (LI), and surveys of customer usage patterns for marketing purposes. [DIFFUSE, ANGEL, DSTC]

Using multiplayer game engines to visualise and control dynamic network state
Network operators are continually challenged by the task of identifying, and subsequently reacting to, anomalous (and potentially malicious) Internet Protocol (IP) traffic entering, traversing or leaving their systems. Recognising and reaction to anomalous network events (such as by updating firewall rules, re-routing traffic, etc) is often labour-intensive except in simple, clearly defined computer-controlled scenarios. Imagine a virtual world where real-world events (like port scans) were mapped into dynamic behaviour of objects inside the virtual world. Imagine further that avatars inside the virtual world could trigger changes to the real-world by interacting with those objects inside the virtual world. Multiplayer game engines provide a convenient platform on which to create exactly this kind of environment, with the potential for participants to join from desktop PCs, laptops or networked tablets and PDAs. We havedemonstrated the basic idea using Quake III Arena , but more work is required to explore and characterise the limits and capabilities of building and using such "virtual" network operation centres. [L3DGE]

Development and/or detection of covert channels across IP networks
Covert channels aim to hide the very existence of certain communication between two parties by overloading the semantics of an existing means of communication (overt channel). Their use is often motivated by the existence of an adversarial relationship between two parties. Covert channels in computer networks are analogous to techniques for hiding information in audio, visual or textual content (steganography). While steganography requires some form of content as cover, covert channels require some network protocol and traffic as a carrier. In this project we will study and model the performance characteristics of different covert channels when implemented in common IP network environments. Concurrently (or as an alternative) we will study the potential for covert channels to be detected and suppressed without interfering with the overt IP data channel. [Covert]

Network support for interactive online games / immersive environments
Interactive, real-time online games represent a considerable challenge for ISPs. In an ideal world with excess funding an ISP could simply over-provision their entire network. But the real world demands that new online game players must be supported on existing links - shared with everyone else's web surfing, file sharing, and email traffic. Network engineering is a balancing act. A proactive ISP tries to maintain packet latency, loss, and throughput expectations of its customers by adapting their internal resources to the offered load. I am interested in research that provides new insights into topics such as the latency tolerance of players, network layer mechanisms for cheat-mitigation, the hidden network-layer impact of server discovery protocols, synthetic construction of realistic game traffic simulations, passive real-time detection of live game traffic in ISP access networks, and the use of 3D game engines for interactive visualisation of network activity. [GENIUS,SONG,BITSS]

Applying for candidature under my supervision:
Before sending me an email, check Swinburne's requirements at this page. Then write a 1 or 2 page research proposal (including references) showing how your research interests map into one of my projects above. This information will help me decide whether to have further discussions with you about possible candidature and scholarships.

Last Updated: Monday 22-Aug-2011 14:45:33 AEST | No longer maintained. Pre-2018 was maintained and authorised by Grenville Armitage, garmitage@swin.edu.au