This is a new and exciting research activity in my group. The aim is to fundamentally understand the data privacy problem from an information-theoretic perspective and use this understanding to develop new methods that strongly and provably mask sensitive personal information in large datasets (such as government census or medical datasets) against re-identification and inference attacks, whilst providing high-fidelity useful data to the users (such as medical researchers, data analysts or other scientists).
The research is primarily funded by the prestigious Australian Research Council (ARC) Future Fellowship Scheme between 2020 and 2023 and builds upon on my extensive research experience and track record in information theory, index coding, network coding (including communication for omniscience, source compression, and information theoretic security), as well as optimisation of wireless communication systems, as per outlined below. I was also recently awarded a Data61 (@CSIRO) Collaborative Research Project (CRP) to carry out research on complementary aspects of this problem.
Funded by these two projects, a number of PhD scholarships and top-ups are available in my group to carry out research on fundamental information-theoretic aspects of data privacy, utility and security. I am looking for highly motivated PhD students with excellent academic record, promising research potential, and strong mathematical, analytical and communication skills. As a guide, students in top 5% of their cohort are deemed competitive. Please email me and include your CV, academic transcripts, with a brief statement of your research interests and prior experience (including publications or research thesis).
Collaborators: Prof. Medard (MIT) and her group, Prof. du Pin Calmon (Harvard) and his group, Dr. Thierry Rakotoarivelo (at Data61, CSIRO), Dr. Ni Ding (PhD graduate of ANU, formerly at Data61, CSIRO, now a Doreen Thomas Research Fellow at University of Melbourne), Mr. Yucheng Liu (ANU), .
I am actively looking for motivated students to work in this research area:
Index coding studies the problem of broadcasting a set of n messages from a server to receivers with side information. It was first studied in its modern sense in 1998 by Birk and Kol. Index coding is an NP-hard open problem with deep connections to network coding, distributed storage, interference alignment and more generally, to network information theory. It has applications in cached content distribution, satellite broadcast and even automated vehicular communication networks. Variations of index coding include pliable index coding and secure/private index coding.
In the past four years, I have been actively studying fundamental performance limits and achievable coding scheme for single-server (centralized) index coding, as well as multi-server (distributed) index coding. In distributed index coding, there is more than one server that can broadcast messages to receivers. Each server contains a subset of messages. Distributed index coding has applications in distributed storage/broadcast networks and vehicular communication networks where data is geographically or otherwise distributed among many servers. My paper entitled ``distributed index coding”, published in ITW 2016, is often cited as one of the first few papers on this topic. Index coding can be viewed as a form of source compression. I also have worked on security aspects of index coding. See for example, ITW 2018. A consolidation of our results in distributed index coding, appeared in this 2020 IEEE Trans. Inf. Theory paper.
Collaborators: Prof. Young-Han Kim (UCSD), Dr. Fatemeh Arbabjolfaei (PhD graduate of UCSD), Dr. Badri Vellambi (formerly at ANU, now at University of Cincinnati), Dr. Lawrence Ong (University of Newcastle), Prof. Joerg Kliewer (NJIT), Mr. Yucheng Liu (ANU), Mr. Yuxin Liu (formerly at ANU).
I consider motivated students to work in this research area:
Communication for omniscience (CO) is a fundamental problem in information theory and was first studied in 2004 by Csiszar and Narayan. In CO, a group of users each observe part of a source. They then wish to exchange their observations in the most communication efficient manner (least sum rate, for example) such that in the end, all users know the entire source of information. CO has deep connections to the capacity of secret key sharing, multivariate mutual information, automated information clustering, and distributed source compression. A special case of CO is called cooperative data exchange (CDE) and deals with exchange or broadcast of packets (integer units of information) between wireless clients using minimum number of network coded transmissions such that all the packets will become known to all users at the end. Although the solutions to CO and CDE problems are well formulated, their efficient computation or approximation and many other practical problems, such as fairness in the data exchange process remain open.
I have co-authored some of the first papers on CDE using network coding (see for example ITW 2010 and ISIT 2010). I also studied throughput/delay tradeoff in CDE (see for example TVT 2016.) More recently, I have contributed to the more general CO problem. My contributions include developing computationally efficient algorithms for the CO problem, as well as understanding its game theoretic aspects and fairness (see for example TIT 2018, AusCTW 2016 and ISIT 2018).
Selected Collaborators: Prof. Alex Sprintson (Texas A&M University), Dr. Salim El Rouayheb (PhD graduate of Texas A&M University, now at Rutgers University), Dr. Ni Ding (PhD graduate of ANU, formerly at Data61, CSIRO, now a Doreen Thomas Research Fellow at University of Melbourne).
I consider motivated students to work in this research area for its new applications in automated vehicular networks:
Since the publication of the landmark paper in network coding, ``Network Information Flow” by R. Ahlswede, N. Cai, S.-Y.R. Li and R.W. Yeung, network coding has contributed to our understating of fundamental performance limits of many wired and wireless data networks. It is well known that network coding can enhance the throughput performance of many wired and wireless networks, but oftentimes this comes at the cost of increased decoding latency. That can happen for example in generation-based random linear network coding (RLNC) where almost all packets can be decoded only after the transmission of a full block.
In the past ten years, I have been actively researching the fundamental tensions between the throughput and decoding latency of network coding. For example, in Netcod 2015, co-authored with Yu and Sprintson, we showed for the first time that finding a linear network code that yields the minimum average packet decoding delay is an NP-hard problem. More recently in ITW 2018, Yu and I showed how weak/strong throughput approximation is related to weak/strong decoding delay approximation and proved that memoryless network coding schemes (including instantly decodable network coding (IDNC)) cannot achieve strong or even weak throughput approximation. Through my research, I have also developed novel practical network coding schemes that can strike a better balance between throughput and decoding delay or packet drop rate in network coding (see for example TCOM 2014 paper 1, TCOM 2014 paper 2 and TCOM 2014 paper 3). Also see my paper in TCOM 2017 on feedback-free RLNC for adaptive layered video transmission.
Selected Collaborators: Prof. Muriel Medard (MIT), Dr. Sameh Sorour (Now at Queens University, Canada), Dr. Neda Aboutorab (Now at UNSW), Dr. Danail Traskov (PhD graduate of TUM, Germany).
I am less active in this area these days. However, I have co-authored a number of key papers on spatially localised filtering, equalisation, and convolution on the unit sphere and a book on Hilbert Space Methods in Signal Processing. If you are a motivated student with a good purpose and preliminary research background on this topic, please contact me.
I am less active in this area these days. However, I have co-authored a number of key papers on cooperative communications and relaying, channel estimation, finite-state Markov modeling of fading channels, cross-layer communications, and an award-winning paper on massive MINO. If you are a motivated student with a good purpose and preliminary research background on this topic, especially on information-theoretic understanding of future wireless communication networks, please contact me.