Machine Learning and its applications to large scale networks

In this project, we consider using machine learning (e.g., reinformacement learning techniques like multi-armed bandit theory, markov decision processes, game theory,..etc to undersand and design robust network protocols or architectures to address the future networking needs. Some of the issues that we are studying are:

  • Multi-path TCP controls and its interplay with SDN networks
  • Multi-channels access and resource allocation for future wireless networks
  • Crowdsourcing and resource allocation
  • Multi-rate video streaming and its protocols

We take both the analytical as well as system approach in examining the above issues.

Data Analytics for large scale networks (e.g, online social networks)

We take both the network science and distributed systems approaches to examine issues of online social networks. Some of the issues that we are studying are:

  • influence and information spreading;
  • advertisement and recommendations;
  • privacy and security;
  • complimentary role of online social networks and P2P technology;
  • dynamics and evolution of OSN;
  • deep learning approach in motif discovery
  • information exploration and data mining
  • Storage and index issues for large scale graphs

We take both the analytical as well as system approach in examining the above issues.

Network economics

In recent years, there has been a lot of interests on the multi-disciplinary research on Internet and its economic issues. To handle this line of research, we propose to use game theory, pricing theory, approximation algorithm as well as machine learning techniques to understand the intricacy and the inter-dependency of networks/economics/social issues. We have been looking at issues like:

  • Bundling sales for information goods
  • Efficient approximation algorithms and data mining approach for bundling products
  • Product selection and advertisement
  • Pricing and incentive issues for online rating systems
  • Use pricing to provide service differentiation;
  • Profit re-distributions for ISPs;
  • Product Advertisement;
  • Interactions of competing ISPs;
  • Cooperation among ISPs and content providers;
  • Charging, billing and accounting for future networks;

System/Network Security

In this project, we examine various system security issues which are related to mobile computing and Internet-of-Things (IoTs) such as smart homes, smart offices or building. Note that for this line of research, it requires students to have a good background on software security, compiler, programming languages and operating systems. We have been examing issues like:

  • Malware detection (both static and dynamic analysis)
  • Information-flow tracking system for Android systems
  • Self-randomizing address space layout to defend against buffer/memory overflow attacks
  • Disk and information encryption for smartphones
  • distributed and randomized algorithms to detect sybil and dishonest network users.

Designing Future Network Architecture

One can forsee the huge increase of future network traffic and this requires some careful and fundamental design changes in our current network architectures. In this project, we examine various fundamental issues which can change the landscape of future network architecture:

  • Distributed caching algorithms
  • Distributed hierarchical pricing for cache resources
  • Bandwidth brokerage for Inter-domain traffics
  • Importance sampling and Markov chain Monte Carlo fast simulation for network architectures
  • Distributed SDN control and its implication to data center networks (DCNs)

Internet-of-Things (IoTs): Theory and applications

In this project, we consider both the traffic theory and applications for IoTs. In particular, we focus on the issues on security, scalability, software reuse and programmability. We consider:

  • System and software component security
  • Audit and traceability of apps on IoT environment
  • Dynamic traceback and analysis for component software
  • Micro-payment algorithms and their scalabilty, robustness and security issues.