Zahir Tari

Supervisor interest areas: Cybersecurity. Performance. Scalability. Reliability

Supervisor projects

• Effective Countering cyber threats over Smart-Grid Infrastructures.  Smart-grids are the cornerstones of future energy infrastructure. Today's ageing power grids have had their inadequacies exposed by challenges such as: the rising cost of fuels, need for lower greenhouse emissions, and the rise of renewable energy. Tomorrow’s Smart Grids will incorporate increased sensing, communication, and distributed control systems to accommodate renewable generation, EV (Electric Vehicle) loads, peak demand reduction and energy loss minimization. Of the various impediments affecting smart-grids, security is the foremost concern for all stake holders, estimating economic losses at $6B annually. This project aims to research a distributed and multi-granular framework for effective data security that is achieved through anomaly detection over smart-grid sensing systems. This will evaluate cyber threats from a data centric point of view for effective protection, and the framework will be compatible with and cater to the dynamic and non-homogeneous data typically present in multi-faceted smart-grid environments.

• Preventing Exfiltration of Sensitive Data by Malicious Insiders or Malware. Data exfiltration is the unauthorised leakage of data from computers by sophisticated malware and malicious insiders. Data exfiltration is a serious problem since it may have catastrophic effect on businesses, governments as well as individuals if such exfiltration involves sensitive data. Examples include exfiltration of data involving business inventions, national intelligence, classified research, individual’s credit card and biometric profile. Specifically, data exfiltration has resulted in huge economic losses as well as unprecedented breaches of national security. A study by the Ponemon Institute reported that the average per-incident cost of reported data leakage by businesses was $14 million, and the number of reported data breaches surpassed all previous years. The aim of this project is to develop solutions to detect sensitive data exfiltration attempts by malware, as well as human users, and block those attempts without affecting legitimate users’ normal usage of computers.

• Low-Latency High-Throughput Computational Models for Heavily Data-Driven. The need to process a huge volume of data during in a small amount of time is dramatically increasing especially as the size of the data moves into Exabyte in the near. While use of such applications was previously confined to the finance sector, it is becoming now prevalent in almost every industry where analytical processing over massive data sets can solve business problems. To meet such low-latency requirements of data mining and machine learning applications, datacentre providers must expand the computing capacity of the underlying infrastructure by exploiting graphics processing units (GPU) and Field Programmable Gate Arrays (FPGAs) as new hardware accelerators, the so-called heterogeneous datacentres. However, there is no mechanism that can appropriately project the complex characteristics of modern applications emerging in enterprise/scientific domains into the available computing capacity of a system with hundreds or thousands of heterogeneous computers. Additionally, using existing resources allocation solutions in heterogeneous datacentres result in significant resource wastage. The general aim of this project is to investigate innovative solutions/methods to control and to make use of the capabilities of the new hardware accelerators in a heterogeneous computing systems to substantially enhance the resource efficiency when running data-driven applications.