Deakin-Coventry Cotutelle - Protective Security and Defence

Protective security and information: Understanding and mitigating potential security threats that emerge from businesses online

Advancements in data and technology, coupled with how organisations have evolved their businesses in response to Covid 19, has enhanced the amount of information that organisations have put online. While this has been intended to enhance things such as user experience, accessibility, and engagement with customers and visitors, it may also have exposed potential vulnerabilities that could jeopardise the safety and security of people, and provide terrorists and other criminals with significant knowledge without having to leave their home or have significant technical skills. In contrast, it may also present opportunities for first responders, government, or the risk management and insurance sector.

The purpose of this research is to address this gap in the current knowledge base by first, focusing on how this information could be an important enabler for understanding, countering and mitigating particular security threats, and second by exploring how the resilience of organisations, places and individuals could be compromised through certain unintended consequences. Outcomes from the research could include the research and development of a co-created framework for understanding and mitigating risks associated with including potentially sensitive information online, technical approaches for first responders to identify such information, or other methods that could mitigate the risk for the organisation that has included the information online or for third parties.

Leadership and decision-making: The potentiality of emerging technology for effective and efficient responses to security-related incidents

In organisations, fire drills are embedded in policies that protect employees against the potential risk of fire. While fire continues to be a threat, it is well understood, and most organisations are prepared to respond. However, other threats- like terrorism, insider risk, and espionage - are less understood and continue to evolve, leaving organisations unprepared to respond to events that challenge the security and safety of the organisation and its employees. In dealing with such incidents it is fundamental that senior level executives and board members understand their role, have the necessary skills, and regularly test against different security-related scenarios effectively and efficiently. Emerging technologies provide the potential bridge between learning and real-life experience in a risk-free environment and offer senior leaders the ability to understand unintended consequences of decisions, inject different and complex challenges during each simulation, as well as understand the overall performance of the senior leaders.

This research will research, develop, test and refine cutting edge approaches for simulating security incidents in non-security critical organisations/sectors (such as education institutes/places of worship/hospitality/etc) and is designed to enhance board level decision-making. These methods may include the use of emerging technologies such as virtual/mixed/augmented realities, artificial intelligence and machine learning and the research.

Towards safe virtual security assessments

There is a growing need to reduce the amount of time that it takes to assess the security and resilience of crowded places and/or critical national infrastructure. One potential approach is through the development and testing of a virtual environment that allows assessors (such as Counter Terrorism Security Advisors) to reduce the amount of time on site while ensuring that the integrity and proportionality of security assessments are not compromised. This would need to be grounded in digital security by design principles and potentially co-created with security assessors.

Securing political and governmental actors from targeted cyber attacks

Involvement in politics carries with it increased risks to personal cyber-security. In recent years political elites such as former British Prime Minister Liz Truss, the Prime Minister of Spain, and members of Cabinet in Greece have had their personal devices compromised. Cyber-security for those who work in political systems, either as elected representatives or civil service roles is vital for the overall strength of democracy. This project examines strategies to improve the cyber-security of political and governmental actors.

Non-state actors in cyber policing: Assessing the role of private investigators

This project examines the emerging field of private investigators in cybercrime. It aims to map this field, examine the types of cybercrimes they investigate, how they interact with state actors and cybercrime victims. It may adopt an international and comparative perspective.

A framework for data privacy protection in immersive technologies

Data leakage is an emerging problem in immersive technologies (VR/XR). This project aims to design a privacy preserving framework for immersive technologies and to examine the feasibility of adopting the framework to guide privacy by design.

Enhancing Protective Security in E-Safety: Safeguarding Digital Users in the Online Environment

This PhD project aims to address the challenges and develop strategies for enhancing protective security in the context of e-safety, focusing on safeguarding digital users in the online environment. The research will investigate the evolving landscape of digital threats, including cyberbullying, online harassment, identity theft, and malicious content dissemination. The project will involve analysing existing e-safety frameworks, policies, and technologies, and identifying gaps and limitations. Innovative approaches will be explored to develop proactive measures such as user authentication methods, content filtering algorithms, and effective reporting mechanisms. The study will contribute to the development of comprehensive and effective protective security measures to promote e-safety and protect individuals in the digital realm.

Enhancing Risk Management and Decision-Making through Trust-Based Collaborative Systems: Mitigating Uncertainties in Protective Security and Defence

The thesis will aim to enhance risk management and decision-making in protective security and defence by implementing trust-based collaborative systems. In the face of uncertainties in evolving threats, such systems provide a framework for effectively mitigating risks and making informed decisions. By fostering trust among stakeholders and enabling trusted collaboration, these systems facilitate sharing of information, expertise, and resources.

The research will explore various techniques and methodologies for establishing trust, analyzing uncertainties (e.g., Epistemic and Aleatoric), and leveraging collaborative approaches to enhance the overall effectiveness of risk management and decision-making processes in the context of protective security and defence.

Zero Trust based Continuous Authentication towards Protective Security

Australian Government has developed Protective Security Policy Framework (PSPF) ^[1] to provide guidance and establish policies for protecting government assets, information, and people. One of the key requirements of PSPF is safeguarding data from cyber threats. To mitigate this challenge, Multi-Factor Authentication (MFA) is highly effective in preventing unauthorized access attempts by adversaries. While traditional authentication models that rely on static credentials are increasingly vulnerable to sophisticated attacks, Zero Trust Architecture (ZTA) based authentication addresses these challenges by employing continuous verification, contextual factors, and adaptive authentication methods.

This project aims in developing a continuous authentication platform that offers identity assurance and compromise detection, alerting IT security personnel to potential threats in real time. As a part of this project, a lightweight continuous authentication protocol will be developed to authorize users to access resources using at least one device-level authentication signal alongside an identity signal through behavioural biometrics. Such signals can also be used as a highly phishing-resistant MFA option that does not require the use of a mobile device, one-time code, or push notifications which makes it highly suitable for the usages in Government buildings and premises.

^{[1] The Australian Government's Protective Security Policy Framework, available Online https://www.protectivesecurity.gov.au/ (Accessed 6th July, 2023)}

Enhancing zero trust control plane resilience to cyberattacks

Zero trust architecture depends on several trusted components in the control plane. These components are erroneously assumed that they cannot be compromised by attackers, but recent works have identified possible threats and attacks to zero trust control plane. This project aims to investigate vulnerabilities and threats to zero trust control plane and develop theoretically and practically sound algorithms and methods that ensure zero trust control plane resilience to cyberattacks.

Improving the Robustness of Deep Learning Models for Tabular Datasets

Robustness of Artificial Neural Networks (ANN) to benign or malicious input manipulations such as covariate/concept drift and adversarial attacks on tabular datasets has acquired a lot of attention in the last few years. This is because ANN models are deployed in many applications and constitute important component of critical decision making. It has been shown recently that by processing ANN’s input, one can control their robustness to both benign and malicious transformations as evident by some of key state of the art techniques such as D2A3, gD2A3, D-Layer, etc. proposed by our group ^[1][2]. This project proposed in this work is the extension of these techniques.

This project will build on D-Layer – which builds an automatic equal frequency discretization of input layer in ANN models. Despite its effectiveness, D-Layer cannot handle concept drift. This project will explore the integration of Hoeffding Bound trees to integrate incremental learning approach of Decision Trees in ANNs. The second stream of research in this project will explore learning better representation of cut-points. Existing methods representation of cut-point is some threshold in an input space. By leveraging embedding layers, we aim to learn a representation that is based on latent space representation – hoping to obtain a cut-point that is more resistant to input manipulations. Finally, this project will study the integration of generative models in ANNs to ward-off any adversarial attacks, as an extension of gD2A3 approach. The model will explore developing a Bayeisan Network for the intermediate layers of ANN.

^{[1] Zaidi et al. Improving Neural Network’s Robustness on Tabular Data with D-Layers, DMKD 2023
[2] Zaidi et al. Leveraging Generative Models for Combating Adversarial Attacks on Tabular Datasets, PAKDD 2023}

Privacy-Preserving Federated Learning for Intrusion Detection in IoT Networks

This project aims to enhance privacy-preserving intrusion detection by leveraging federated learning in IoT networks. The objective is to develop a distributed and collaborative learning framework that enables IoT devices to collectively learn intrusion detection models without compromising sensitive data. By aggregating local model updates instead of sharing raw data, this approach ensures privacy preservation while improving the accuracy and robustness of intrusion detection systems in IoT environments. The project will explore techniques such as differential privacy, secure aggregation, and cryptographic protocols to address privacy concerns in federated learning for intrusion detection.