Securing Our Future

Since Concorde's retirement, the vision of two-hour international travel has been extremely limited by the immense drag and reduced lift efficiency at hypersonic speeds. In order to overcome these limitations and improve the drag quality of our peers' previous design, Cradle Computational Fluid Dynamics (CFD) simulations were employed, allowing for the iterative testing of both theoretical and experimental techniques in the aerospace field. Simulations allowed for exploration into the Busemann biplane, a theoretical concept never explored on real aircraft, allowing for the manipulation of shockwaves to near-entirely eliminate wave drag. Separately, CFD allowed for investigation into Waveriding principles; leveraging compression lift and flow separation to reduce drag and increase lift, with additional research confirmation techniques such as the aerospike and trailing blunt-edge geometries. With these new low-drag aircraft concepts showing improved drag results compared to their predecessors, this work paves the way for future commercial hypersonic aviation prospects.

Project by:

• Mitchell Guidera 
• Darcy Marker 
• James Mason 
• Samuel Green 
• Reilly Bennet 

Adelaide University is developing a human-powered submarine to compete in the 2026 European International Submarine Race. The Hull is critical: it must protect the pilot, support all subsystems, and be shaped for drag reduction and propeller efficiency. Despite strict limits on budget and time, the team produced a race-ready Hull through innovation and rapid refinement.

From the outset, a systems engineering approach brought together the pilot, gearbox, propeller, and control mechanisms in a cohesive design. The Hull’s modular sections were built from an innovative composite that combined 3D‑printed moulds with fibreglass reinforcement. These moulds were designed to bond permanently to fibreglass and allow resin infusion, improving strength while also reducing weight and cost.

Computer simulations and physical testing guided the development cycle. Computational Fluid Dynamics refined the Hull shape, while Finite Element Analysis confirmed structural integrity under race conditions. Material testing and construction trials further validated the composite’s performance.

Project by:

• Jaidyn Willis 
• Ulysses Hill 
• Thy Tao 
• Jingya Liu 

Active directory (AD) security is a complex domain where most of the current analysis requires manual work from human security expert. This project leverages Large Language Models (LLMs) to help automate the analysis of AD systems for security vulnerabilities. We focus on fine-tuning the Mixtral 8x7B  to build an AI agent that can scrutinise the AD graph database and identify potential loopholes in the system. The model shows promising results and can be implemented to reduce security assessment time, allowing security experts to focus more on the solutions rather than searching for issues in the system.

Project by: Tu Vu

The Southern hairy-nosed Wombat is a native burrowing marsupial residing in semi-arid regions of South Australia, and some areas of Western Australia. While well-adapted to harsh environments, the species is one of many facing challenges from climate and global change. Species distribution modelling allows us to predict habitat suitability for the SHN Wombat under future climate scenarios. Geographical information systems, satellite imagery and field validation are used to create models which link current wombat warren occurrences with climatic and environmental data to predict presence/absence of warrens based on biophysical requirements. By applying these models to future climate scenarios we can predict the change in habitat suitability and better determine how to help prevent extinction. This project expands upon Mike Swinbourne's PhD and Natarsha McPherson's honours and current PhD work.

Project by: Oliver Fulcher

How secure are the robots we send to space? What additional challenges does the environment of space add to the task of attacking and defending these systems? This project explores the cybersecurity challenges faced in space, in particular for autonomous systems like robots, through threat modelling, using attack trees. Additionally, this project investigates the emulation of spaceROS, a software framework that is designed to be used to develop autonomous space systems based on the  popular  open-source framework ROS (robot operating system) and exploring different potential attacks that could be carried out against the system.

Project by: Isabelle Colby

With the introduction of USB 3.0, hubs changed from a shared-bus/repeater communication model to packet-switching, making previously trivial, device-to-device monitoring difficult and under-researched. We investigate whether a malicious USB device can monitor communication timings of a system's network interface to infer a user's web-browsing activity compromising privacy on USB Hubs. Following from this, we investigate whether a malicious device can impersonate neighbouring devices by monitoring their communication timings, then using this information to pre-empt them by injecting spoofed transmissions at precise moments. This was achieved by constructing and iterating on malicious USB 3 devices with Field Programmable Gate Arrays (FPGAs) then analysing the communications with an external analyser. Side-channel timing analysis through statistical and machine learning models and manual analysis of the collected data was used to identify breaches of confidentiality and authenticity.

Project by: Sidney Bruneder

Imagine flying an aircraft and putting it through all sorts of manoeuvres and varying missions - Military aircraft do this all the time. Each different mission can change how the aircraft is stressed, and how it accumulates fatigue damage. Predicting this is usually expensive and relies on the manufacturer’s proprietary data. Our project aims to predict how much fatigue damage an aircraft accumulates during different missions and compare it to what happens in flight. This means keeping aircraft safer by developing prediction models using historical flight records, predicting crack growth in a particular flight – and in the end, comparing this to some real-world data from a real flight to see how well predictions match reality. Yes, we went up in the air with this project. So, the question is, how safe is your flight?

Project by: 

• Abigail George 
• Joanna George 
• Jake Bonner 
• Zach Al-Farabi 

Recent developments in drone warfare have demonstrated the limitations of conventional radar systems. Small UAVs with low radar cross section (RCS), slow movement, and low-altitude flight paths are difficult to distinguish from clutter. The project included the integration of continuous wave (CW) radar operating in the terahertz (THz) range with a Risley prism beam steering system. The design aimed to achieve fast scanning capabilities to enhance previous THz beam steering implementations. In parallel, micro-Doppler signature processing was used to extract drone velocity and motion patterns. Laboratory experiments using a drone mounted on a raster scanner were conducted to simulate different types of motion. Based on this work, algorithms were developed to process radar returns and extract micro-Doppler features from rotor blades and thereby determine the velocity. The investigation confirmed the feasibility of a terahertz radar system for drone interception. It demonstrated that a Risley prism achieved short scan times and large area coverage while extracting drone velocity and motion patterns. These findings highlight the potential for developing short-range defense systems capable of detecting and classifying small drones. 

Project by:

Mariam Abd-alghany 
Blake Hore 
Vy Huynh 

Militaries today are moving toward using many small, low-cost drones instead of relying only on large aircraft with human pilots. Our sponsor Praetorian Aeronautics asked us to explore a new type of drone that has two main wings (a tandem wing) and can take off vertically like a helicopter or normal drone but also fly around like a regular airplane. The aim of our project is to create a design process that can quickly turn ideas into working products for this unusual aircraft type. We began with simple maths and physics calculations, then moved to more advanced methods like computational fluid dynamics (CFD) and wind tunnel testing. At the same time, we built digital twin models, which are computer versions of the drone that bring together all our test data. These models allow us to simulate and predict how the aircraft performs across its different flight modes.

Project by:

Fikry Aljawahari 
Joshua Carrigan 
Kieran Livingstone 
Vasudev Nair 

People can trick AI models and produce harmful content using clever Jailbreak prompts. As AI is used widely, these attacks pose significant risks to users and organisations. This project builds a real-time team of AI agents that work together as a prompt filtering system and analyse the prompt. A leader agent and along with three agents, analyses the prompt. If the prompt seems malicious, the system blocks it and informs the user, while the safe ones go through. In testing across various cases, the system successfully prevents harmful outputs while remaining user-friendly, making the AI interactions safer without interrupting the daily normal use. In continuing work, we expand the system to multiple domains.

Project by: Jyothis Joy

The Adelaide University Submarine Team (AUST) is set to become the first Australian team to compete in the European International Submarine Race in 2026, with intention to compete annually on an ongoing basis. This global competition challenges students to design, build and race their own human-powered submarines. This project focuses on a critical part of submarine performance, the propeller. An investigation into propeller performance in the context of a human-powered submarine was conducted, including a variety of innovative propeller shapes that will improve performance at future races. A combination of computational fluid dynamics simulations and physical testing has been undertaken to compare the performance of each propeller design. We present a recommendation of propellers that show strong potential for future application by the AUST in future International Submarine Races.

Project by: 

• Michael Smallridge 
• Keifer Potts 
• Shaghaf Abumustafa 

With the increasing use of drones on the battlefield for strike and reconnaissance missions, the need for an affordable and effective countermeasure is growing. Jamming the communications between the drone and its controller is a viable solution to the drone menace. The project, sponsored by industry partner Raytheon Australia, aimed to understand how drone communications work, and then design bespoke jamming techniques against a selected range of commercial-off-the-shelf drones. A test environment was developed to allow for the recording of drone radio signals and the transmission of jamming signals. Software based solutions, enabled by Raytheon supplied hardware, were developed alongside rigorous test procedures to experimentally verify jamming techniques across a spectrum of drone types. The project results include a database of drone communications signal characteristics and suite of tested jamming techniques. This outcome provides the basis for the development of a counter drone system.

Project by: 

• Jacob March 
• Jorgie McKenzie 

High-powered laser systems are essential in a wide range of fields, including defence, remote sensing, medical procedures, manufacturing, and laser fusion. Achieving high-power lasing can be accomplished through electro-optic Q-switching techniques. However, existing Q-switch driver systems are often costly, complex, or unreliable — where modularity and thermal stability are essential. The goals for the project were to repair and improve an electrical Q-switching driver which was first built within the University of Adelaide. A combination of diagnostic, troubleshooting, measuring, and retrofitting was performed to repair the driver and improve its capabilities. We present a device that is capable of supplying high voltage and high frequency pulses to an optical crystal for the purpose of generating high powered laser pulses. This project demonstrates that a cost-effective, reliable, and user-friendly Q-switch driver can be built from low-cost components.

Project by: Liam Siles Wright

Imagine a swarm of explosive kamikaze drones launched from a shipping container off the Australian coast. Is our military prepared to respond? With FPV drones becoming an increasingly devastating tool in modern warfare, a fast, affordable and scalable defensive solution is urgently needed.

We designed and built the PI-UAV – a Physical Interception Unmanned Aerial Vehicle – to intercept and neutralise hostile drones by colliding mid-air, then returning safely for another mission.

Our prototype uses aerospace-grade, 3D-printed titanium to create a lightweight yet high-strength frame capable of withstanding repeated impacts. The result is a portable, cost-effective system offering a promising countermeasure to today’s RF-silent, explosive UAV threats. 

Project by:

• Aleksander Pollok 
• Oskar Jakubowski 
• Brody Moylan 

This project investigates the scope of trained AI model to understand threats in a computer network, also how Large Language Models (LLMs), like GPT and LLaMA, can improve network intrusion detection. With current trends, traditional security tools find it difficult to cope up with today's complex, fast-evolving attacks, often missing threats or generating too many false alarms. LLMs offer a new solution by interpreting network logs and behaviours like sentences, learning to spot danger more intuitively. We tested LLM-powered detectors on real-world security datasets and compared them with classic methods. Results show that LLMs significantly improve detection accuracy and recall, especially for subtle and novel attacks. The proposed system balances performance with efficiency, by combining numeric and textual features, and optimizing it for edge deployment. To protect real-world digital infrastructure, this work paves the way for smarter, adaptable, and more trustworthy cybersecurity tools.

Project by: Tuhin Anand 

Secure systems of the future will be powered by light. Photonic integrated circuits (PICs) put lasers, amplifiers, and waveguides onto a chip smaller than your fingernail - delivering robust and portable precision measurement capabilities.

This project focused on the semiconductor optical amplifier (SOA), a component that boosts signals while keeping them sharp under demanding conditions. Operating at 780 nm for miniature atomic clocks, we carried out an initial characterisation of its frequency modulation bandwidth and extinction ratio.

Preliminary results revealed record-high signal clarity (33–38 dB), setting benchmarks for clean, reliable on-chip control. We also quantified trade-offs between High Power and Low Noise designs, linking device architecture to practical field applications.

Project by: Arth Tiwari

This project researched modelling approaches for the computational fluid dynamics (CFD) analysis of hypersonic vehicle thermal protection systems (TPS) using desktop hardware and ANSYS Fluent. Hypersonic vehicles fly at speeds greater than Mach 5, five times the speed of sound, which presents unique engineering challenges, such as high temperatures generated due to high friction and pressure gradients. Current modelling approaches require excessive and expensive computing power and can be inaccurate. New modelling approaches for hypersonic flight were validated against five existing studies to assess accuracy. These setups were refined to reduce simulation run time and computing power, with the findings summarised into guidelines. A parametric study was then performed using these guidelines to assess heating trends across a range of shapes and operating conditions. The successful development of CFD guidelines will accelerate the advancement of hypersonic research through recommendations of geometries, domains, meshes, and solver set-up.

Project by:

• Jacob Casey 
• Samuel Bush 
• Owen Matz 
• Markus Chandler 
• Elise Myatt 

Any person who has sat aboard an aircraft has experienced it and to some extent, dreaded it. Going through security checkpoints. They're stressful, busy and bog you down because you accidentally forgot to remove your belt. We want a better, more efficient and more accurate solution to address this. To that end, we are investigating the use of Terahertz waves to replace existing security systems using millimetre-wave scanners. With faster scanning, better resolution and no issues with harmful radiation, it's a promising emergent solution. This project is developing a body scanning system utilising frequency-modulated continuous wave (FMCW) radar scans directed through a galvanometer to scan a person for concealed objects. This essentially converts a manual pat-down search into an automatic one that is quick, accurate and non-invasive, whilst not compromising on security. This can streamline security checkpoints to be more efficient and make the travel experience much less stressful.

Project by:

• Oliver Moors 
• Jonathan Miteff 
• Jacob Maegraith 
• Joshua Vidale 

Modern cyber-attacks have become increasingly sophisticated and persistent, additionally the activity of cyber-criminals is easily overlooked amongst overwhelming volumes of network traffic. The effects are tangible: individuals are targeted in data breaches, and headlines frequently report high-profile cyber-crimes on businesses. Our project proposes harnessing artificial intelligence to detect malicious behaviours hidden within network logs. We modelled typical cyber-attack phases on a simulated corporate network made of seven devices to generate training data. Using machine learning and statistical techniques, the malicious behaviour was identified from amongst the normal traffic and process logs. The techniques we developed were incorporated into an analytics tool to allow an analyst to string together pipelines and interpret results visually. The ability to identify threats at each stage of the attack cycle may help us to stop attackers in their tracks by preventing a breach as it unfolds.

Project by: 

• Byron Betts 
• Sophie Davidson 
• Liam Jacobsen
• Austen Lindsay
• Dev Mer

3D printing, most often associated with plastic extrusion, has evolved to be able to print the next best thing: metals. Using this technology, intricate structures can be printed from aluminium without the need for multiple parts, reducing waste. Whilst this appears to be the perfect manufacturing method for aluminium parts, it is costly and the final material properties are not well understood. Given this lack of knowledge and the Defence Science and Technology Group's interest in this technology, our goal was to determine the material properties of printed aluminium. To accomplish this, samples were printed with varying print parameters and a mild heat treatment cycle, then analysed to determine the microstructure, porosity, mechanical properties and surface roughness. Recommendations have been made to DSTG regarding printing and heat treatment conditions to create ideal material properties for future production of parts with this technology.

Project by: 

• Thomas Hall 
• Eli Bonner 
• Alexander Nacov 
• Tayla Grant 

The Internet is an essential system that is fundamental in our everyday lives, yet its structure and behaviour is only partially understood, largely due to its size and complexity. This project aims to better understand this structure by creating more detailed maps of Australia’s Internet specifically. This approach involves collecting real information from different Internet providers, including their networks, the cables and facilities they use, and how traffic flows through these points. By considering many networks, this allows us to gain a more comprehensive understanding of the network, and to identify potential bottlenecks and locations that could cause major outages if they shut down (from natural disasters or targeted attacks). The outcome is a comprehensive, multilayered view of Australia’s Internet infrastructure, that helps us understand the system better and informs future steps to make the network more reliable, efficient, and resilient.

Project by: Benjamin Lang

Human-powered submarines are designed for international competitions where a single pilot pedals underwater to propel the vessel forward. While propulsion is provided by the pilot, steering remains a critical challenge, as the submarine must safely manoeuvre up, down, left, and right in a dynamic underwater environment. The aim of this project was to design an effective control system to enable precise and reliable steering without the use of conventional engines or electronics.

To achieve this, hydrodynamic principles were applied to explore how water flows across different fin shapes and sizes. Computer-based simulations were then carried out to test and compare control surface designs under realistic operating speeds.

The outcome of this project is the development of a control system design that balances responsiveness, stability, and ease of use. This system is now ready for integration into the submarine, supporting the University of Adelaide's entry in the international submarine race.

Project by:

• Sung Hoon Ok
• Toby Willis

Hidden software is baked into modern CPUs, and it can only be read directly from the chip. This code that bridges hardware and software underpins computer security; failures here ripple across everything built on top, from personal devices to global cloud services.

We physically reverse-engineer chips to expose and image the memory that stores CPU firmware (microcode). For the chips to be suitable for scanning electron microscopy (SEM), layers over 100x smaller than hair must be removed and kept consistently flat and damage-free. This reveals bright/dark cells that can be translated into data bits.

We compare chemical etching with controlled polishing, and their imaging results. The project establishes a repeatable workflow for CPU imaging, produces SEM images, and successfully extracts portions of microcode from AMD Zen. While a full read-out was not achieved, the results demonstrate feasibility and identify next steps: higher-uniformity delayering, automated decoding and improving imaging contrast.

Project by: Nicholas Hassan

Memory corruption bugs are a consequential attack vector for malevolent actors. Address sanitizers, traditionally a compiler plugin executed at runtime, aim to detect and prevent memory safety issues in programs. An address sanitizer inserts checks around memory accesses, and crashes the program upon detecting an improper memory access. Sanitizer are especially helpful in automated testing methods, such as fuzzing, to detect bugs otherwise missed and when triaging discovered crashes. This study seeks to build the first sanitizer designed specifically for monolithic firmware. Sanitizers are operating system (OS) and instruction-set-architecture (ISA) dependent, making designing techniques for firmware which is compiled for various OSs and ISAs naturally a challenge. This research project develops techniques for an effective and efficient sanitizer supporting binary-only testing where source code may not be accessible.

Project by: Douglas Cooke