As Australia accelerates its transition to renewable energy, the security of the systems that keep our power grid running has never been more important. Digital substations, the backbone of modern electricity networks, offer enormous advantages in efficiency, automation and real‑time monitoring. But with this digital transformation comes a new frontier of cyber risk.
One of the most concerning threats is GOOSE spoofing, where attackers manipulate the high‑speed communication messages that control protection equipment inside substations. A successful attack could disrupt grid operations, damage equipment or trigger widespread outages. Protecting these systems requires cybersecurity solutions that are as fast and adaptive as the threats themselves.
Researchers here at QUT’s Energy Transition Centre have taken a major step forward in addressing this challenge. In a recent study published in IEEE Xplore, Gowri Sankar Ramachandran, Mahinda Vilathgamuwa, Trinal Fernando, Dhammika Jayalath and colleagues demonstrate how deep learning models, specifically DNN and LSTM architectures, can detect GOOSE spoofing attacks with high accuracy and real‑time responsiveness.
Their approach combines traditional message‑level attributes with physical system measurements, creating a multi‑modal dataset that allows the models to recognise subtle anomalies that rule‑based systems often miss. This fusion of cyber and physical data gives the models a richer understanding of what “normal” looks like inside a digital substation – and when something is wrong.
The results are promising. The models show strong potential for scalable, resilient anomaly detection that can adapt to evolving threats. For critical infrastructure operators, this kind of capability is essential. As the grid becomes more decentralised, more automated and more digitally interconnected, cybersecurity must evolve just as quickly.
This research lays important groundwork for the next generation of secure, intelligent power systems. It highlights the role advanced AI can play in safeguarding the infrastructure that underpins Australia’s energy transition, ensuring that as we modernise our grid, we also strengthen its resilience.
Want to learn more?
If you’d like to explore the full technical details, the complete research paper is available via QUT ePrints: Anomaly Detection for GOOSE Spoofing Attacks in Digital Substations Using Deep Learning Models: A DNN and LSTM Approach