Multimodal Deep Learning for Phishing Detection

About the research project

Phishing is one of the most persistent and rapidly evolving cybersecurity threats, in which attackers deceive individuals or organisations into revealing sensitive information, transferring money, installing malicious software, or accessing fraudulent websites. The increasing availability of generative artificial intelligence (AI) has further amplified this threat by enabling attackers to create highly convincing, personalised, multilingual, and context-aware phishing messages and websites at scale. Consequently, conventional phishing detection techniques based primarily on predefined rules, blacklists, URLs, or manually engineered features may struggle to identify sophisticated and previously unseen phishing attacks.

Significant research has investigated machine learning and deep learning for phishing detection. Existing approaches have applied decision trees, support vector machines, convolutional and recurrent neural networks, and, more recently, Transformer-based and large language models to analyse phishing emails, URLs, web pages, and other communication characteristics. However, important limitations remain. Many existing studies focus on a single data modality, rely on relatively static datasets, or evaluate models under controlled conditions. Their ability to detect emerging and AI-generated phishing attacks, generalise across domains and languages, resist adversarial manipulation, explain detection decisions, and maintain performance as phishing strategies evolve remains insufficiently investigated.

This PhD project proposes to develop a robust, explainable, and adaptive deep learning framework for phishing detection in the generative AI era. The anticipated methodology will comprise five stages. First, representative datasets of phishing and legitimate communications will be collected and constructed, incorporating textual content, URLs, HTML/webpage characteristics, screenshots, and, where feasible, AI-generated phishing samples across different domains and languages. Second, deep learning and Transformer-based models, including appropriate large language models, will be developed to learn semantic, visual, structural, and behavioural features. Third, multimodal fusion techniques will be investigated to determine how complementary information from different sources can be effectively integrated. Fourth, robustness and adaptability will be studied through adversarial testing, cross-domain and cross-language evaluation, few-shot learning, and continual learning to assess detection of previously unseen and evolving phishing attacks. Explainable AI techniques will also be incorporated to provide interpretable evidence supporting detection decisions. Finally, extensive experiments will compare the proposed approaches with established machine learning and deep learning baselines using precision, recall, F1-score, false-positive rate, robustness, generalisation, and computational efficiency.

The expected outcome is a scientifically validated phishing detection framework that can identify sophisticated and emerging phishing attacks more accurately, robustly, and transparently than existing approaches.

Primary supervisor

Meet Doctor Shuxiang Xu

Funding

Applicants will be considered for a Research Training Program (RTP) scholarship or Tasmania Graduate Research Scholarship (TGRS) which, if successful, provides:

• a living allowance stipend funded by University of Tasmania of $34,315 per annum for 3.5 years (PhD) or 2 years (Master by Research)
• a relocation allowance of up to $2,000
• a tuition fees offset covering the cost of tuition fees for up to four years (domestic applicants only)

A tuition fee offset may be offered to eligible international applicants following competitive assessment 

As part of the application process you may indicate if you do not wish to be considered for scholarship funding.

Other funding opportunities and fees

For further information regarding other scholarships on offer, and the various fees for undertaking a research degree, please visit our Scholarships and fees on research degrees page.

Eligibility

Applicants should review the Higher Degree by Research minimum entry requirements.

Ensure your eligibility for the scholarship round by referring to our Key Dates.

Additional eligibility criteria specific to this project/scholarship:

• Applications are open to Domestic/International/Onshore applicants

Selection criteria

The project is competitively assessed and awarded. Selection is based on academic merit and suitability to the project as determined by the College.

Additional essential selection criteria specific to this project:

• Applicants should demonstrate sound knowledge and practical expertise in machine learning processes, including fundamental concepts such as supervised and unsupervised learning, model training and evaluation, feature engineering, and performance assessment.
• Applicants should also possess sound programming skills in at least one of Python, C, or Java, with the ability to develop, test, debug, and implement software solutions.

Additional desirable selection criteria specific to this project:

• Applicants should have knowledge and practical experience in deep learning, including neural network architectures, model training, optimisation, and evaluation.
• Familiarity with large language models (LLMs) and their applications in natural language processing would be highly desirable.
• Experience with deep learning frameworks such as PyTorch or TensorFlow, as well as knowledge of Transformer-based architectures, would be advantageous.
• An understanding of applying deep learning techniques to cybersecurity, text classification, anomaly detection, or related problems would also be beneficial.

Application process

1. Select your project, and check that you meet the eligibility and selection criteria, including citizenship;
2. Contact Doctor Shuxiang Xu to discuss your suitability and the project's requirements; and
3. In your application:
   • Copy and paste the title of the project from this advertisement into your application. If you don’t correctly do this your application may be rejected.
   • Submit a signed supervisory support form, a CV including contact details of 2 referees and your project research proposal.
4. Apply prior to 1 October 2026.

Full details of the application process can be found under the ' How to apply ' section of the Research Degrees website.

Following the closing date applications will be assessed within the College. Applicants should expect to receive notification of the outcome by email by the advertised outcome date.

Apply nowExplore other projects