Real-Time SCAM SMS & Call Prevention

Engineering, IT, Mathematics and Statistics

ABOUT THE INDUSTRY PARTNER

Telstra is Australia’s leading telecommunications and technology company, offering a full range of communications services and competing in all telecommunications markets. In Australia Telstra provides around 22.5 million retail mobile services and 3.4 million retail bundle and data services.

WHAT’S IN IT FOR YOU?

  1. Career Advancement: A successful project could lead to further opportunities in the Telco or the broader telecom and cybersecurity industry. Proving skills on a high impact use case makes the student competitive for senior analytics, data science, or research roles.
  2. Industry Mentorship: The student will regularly interact with the Telco’s fraud prevention and analytics teams, receiving mentorship on real-world data challenges, big data infrastructure, and system deployment within a major telecom network.
  3. Hands-on Real-World Impact: By working on the frontline against scams, the student will help protect Australians from fraudulent activities, contributing to societal good and learning how to bridge cutting-edge research with operational solutions.
  4. Access to Industry Data: The student will gain access to large-scale telecommunications data—such as call detail records (CDRs), SMS logs, and network signalling data—to develop and test cutting-edge fraud detection approaches.

RESEARCH TO BE CONDUCTED

Telstra play a pivotal role in blocking these scam communications before they reach consumers. As part of its “Cleaner Pipes” initiative, Telstra reported blocking over 185 million suspected scam calls in 2022 alone, averaging 15 million scam calls blocked each month.
Despite these efforts, scammers continually adapt. A more dynamic, intelligence-driven system that uses statistical modeling, graph-based techniques, and machine learning is needed to proactively detect new scam behaviors with minimal delay.

Objectives:

  1. Develop a Real-Time Detection Pipeline: Design a system that identifies scam calls and SMS in near real-time, ideally before the call connects or the message is delivered.
  2. Integrate Statistical and Graph Methods: Leverage Bayesian networks for probabilistic reasoning, along with graph-based clustering and anomaly detection to uncover suspicious patterns (e.g., fraud rings).
  3. Apply Machine Learning: Train and evaluate supervised/unsupervised ML models on large-scale telecom data, ensuring high accuracy and low latency.
  4. Explainability: Incorporate interpretable AI techniques so that the system’s decisions can be understood and validated by fraud analysts, compliance teams, and potentially regulators.

SKILLS WISH LIST

If you’re a postgraduate research student and meet some or all the below we want to hear from you. We strongly encourage women, indigenous and disadvantaged candidates to apply:

  1. Machine Learning & Statistics: Familiarity with classification, clustering, anomaly detection, and probabilistic reasoning (especially Bayesian networks).
  2. Graph Analytics: Experience with graph theory, clustering, community detection, and/or graph neural networks (NetworkX, igraph, PyG, etc.)
  3. Programming & Data Science: Proficiency in Python (pandas, NumPy, scikit-learn, PyTorch/TensorFlow), data engineering (SQL, Spark), and version control (Git).
  4. Cybersecurity & Telecom (Nice-to-Have): Understanding of telecom fraud, scammer tactics, and relevant cybersecurity frameworks.

Interdisciplinary Approach & Potential for Multiple Students
Given the cross-disciplinary mathematical approach required for developing an effective realtime scam detection system, the project may benefit from the involvement of two PhD students with complementary expertise. The core methodologies span multiple domains, including:

  • Graph Analytics & Network Theory – Modelling call/SMS interactions as graphs, performing community detection, anomaly detection, and graph-based fraud clustering.
  • Bayesian Networks & Probabilistic Inference – Utilizing Bayesian models to handle uncertainty in scam detection, incorporating probabilistic reasoning to improve real-time decision-making.
  • Machine Learning & Real-Time Processing – Designing, training, and optimizing supervised/unsupervised learning models to classify fraud patterns at scale.

To ensure depth and effectiveness in each of these areas, Telstra is open to accommodating two PhD students under the APR Intern program if required. A possible division of expertise could be:

  • Student 1: Specializes in Graph Theory, Graph Analytics, and Community Detection, focusing on scammer behaviour clustering, anomaly detection, and fraud network visualization.
  • Student 2: Specializes in Statistical method, Bayesian Networks, Time series Analysis in developing robust scam classification systems and real-time decision frameworks.

By integrating insights from both mathematical disciplines, the project can achieve a holistic, high-accuracy detection system that leverages the strengths of both graph-based approaches and probabilistic AI techniques.

RESEARCH OUTCOMES

1. Functional Real-Time Fraud Detection System

  • A prototype that processes incoming call/SMS data in a streaming environment, computes relevant features, and classifies fraudulent communications with minimal latency (goal: under 1 second).
  • Incorporation of graph-based metrics (e.g., centrality, community detection signals) into the decision-making process, alongside probabilistic reasoning from Bayesian networks.

2. Performance Benchmarks

  • Detection Accuracy: Achieve high true-positive rates (TPR) and minimize false- positive rates (FPR), ensuring legitimate calls/SMS aren’t over-blocked.
  • Latency Thresholds: Demonstrate that the approach can scale to millions of daily call/SMS records with decisions made before connection.
  • Explainability: Provide insights into why a call/SMS was flagged or allowed, building trust and enabling compliance/regulatory reviews.

3. Scalability & Deployment Report

  • Documentation on how to integrate the prototype with the Telco’s existing “Cleaner Pipes” and network infrastructure.
  • A roadmap detailing additional steps needed for production, e.g., containerization, cloud deployment, or further refinements to handle even greater throughput.

4. Academic and Industry Contributions

  • Potential conference/journal publication describing novel or effective methods for real-time telecom fraud detection.
  • Internal knowledge transfer: training Telco staff in the usage, maintenance, and continuous improvement of the solution.

ADDITIONAL DETAILS

The intern will receive $3,300 per month of the internship, usually in the form of scholarship payments.

It is expected that the intern will primarily undertake this research project during regular business hours and maintain contact with their academic mentor throughout the internship either through face-to-face or phone meetings as appropriate.

The intern and their academic mentor will have the opportunity to negotiate the project’s scope, milestones and timeline during the project planning stage.

Please note, applications are reviewed regularly and this internship may be filled prior to the advertised closing date if a suitable applicant is identified. Early submissions are encouraged.

LOCATION:
Remote
DURATION:
6 months
CLOSING DATE:
09/04/2025
ELIGIBILITY:
PhD & Masters by Research students, both domestic & international
REF NO:
APR - 2724

INTERNSHIP CONTACT

CONNECT WITH APR.INTERN

Suggested Internships

NATURAL HAZARDS RESEARCH AUSTRALIA – WA (APR – 2722)

Location:
Perth, WA
Compare and Contrast LA & WA Wildfire Resilience, Preparedness and Response

TELSTRA (APR – 2711)

Location:
Remote
SecureSpeedCore – Shifting the Rust in Telco with Rust in Software

COHEALTH (APR – 2717)

Location:
Melbourne, VIC
Developing a Culturally Responsive and Psychologically Safe Framework for Engaging Families, Carers, and Supporters in Mental Health Care