Secure AI for Robust Anomaly Detection (SARAD)

Posted by:
Categories:

An AI solution to protect your AI models from adversarial attacks

Secure-AI for Robust Anomaly Detection (SARAD) is an accurate, secure and trustworthy AI cybersecurity solution, designed to addresses the significant risk that adversarial attacks pose to AI models.

It offers advanced features that prevent attacks on AI models.

What is SARAD?

Anomaly detection is vital for fraud prevention in finance, equipment monitoring in manufacturing, and identifying health risks in healthcare. While AI-based anomaly detection is state-of-the-art, it faces two critical challenges; security vulnerabilities and lack of explainability.

Adversarial attacks on AI models pose significant risks to organisations including financial losses and safety hazards. In manufcaturing, these attacks can manipulate sensor data, leading to costly errors. In healthcare, the lack of explainability erodes trust, as patients and professionals need clear insights into AI decisions for informed and safe treatment choices.

A 2024 Survey revealed that 77% of IT and data science leaders reported AI breaches, underlining the urgency for secure AI systems. Recent UK government policies have further emphasized the need for AI cybersecurity, creating a favourable environment for innovation in this area.

Therefore, we have developed SARAD, a Secure-AI Anomaly Detection System, that offers additional features, trust and prevents attacks on AI models.

Secure AI cybersecurity countermeasures

AI countermeasures have been built into the system’s defenses to allow it to identify and block attempts to manipulate data or trick the AI into making wrong decisions.

This includes training the AI model to recognize and reject suspicious data, ensuring it stays reliable even in challenging situations.

Explainable AI techniques

SARAD does not just identify anomalies; it also explains why it has done.
For example, if it detects unusual activity, it will provide clear, simple insights into what caused the issue, helping users understand the situation and giving them the knowledge and confidence to act accordingly.

More information

If you would like more information on implementing SARAD at your organisation, please contact:

Dr Angel Jimenez-Aranda
Associate Professor in Digital Transformation
Salford Business School, University of Salford
A.Jimenez-Aranda@Salford.ac.uk

Dr Tarek Gaber
Senior Lecturer in Cyber Security
School of Science, Engineering, and Environment
University of Salford
T.M.A.Gaber@Salford.ac.uk