Research News Last updated 09 April
探花直播 (BCU), in partnership with Covatic, has unveiled groundbreaking research to protect AI models from cyber threats.
AI models are now integral to high-stakes industries like healthcare and autonomous vehicles.
These systems exhibit high accuracy in analysing images, often surpassing human performance at quicker rates.
However, they remain vulnerable to deceptive or adversarial attacks 鈥� malicious attempts to deceive AI systems by subtly altering input data.
One such method, known as a 鈥榖lack-box attack鈥�, allows cyber attackers to test an AI model repeatedly to gather intelligence and find ways to manipulate its decisions.
This could allow an AI-powered self-driving car to misread a stop sign as a speed limit sign; or misdiagnose a patient based on the images it鈥檚 provided.
Groundbreaking from BCU, which was recently published in , has introduced a new defence mechanism for these AI models.
By applying simple and random image adjustments 鈥� like rotations or resizing 鈥� before processing, the AI system becomes more resilient to deception and manipulation.
Compared to regular adversarially trained AI models with no defence, image adjustments saw a 21% improvement in competitive performance and compared to other defence methods such as Random Noise Defence, the margin of improvement increased to between 2.3-4.6% on different AI systems.
Atif Azad, Professor of Artificial Intelligence at BCU, said: 鈥淎s AI plays a bigger role in critical areas like healthcare and self-driving technology, addressing security risks is essential. This research takes an important step toward making AI systems more resilient against cyber threats.鈥�
Dr Umar Daraz, Director of Innovation at BCU, said: 鈥淲e are incredibly proud to be working with our partners from Covatic.
鈥淚t鈥檚 a relationship that has grown to encourage cutting-edge research on AI, data and cyber security solutions, and has helped to create new jobs in the West Midlands. This research will form part of Covatic鈥檚 own applications, which are used by millions of people already.
鈥淭he collaboration has been built on our common goals of innovation and supporting local companies by designing, developing and introducing new tools that were seeded locally, but could have international impacts.鈥�
探花直播-based , is the leading provider of privacy-first audience solutions, and helps companies improve targeted advertising while protecting users' privacy.
Nick Pinks, CEO at Covatic, shared his enthusiasm for the collaboration, stating 鈥淎t Covatic, we are committed to staying at the forefront of technology. This breakthrough has helped us stress-test and better secure our machine learning algorithms from malicious attacks.
鈥淥ur CTO, James Holyhead, worked closely with the BCU team on this project, reflecting our dedication to advancing future technologies, particularly in the advertising industry.鈥�