In response to the demand for protecting the integrity of AI programs and fostering confidence in AI solutions, a research team from Nanyang Technological University, Singapore, and experts in the AI sector have developed a new standard on AI security.
“We seek to build trust in AI for AI practitioners by offering guidance on the essential defenses and evaluations to make AI applications safer. At the same time, we expect that users of certified AI solutions would feel more comfortable doing so, adds Prof. Liu Yang of NTU’s School of Computer Science and Engineering, who also oversaw the standard’s research and development.
Despite the many benefits of adopting AI, cybersecurity concerns like hacking pose a severe risk to AI systems, especially when hackers could get access to sensitive data or disrupt automated processes. However, there aren’t many laws enforcing AI system security.
The International Organization for Standardization (ISO) will use the standard to guide global standardization initiatives in this area, making Singapore one of the first countries in the world to steer AI security improvements.
The new standard describes the various types of attacks that AI systems might encounter, how to gauge an AI algorithm’s security, and what AI experts might do to thwart these attacks. 30 AI and security specialists from business, academia, and the government collaborated to create it over the course of a year.
The standard demonstrates the significance of secure AI systems by highlighting four case scenarios in which security breaches could have catastrophic results. These case studies cover systems for flagging objectionable content on social media platforms, credit scoring systems to protect consumers and financial institutions, AI-enabled disease diagnosis systems, and systems for spotting and protecting computers from harmful software.
If these AI systems malfunction, there might be catastrophic consequences for people’s lives. Users might, for example, be exposed to extremist content on social media platforms, receive a false diagnosis, or have their credit score calculated wrongly.
While this is going on, researchers from the National University of Singapore and the NTU Singapore’s Centre for Environmental Life Sciences Engineering (SCELSE) have created a technique to remove phosphorus from wastewater at temperatures higher than those allowed by currently employed methods by storing the chemical in bacteria.
Temperatures exceeding 25 degrees Celsius, which are more frequent in warm areas, make it difficult for current phosphorus removal procedures to function well. As a result of global warming, this is anticipated to happen in more countries.
The SCELSE-developed method, which is based on bacteria, would help to “future-proof” the removal of the toxin because water reclamation plants in Singapore are home to a variety of microbial species. This is due to studies showing that it effectively eliminates phosphorus from wastewater at 30 and 35 degrees Celsius.
The bacterium genus known as Candidatus Accumulibacter takes phosphate from wastewater and stores it inside of itself as polyphosphate granules. It is neither harmful to humans or the environment. According to researchers, their technology might be applied to both large treatment facilities and small laboratory reactors.
The biological phosphorus removal process may function at temperatures as high as 35 degrees Celsius thanks to the bacteria-based technology. Since alternative methods that involve biological innovations only function at cooler temperatures and will become less effective as global warming causes temperatures to rise, this would help “future-proof” phosphorus exclusion.