Singapore unveils guidelines to secure AI systems

The Singapore International Cyber Week (SICW) 2024 brought together top global policymakers, thought leaders and cybersecurity experts to advance cooperation and knowledge exchange on key cybersecurity issues and solutions.

With more organizations adopting and developing AI use cases, the Cyber Security Agency of Singapore (CSA) has unveiled the Guidelines and Companion Guide on Security AI Systems at the Singapore International Cyber Week (SICW) 2024.

Developed with the aim of helping organizations adopt AI in a secure manner, the guidelines will help organizations identify potential threats. This includes supply chain attacks, and risk such as Adversarial Machine Learning. The guide will also include principles to guide decision-makers and practitioners on implementation of security controls and best practices to protect AI systems.

Singapore has seen an uptake in AI adoption and in fact is the third fastest country in the world to embrace AI after the US and China. Businesses are aware of the benefits that AI can bring to improve productivity and efficiency. However, some organizations are still concerns on the potential risks that are involved in the development and deployment of AI in their systems.

Studies and research have revealed that AI systems can be vulnerable to adversarial attacks, where malicious actors intentionally manipulate or deceive the AI system. For example, the use of deepfake technologies to manipulate AI systems to get data and information out of an organization. These have led to risks such as data breaches, or result in harmful, or otherwise undesired model outcomes.

With that said, CSA is hoping organizations look to implementing AI that is secured by designed and secured by default, as will all software systems. As such, the guidelines were developed by referencing established international industry guidelines and standards, including the Guidelines for secure AI system development (by the UK National Cyber Security Centre and US Cybersecurity and Infrastructure Security Agency), National Institute of Standards and Technology Risk Management Framework, MITRE Adversarial Threat Landscape for AI Systems and Open Web Application Security Project Machine Learning Security Top 10.

Currently, organizations look to harden the AI model. However, this is insufficient to mitigate security risks. CSA recommends systems owners to take a life-cycle approach for a holistic perspective to identify and mitigate the security risks. The guidelines are developed to help organizations in the five key stages of the AI life cycle, from planning and design to development, deployment, operations and maintenance to end of life.

As businesses carefully develop and deploy AI, securing the data and models in each of these stages is crucial. For example, for development of AI, there should be a focus on supply chain security and protection of AI assets while deployment requires secure infrastructure, establish incident management processes and AI benchmarking and red teaming. When it comes to end of life, businesses should ensure secure and proper disposal of data and model artefacts.

A collective cybersecurity approach in ASEAN

SICW also witnesses ASEAN member states committing to enhance collective cybersecurity in the region. The Southeast Asian region is one of the most diverse regions in the world and is also a hot bed for cybercriminal activities. Recent reports by law enforcement agencies and the UN have indicated an increase in cybercrime in the region, which cybercriminals using technology to exploit and prey on the vulnerable.

The 9th ASEAN Ministerial Conference on Cybersecurity (AMCC) which took place during SICW 2024 witnessed the launch of the physical facility of the ASEAN Regional Computer Emergency Response Team (CERT). Apart from information sharing among ASEAN member states on cyber threats and attacks as well as online scams, the center will serve as a dedicated space for in-person activities. This includes cyber exercises, workshops and CERT-CERT cyber capacity building programs to foster cohesive collaboration among the AMS.

Apart from the center, CSA and the United Nations (UN) Office for Disarmament Affairs launched the Norms Implementation Checklist (NIC) initiative under the auspices of the UN-Singapore Cyber Program in 2020 following ASEAN’s commitment to subscribe in principle to the 11 norms of responsible state behavior in cyberspace. The NIC comprises a set of actions that all states can consider and follow to implement the UN norms of responsible state behavior in cyberspace.

As the first regional checklist of its kind, the NIC will serve as a reference not just for ASEAN member states but for countries beyond the region to support the collective efforts to build a safer and more secure cyberspace.