[Column] In the AI Era, Security Threats Are Growing Larger... Zero Trust Will Be the Shield
Featured Column by Our Head of R&D, Dr. Junyeol Lee
In this insightful column, Dr. Junyeol Lee discusses the evolving cybersecurity challenges in the AI era, emphasizing the importance of Zero Trust security models and Dtonic’s proactive approach. Originally published in Electronic Times, you can read the full article here in Korean: [Original Link].
Dr. Junyeol Lee, Head of R&D, Dtonic
In November 2023, an incident occurred in China where ransomware developed by exploiting ChatGPT was used in a security attack. This is a representative case showing that artificial intelligence (AI) can become a threat.
Since then, ransomware attacks have been increasing rapidly. According to the ransomware trend report published in the first quarter of this year by KARA (Korean Anti Ransomware Alliance), the number of ransomware damage cases worldwide reached 2,575, a 122% increase compared to the same period last year.
In particular, the development of Large Language Models (LLMs) provides hackers with sophisticated attack methods. The dark side of AI is demonstrated by variations of 'cryptoworms' with self-propagation capabilities, 'WormGPT' that induces the leakage of confidential information through phishing and email, and 'FraudGPT' which supports the creation of malicious code and phishing pages.
Cloud-based AI can cause more severe damage because it processes data by transmitting it to a server.
As a result, 'On-Device AI', which minimizes security threats in the data transmission process by processing data on its own, is gaining attention, but it is not a perfect solution either. This is because the device can learn incorrect data or acquire biased knowledge in the field.
To compensate for this, 'Federated Learning' is conducted, where On-Device AIs that have learned in different spaces and at different times share information with each other and correct incorrect learning. However, this process inevitably requires communication between the edge and the cloud, which can become a vicious cycle that creates new security threats.
Dtonic, the developer of the AI data platform 'D.HUB' and the On-Device AI-applied 'D.EDGE', has adopted and strengthened the 'Zero Trust' model to avoid such threats.
Based on the principle of 'never trust, always verify', this security paradigm treats all access and communication as a security threat, regardless of whether any user, device, or component is inside or outside the network.
By default, data and resources are inaccessible, and strictly controlled access rights are granted for connection. This process is also applied to all endpoints, requiring users to prove their identity with new secret information at every moment.
A Zero Trust system must comply with international standards related to the Internet of Things (IoT) and robot security (e.g., ISO TR 23482-2, IEC 62443) and domestic regulations. Without such sufficient preparation, the move by companies to advance services and increase work efficiency using AI could instead cause great damage.
Accordingly, Dtonic is developing and strengthening interoperability between its D.HUB and D.EDGE solutions, as well as its security solutions. To operate edge-cloud collaboration and learning, a strong security system based on Zero Trust is also being operated in accordance with major domestic and international standards. This is to prevent customers and partners from falling into unexpected security threats while undergoing their Artificial Intelligence Transformation (AX).
Besides Dtonic, many tech startups are facing the same concerns and making similar efforts, but there are also some disappointing aspects. This is because practical barriers such as the absence of clear guidelines, difficulties in setting roadmaps, and the burden of initial costs still exist.
Fortunately, the Ministry of Science and ICT and the Korea Internet & Security Agency (KISA) recently decided to actively proceed with government projects that comply with the core elements of Zero Trust and its guidelines, as well as to support consulting for companies to adopt Zero Trust.
Zero Trust is a security paradigm that all major tech-advanced countries in the world are paying attention to. If the private sector actively adopts Zero Trust in response to global trends, and the government concurrently provides timely institutional support, it is expected that Korea will also be able to effectively respond to the new security threats that are increasing in proportion to the speed of the rapidly expanding AI industry.
By Junyeol Lee, Head of R&D Center, Dtonic
johnlee03@dtonic.io