AI · 2026-10-24
Suspicious Hack Linked to Lazarus? AI Used in Attacks Targeting PHP Vulnerability
Cybersecurity firm Ciphirma says it detected signs that Korean-speaking operators were preparing attacks on web servers in multiple countries by exploiting a critical PHP vulnerability. There are also findings suggesting they used AI to sort and verify public exploit code, but there is still no confirmed evidence directly linking the activity to North Korea’s Lazarus Group.

Amid the discovery of a critical vulnerability in PHP, a widely used server-side programming language, signs have emerged that someone was preparing to exploit it to attack web servers in multiple countries. There are also analysis results suggesting that artificial intelligence (AI) was used during the attack preparation process, but it is important to note that there is not yet enough evidence to conclude that North Korea’s hacking group Lazarus is behind it.
According to Yonhap News, Singapore-based cybersecurity company Ciphirma identified signs on the anonymous network I2P that Korean-speaking operators were discussing attacks using the PHP vulnerability. The vulnerability in question is PHP2Shell, a flaw that allows an attacker to execute commands on a server by sending a manipulated web request without logging in or undergoing any separate authentication. Its CVSS score is 9.2 out of 10, placing it in the critical category.
Ciphirma analyzed the activity as part of a campaign it dubbed “June 2026 Minjun,” saying that government agencies and financial institutions such as banks were the primary targets. It also said manufacturers, logistics companies, shipping and aviation firms, online trading platforms, trading companies, and construction firms were among the suspected targets. By region, the list reportedly included South Korea as well as the United States, Japan, Australia, Taiwan, India, Singapore, the Middle East, and the European Union.
AI was not used to automatically hack targets directly. Instead, it appears to have been used during the preparation stage. According to Ciphirma, the operators gathered public proof-of-concept (PoC) exploit code, then used their own AI model to rank it by priority and test it in a controlled environment before using it in real attacks. In other words, the group appears to have combined publicly available attack tools with AI to identify which vulnerabilities were most likely to be useful.
That said, the link to Lazarus has not been confirmed. Ciphirma raised the possibility of a North Korea connection, but assessed the confidence level as “medium.” Security experts also pointed out that Korean-language instructions, the nature of the targets, and similarities with past tactics are not enough to prove a direct connection, and that no evidence of overlapping malware or attack infrastructure has been confirmed. In other words, the vulnerability exploitation preparations and any judgment about who is behind them should be treated as separate matters.
On the defensive side, the first priority is to check whether PHP environments affected by the vulnerability are in use and to apply the security update. According to Yonhap News, the PHP development team released an emergency update fixing the flaw on October 21, and South Korea’s Ministry of Science and ICT and the Korea Internet & Security Agency (KISA) issued a “caution” advisory and response guidance on the 22nd. Organizations and companies using PHP should review the relevant notices and verify whether the patch has been applied.
This case highlights both the possibility that AI can be used in cyberattacks and the importance of carefully evaluating the strength of the evidence in threat analysis. While the urgency of the vulnerability and the signs of attack preparation should be addressed quickly, any assessment of whether a particular country or organization is involved should remain cautious until additional evidence is available. Source: Yonhap News, report dated October 24, 2026.
According to Yonhap News, Singapore-based cybersecurity company Ciphirma identified signs on the anonymous network I2P that Korean-speaking operators were discussing attacks using the PHP vulnerability. The vulnerability in question is PHP2Shell, a flaw that allows an attacker to execute commands on a server by sending a manipulated web request without logging in or undergoing any separate authentication. Its CVSS score is 9.2 out of 10, placing it in the critical category.
Ciphirma analyzed the activity as part of a campaign it dubbed “June 2026 Minjun,” saying that government agencies and financial institutions such as banks were the primary targets. It also said manufacturers, logistics companies, shipping and aviation firms, online trading platforms, trading companies, and construction firms were among the suspected targets. By region, the list reportedly included South Korea as well as the United States, Japan, Australia, Taiwan, India, Singapore, the Middle East, and the European Union.
AI was not used to automatically hack targets directly. Instead, it appears to have been used during the preparation stage. According to Ciphirma, the operators gathered public proof-of-concept (PoC) exploit code, then used their own AI model to rank it by priority and test it in a controlled environment before using it in real attacks. In other words, the group appears to have combined publicly available attack tools with AI to identify which vulnerabilities were most likely to be useful.
That said, the link to Lazarus has not been confirmed. Ciphirma raised the possibility of a North Korea connection, but assessed the confidence level as “medium.” Security experts also pointed out that Korean-language instructions, the nature of the targets, and similarities with past tactics are not enough to prove a direct connection, and that no evidence of overlapping malware or attack infrastructure has been confirmed. In other words, the vulnerability exploitation preparations and any judgment about who is behind them should be treated as separate matters.
On the defensive side, the first priority is to check whether PHP environments affected by the vulnerability are in use and to apply the security update. According to Yonhap News, the PHP development team released an emergency update fixing the flaw on October 21, and South Korea’s Ministry of Science and ICT and the Korea Internet & Security Agency (KISA) issued a “caution” advisory and response guidance on the 22nd. Organizations and companies using PHP should review the relevant notices and verify whether the patch has been applied.
This case highlights both the possibility that AI can be used in cyberattacks and the importance of carefully evaluating the strength of the evidence in threat analysis. While the urgency of the vulnerability and the signs of attack preparation should be addressed quickly, any assessment of whether a particular country or organization is involved should remain cautious until additional evidence is available. Source: Yonhap News, report dated October 24, 2026.
Related search terms
#PHP vulnerability attack#PHP2Shell exploit#AI used in cyberattacks#Lazarus Group suspected hacking#Korean-speaking hackers#I2P network cyber threat#KISA PHP security advisory