Skip to content
Bora Haber
Markets
  • US dollar50.44up 0.05%
  • Euro55.34down 0.18%
  • British pound66.55down 0.11%
  • BIST 10011,477down 3.08%
  • Gold (gram)6,043down 3.53%
  • Gold (ounce)US$3,726down 3.58%
  • BitcoinUS$62,702down 1.5%
BREAKING

AI helps uncover PHP's "PHP2Shell" flaw; experts call for rapid-response plans

A critical "PHP2Shell" flaw in PHP, widely used on websites, was found through AI-assisted research. Experts stressed the role of human guidance and layered security.

Nehir Gökçe

· 3 min read

Dark screen showing colorful lines of code: CSS style rules on the right and a blurred HTML login form on the left.
A code editor screen; web application flaws are often found by reviewing source code (representative image). Temsili görsel · Fotoğraf: Harold Vasquez / Pexels

Critical security flaws have been found with the help of artificial intelligence in PHP, a programming language widely used in website infrastructure. The development has renewed attention to risks and solutions ranging from server takeover to the protection of personal data. Organizations increasingly use the analytical capabilities of AI to detect and close vulnerabilities in web-based systems.

The "PHP2Shell" flaws, rated high in severity under the Common Vulnerability Scoring System (CVSS), were identified independently through AI-assisted research. This also shows the power of the technology in spotting complex software bugs. Experts point to the speed of AI in code review, while saying the human factor and proper guidance are decisive for success.

"AI can try millions of paths"

Assoc. Prof. Kübra Kalkan Çakmakçı of the Computer Engineering Department at Özyeğin University's Faculty of Engineering said AI played an important role in detecting the flaw and made analysis of complex code easier. She explained that security flaws do not emerge in a single step; a bigger problem can be noticed when several bugs are linked one after another. Saying AI also speeds up testing and debugging, Çakmakçı said more such detections are expected.

In detecting the PHP2Shell flaw, AI was able to analyze many layers very quickly at the same time. While humans can work on and think through a few stages, AI can try millions of paths and methods.

Assoc. Prof. Kübra Kalkan Çakmakçı

According to Çakmakçı, the flaw's CVSS score was announced as 9.2 out of 10, so it was announced as a serious flaw. The problem is not one that arises directly on individual users' computers; users can be harmed through the websites they use. Because PHP sits in the application layer, an attack there may not be detected by companies' infrastructure layers. She therefore said a holistic structure covering all layers and a rapid-response protocol for critical flaws are needed.

Kaspersky expert: The flaw has existed for years

Fatih Şensoy, Senior Security Researcher at Kaspersky's Global Research and Analysis Team, said the flaw found in PHP has existed for years. According to Şensoy, it can go as far as taking over a server running the affected version; code can be executed and database values accessed. If stored as plain text in the database, user information, session data and credit card details can also be at risk.

Şensoy said storing personal data encrypted in a separate database on the server adds an extra layer, but that if a frequently used site falls into an attacker's hands, control passes to them. Stressing that AI did not find the flaws alone, he said the experts working on the Patch the Planet project, who were invited to it, know well how to guide AI and where vulnerabilities can arise.

Time is critical

Şensoy said time is very valuable in cybersecurity. He said releasing a patch for a flaw is a process of 2-3 days, yet even 30 minutes matters. He noted that defenders constantly monitor flaws, and anonymous attackers also aim to weaponize them quickly and use them against targets.

When we face such an incident, having a plan for how we can react quickly will always minimize our damage.

Fatih Şensoy

Bora Haber previously reported: Cybersecurity event highlights AI agent risks and the need to limit their authority and Google opens cybersecurity-focused Gemini 4 Carbon to selected experts.

Science guides

Take a quiz, vote in a poll

Video news

Comments

Sign in to comment.

More on this topic

See all