US government says hackers are now exploiting critical PHP web bugs

U.S. cybersecurity agency CISA has confirmed that hackers are now actively exploiting a pair of critical security vulnerabilities in the widely used PHP web software and has ordered civilian federal agencies within the U.S. government to fix their systems within the next three days.

The urgent new warning to patch comes less than a week after the makers of the decades-old PHP web programming language rolled out patches for the two bugs, which allow hackers to remotely break into a vulnerable system over the internet without the need for passwords. The PHP software is reportedly used by some 70% of web-facing servers, putting major websites and applications at risk. The severity of the bugs prompted some governments and big companies to take their sites offline to fix them.

While the bugs are rated critical, meaning they can be highly damaging if exploited, security researchers said that developing a working exploit to hack into vulnerable systems was initially challenging — but not impossible. Confirmed in-the-wild exploitation now puts extra pressure on organizations to patch their systems to prevent malicious compromises, disruption, and data breaches.

CISA did not provide details about who was exploiting the bugs in its Monday statement, but its statement comes soon after security researchers said they found evidence that the PHP bugs were being exploited.

The bugs, known as PHP2Shell, exploit vulnerabilities in the software that went undiscovered for years. The bugs were initially uncovered two years ago but the nonprofit foundation that maintains the software said there was no practical way for hackers to exploit them, leaving the bugs unpatched. That was until earlier this year, when researchers at security research firm Calif uncovered a way to exploit the two dormant vulnerabilities with the assistance of an AI model.

The two bugs in PHP are among countless vulnerabilities that have not yet been discovered or have been deprioritized for years but now pose a risk due to advances in AI models’ ability to find and exploit vulnerabilities, including by chaining multiple flaws together.