PHP2Shell: CISA confirms active exploitation of critical PHP flaws patched Oct 21; federal agencies get three days to patch
Matéria consolidada. Reúne 19 matérias que o Tifon tinha publicado separadamente sobre a mesma história. A IA reescreveu o relatório com o que as 8 fontes trouxeram em 28/10/2026 00:12; os links antigos trazem para cá.
Fontes: techcrunch.com ×2 ↗ theregister.com ↗ securityaffairs.com ↗ wired.com ×2 ↗ itsecurityguru.org ×2 ↗ olhardigital.com.br ×4 ↗ digit.fyi ↗ darkreading.com ↗
Two critical memory-corruption vulnerabilities in the PHP engine, collectively dubbed PHP2Shell, were fixed in emergency security releases on Wednesday, October 21, 2026, after details leaked ahead of the planned disclosure (darkreading.com, itsecurityguru.org, digit.fyi). The PHP Foundation released four updates that day, including PHP 8.5.12 (olhardigital.com.br, techcrunch.com). Each flaw is rated CVSS 9.2 and allows unauthenticated remote code execution, according to CERT-EU (olhardigital.com.br). On October 27 CISA confirmed in-the-wild exploitation of the two flaws and ordered civilian federal agencies to remediate within three days. It gave no details on who is exploiting them, and its statement followed researcher reports of exploitation evidence (techcrunch.com). Earlier, Amazon Web Services (AWS) honeypot logs had shown at least one actor attempting exploitation from October 17, four days before the fixes and public disclosure at 10 am ET (wired.com, securityaffairs.com, theregister.com). AWS CISO CJ Moses said PHP2Shell was being broadly exploited (wired.com). Cloudflare reported millions of mostly opportunistic scanning attempts, and CEO Matthew Prince said there were thousands in the 10 hours before disclosure, tens of thousands within 45 minutes of the announcement and hundreds of thousands within 2 hours (digit.fyi). As of October 23 the PHP Foundation and the experts The Register spoke to had seen no successful exploitation (wired.com, theregister.com). That position predates the CISA confirmation. The NSA warned about the flaws on October 21 (techcrunch.com), and the UK NCSC urged organisations to identify affected systems and update immediately (itsecurityguru.org). A third vulnerability was disclosed later; the PHP Foundation says it needs an unlikely combination of configuration and application code and will ship in the next regular release (theregister.com). The impact was wide. Government services in France, Denmark, Finland and the Netherlands, UK council sites, NHS Scotland and Wikipedia went offline (itsecurityguru.org), as did some services such as the Final Fantasy XIV login server during patching (theregister.com). In Brazil, Anatel and the federal courts TRF1, TRF4 and TRF6 suspended external access to systems including the Sistema Eletrônico de Informações (SEI) on October 22, after an alert from the Brazilian CISC on October 21 (olhardigital.com.br).
Entenda a falha
A vulnerabilidade
PHP2Shell is the name for two memory-corruption flaws in the PHP engine itself. CERT-EU rates each CVSS 9.2 and says an unauthenticated attacker can get remote code execution over the network, with no account. The impact is data theft, defacement or full server takeover. The articles give no CVE IDs. Reporting says every version since PHP 5 is affected; The Register says roughly 5.4 onward. Fixes shipped on October 21, 2026 in four releases, including PHP 8.5.12. PHP 8.1 and older are unsupported and must be upgraded. Updating WordPress or other frameworks does not help, because the bug is in the PHP runtime. A third flaw, disclosed later, needs an unlikely mix of configuration and application code, can be mitigated by a setting and will be fixed in the next regular release.
Como funciona
The articles do not describe the exact code path, so the mechanism is only known in outline. A remote request reaches PHP running on the server, with no login needed. The flaws corrupt memory that PHP trusts, which lets an attacker redirect execution inside the PHP process and run code with that process's privileges. The two bugs have to be chained, and the chain is complex. That is why working exploits were slow to appear. VulnCheck showed that ASLR can be bypassed on x86, so hardening settings alone do not guarantee safety. Real impact depends on server configuration and privileges.
Por que é vulnerável
Both are memory-safety bugs in the PHP core, in code that commentators say dates back to 2004. The articles do not name a more specific root cause. They explain why the bugs lasted so long: when first signalled in 2024 there was no known way to get remote code execution from them, and the fix then stalled and restarted for lack of resources, trouble making decisions in the security team and few experts on PHP internals. The PHP Foundation called this a structural process failure. The patch is reportedly about 13,000 lines of C, according to The Register. The articles do not detail what it changes.
Como foi descoberta
The first signs date to a 2024 external audit by Quarkslab, but the fix stalled. Calif researchers (Ryan Austin) identified the flaws and built exploits in early August 2026, using frontier AI (GPT-5.6 Sol) via OpenAI's Patch the Planet program plus human expertise; Olhar Digital says Calif reported one flaw on August 11. OpenAI and Wiz later found them independently with AI, and Wiz reported a second flaw on August 24. Disclosure was coordinated, but details leaked around October 20 (surfacing on Hacker News), and PHP released fixes on October 21. AWS honeypots show an actor trying exploitation from October 17, and it is unclear how that actor got the information. CISA confirmed exploitation in the wild on October 27 without naming actors. Before that, Reuters reported (via Olhar Digital) that CISA had seen none, and the PHP Foundation saw no successful exploitation as of October 23.
PoC pública
Wired reports that proof-of-concept exploits for PHP2Shell began appearing publicly on October 21, some apparently hours before the 10 am ET disclosure, and that many circulating exploits seem ineffective. Authors and repositories are not named. VulnCheck also reportedly developed an exploit showing an ASLR bypass on x86, but the articles do not say it is public. Wiz warns that AI can adapt exploits to different PHP versions and memory layouts, and attackers also reverse-engineered the fixes. Combined with millions of scan attempts reported by Cloudflare and CISA's confirmation of exploitation, patching is urgent.
Explicação gerada por IA a partir da matéria e dos artigos, em 28/10/2026 00:28.
Detalhes técnicos
- The flaws are in the PHP runtime itself, so updating WordPress or other frameworks does not fix them. PHP must be updated and the services that run it restarted (itsecurityguru.org). WordPress says the issue is not WordPress-specific (techcrunch.com).
- Both flaws are memory-corruption bugs that let an attacker shape memory PHP trusts and hijack execution, leading to RCE, site takeover, data theft and malware delivery to visitors. The attack chain is complex, which is why working exploits were slow to appear (theregister.com). Impact depends on server configuration and privileges, and consequences include data theft, malware installation, defacement and pivoting to other systems (securityaffairs.com). CERT-EU confirms remote code execution without authentication or an account, with possible data theft, web page defacement or full system compromise (olhardigital.com.br). The NSA described the exposure as access to the server file system or the database used by the web application (techcrunch.com).
- PHP rates both flaws 9.2. Commentators say they sit in code dating back to 2004 (over a decade old), and the folded reporting says all versions since PHP 5 are affected (olhardigital.com.br). The Register reports that every PHP release of the past decade, back to roughly 5.4, is affected. PHP 8.1 and older are unsupported, so those sites must be upgraded (techcrunch.com).
- VulnCheck developed an exploit showing that ASLR can be bypassed on x86 machines, so configuration-based mitigations are not a guarantee (theregister.com).
- The patch for the current version is reportedly about 13,000 lines of C (theregister.com).
- A third vulnerability was disclosed later. It requires 'a very unlikely combination of configuration and specific PHP application code', can be mitigated with a configuration setting, and will ship in the next regular release (theregister.com).
- Discovery: The flaws were found with AI assistance. Calif researchers identified them and built exploits in early August, using frontier AI tools through OpenAI's Patch the Planet program together with expert human knowledge. OpenAI and Wiz researchers later found them independently with AI (wired.com, theregister.com). Calif found one flaw with GPT-5.6 Sol (techcrunch.com) and reported it on 11 August (olhardigital.com.br). Wiz used its Wiz Atlas AI agent, Gemini 4 Argon and GPT-5.6 Sol (techcrunch.com) and reported a second flaw on 24 August (olhardigital.com.br). Later reports say Calif and OpenAI discovered and reported the flaws independently in August 2026, and that Wiz found further ways to exploit them (olhardigital.com.br). Wiz called them among the most serious PHP vulnerabilities ever found, possibly in software history, unanswered for over a decade because discovery and exploitation were complex without AI (olhardigital.com.br). Calif later used an AI model to exploit the two dormant bugs, including chaining them together. Building a working exploit was initially difficult but not impossible (techcrunch.com).
- Earlier identification: The flaws were first identified about two years ago, but the maintaining foundation judged them practically unexploitable and left them unpatched (techcrunch.com). Some of the underlying issues were known in 2024 but were judged too complex to exploit, so AI effectively turned a long-dormant weakness into a present-day crisis (wired.com). The PHP Foundation confirmed that the first signs date to a 2024 external audit by Quarkslab, and that the fix stalled and restarted repeatedly because of limited resources, decision-making difficulties in the security team, the lack of a known remote command execution method at the time, and a shortage of experts who know PHP internals. The Foundation called this a structural process failure (olhardigital.com.br).
- AI use by the PHP Foundation: It uses AI tools such as Claude Mythos to analyze code and triage reports, but says they cannot determine the best way to structure a fix (olhardigital.com.br).
- Experts say AI magnifies existing discovery and disclosure dynamics rather than doing anything new. Sophisticated AI-aided finds still need skilled humans, and Calif's Ryan Austin wrote that no model can find 'all bugs' with simple prompts. Jake Williams (Hunter Strategy) expects more long-lived bugs to surface, but notes that few targets match PHP's internet-facing reach (wired.com). OpenAI expects similar findings to increase sharply in the coming months (itsecurityguru.org). Wiz warns that LLMs can now help adapt exploits to different PHP versions, configurations and memory layouts (techcrunch.com).
- Reach: Wiz estimates that about 70 percent of the Internet is affected, given PHP's prevalence in platforms such as WordPress. This is an estimate, not a measured count (digit.fyi). WordPress and most major CMSes are built on PHP, and PHP is estimated to run about 70% of websites (theregister.com), over 33 million sites (techcrunch.com), including platforms such as Moodle and Magento (olhardigital.com.br). TechCrunch (October 27) says PHP reportedly runs on about 70% of web-facing servers. Many organizations may not know they run PHP inside third-party products, CMSs or hosting platforms (olhardigital.com.br).
- The leak surfaced on Hacker News, and PHP moved its security release up by a day. The source of the leak is unknown and no threat group involvement is evidenced (itsecurityguru.org). The Register adds that, after leaks and a purported exploit video, the Foundation warned of emergency patches on Monday and released them on Wednesday.
- It is unclear how the actor active on October 17 obtained exploit information while the flaws were under embargo. That activity came from an IP address linked to a commercial anonymization network (wired.com).
- In the hours before disclosure, AWS honeypots received multiple anonymized attempts with apparently different goals, likely from multiple actors. AWS's technical analysis is ongoing (wired.com). Detected attempts do not automatically mean successful compromises (securityaffairs.com).
- Cloudflare telemetry shows exploitation attempts began before public disclosure, which indicates the leak was exploited ahead of patching (digit.fyi). CISA's October 27 confirmation shows exploitation is occurring but does not identify the actors, targets or techniques (techcrunch.com).
- Proof-of-concept exploits began appearing publicly on October 21, some apparently hours before disclosure. Attackers also reverse engineered the fixes (wired.com). Experts warn attackers may reverse-engineer the patches, and Don Smith (Sophos CTU) expects exploitation to be prioritized within hours or days of release, possibly with AI assistance (olhardigital.com.br). AWS argues AI-assisted analysis makes turning a patch diff or advisory into a working exploit cheaper and faster (wired.com). Security Affairs cautions against assuming every attack is AI-driven (securityaffairs.com).
- According to Wiz, frontier AI models let a single researcher build a working exploit in a week, work that previously would have taken a world-class team months (digit.fyi).
- PHP is often embedded in third-party products and applications, which makes asset discovery difficult (digit.fyi). Forgotten PHP is likely to be found in vendor appliances (NAS, firewalls, VPN portals, printers, building-management panels, OT/ICS HMIs), bundled runtimes in control panels, containers on old base images, marketing microsites, acquired-company infrastructure, admin tools (phpMyAdmin, Roundcube, Nextcloud), and origins reachable directly behind a CDN or WAF. Older and end-of-life PHP versions in appliances and containers may never receive an official fix (itsecurityguru.org). Legacy servers on old PHP versions have no direct update path and remain exposed (theregister.com).
- Suggested hunting indicators: php-fpm, httpd or apache2 spawning sh, bash, curl, wget or base64; new or modified .php files in web roots; outbound connections from web worker processes (itsecurityguru.org).
- One expert warns that attackers could distribute malicious 'patches' that exploit the rush to fix (itsecurityguru.org).
- Impact in Brazil: Anatel suspended external access to SEI and other internet-facing services. TRF1, TRF4 and TRF6 took similar measures, and TRF6 also suspended its electronic judicial process system. The TSE was asked whether its systems are affected. GSI told federal IT managers to update immediately and MGI told federal teams to temporarily take PHP-based systems offline. The federal government's alert urged public bodies to update. A 23 October article reports no major impact in Brazil so far (olhardigital.com.br). French local government sites, including that of the Paris Police Prefecture, were made inaccessible at the request of the national cybersecurity agency (olhardigital.com.br). Governments and companies reportedly took sites offline as a precaution from 21 October (olhardigital.com.br).
- Gaps: No source gives CVE identifiers, IOCs or attribution of the exploitation. TechCrunch (October 27) also gives none.
Remediação
- Apply the latest PHP patches immediately. CISA set a three-day deadline for federal civilian agencies (techcrunch.com). Update to the current PHP release, 8.5.12, and upgrade sites on PHP 8.1 and older, which are unsupported (techcrunch.com).
- Update to the latest PHP version across all deployments, apply the PHP Team's additional hardening steps, and restart services. Use the PHP Team's temporary mitigations if updating is not possible (itsecurityguru.org, digit.fyi). Apply the PHP Foundation emergency patches released on October 21 (theregister.com).
- Prioritize internet-facing PHP servers and monitor them for compromise, disruption and data theft (techcrunch.com).
- Inventory all PHP use, including embedded and third-party components (Accenture, via digit.fyi), prioritising internet-facing systems and not limiting the search to outdated versions. Combine external exposure scanning, host package managers, container/SBOM scans and network behaviour analysis (itsecurityguru.org). Also inventory indirect PHP dependencies such as CMSes and PHP-based applications (theregister.com), and confirm affected versions against official advisories (olhardigital.com.br).
- Systems on old PHP versions with no update path need to be taken down and rebuilt or reinstalled (theregister.com).
- Apply the configuration-based mitigation for the third issue, bearing in mind the ASLR-bypass research (theregister.com).
- WordPress site owners must patch their servers and work with their hosting providers (techcrunch.com).
- Where patching is not yet possible, virtually patch with a Web Application Firewall (Akamai's recommendation, via digit.fyi). Monitor web logs and WAF telemetry for exploitation attempts and treat publicly circulating proof-of-concept code with caution (wired.com).
- Monitor PHP and web server processes for unusual activity or unexpected command execution, and check whether compromise occurred before patching (olhardigital.com.br).
- Review logs for suspicious activity dating back to October 17 and investigate signs of compromise. Restrict access to vulnerable services (securityaffairs.com).
- If mitigations cannot be applied quickly, the NCSC suggests taking the system offline until it is fully mitigated (itsecurityguru.org). The NSA and CISA likewise advise isolating vulnerable systems or taking them offline if updates are not possible (olhardigital.com.br, techcrunch.com). Some governments and other groups took key websites offline before the release for this reason (wired.com). CERT-EU advises isolating internet-facing applications if patching is not immediately possible (itsecurityguru.org). Anatel and the courts kept exposed systems suspended or restricted until patched, and GSI and MGI issued instructions to federal bodies to update or take PHP-based systems offline (olhardigital.com.br).
- Maintain a 'lifeboat' continuity plan for critical operations while vulnerabilities are addressed (Accenture, via digit.fyi). Keep continuous monitoring and tested incident response plans (securityaffairs.com).
- Monitor third-party vendors for updates to products that bundle PHP, and validate patches against official release notes (itsecurityguru.org).
- If compromise is suspected, rotate secrets on exposed hosts, filter multipart requests at the edge, restrict egress from web workers, and isolate EOL hosts (itsecurityguru.org).
- For any site exposed before patching, review who has access to the site and internal systems, remove unused admin accounts, apply least privilege, replace passwords as a precaution and enforce MFA on every login, using unique passwords (itsecurityguru.org, securityaffairs.com).
- AWS says coordinated vulnerability disclosure must evolve and speed up (wired.com).
- Wiz thanked Automattic, Akamai, AWS, Cloudflare, Fastly, Imperva, Microsoft and Vercel for helping validate the patches (techcrunch.com).
Relatos divergentes
- Exploitation status: The earliest report spoke only of reported exploitation attempts (darkreading.com). The AWS-based reporting dates attempts to October 17 and describes broad in-the-wild exploitation (wired.com). itsecurityguru.org cites reports of millions of attempts but says it is unclear whether any succeeded, and digit.fyi, citing Cloudflare, gives attempt volumes without confirming compromises. On October 21 an early report said no exploitation had been observed (techcrunch.com), and a later report cited Cloudflare's CTO on first attacks while CISA said it had not confirmed exploitation (olhardigital.com.br). As of October 23, wired.com and theregister.com said the PHP Foundation and the experts consulted had seen no successful exploitation. The most recent and most authoritative source is CISA (via techcrunch.com, October 27), which confirmed in-the-wild exploitation and set a three-day federal deadline. This supersedes the earlier 'no exploitation' positions. CISA's statement does not say how many systems were compromised or by whom, so the scale of successful compromise remains unconfirmed.
- Exploit code: One source says PoC exploits appeared publicly on October 21 (wired.com). Another says no exploit code was released, only a screenshot of the researchers' exploit tool (itsecurityguru.org). The Register (more recent) says many PoCs exist, some working and some not, and wired.com (October 24) says many appear ineffective. The itsecurityguru.org claim is contradicted by the later reporting.
- Severity and impact description: PHP rates both flaws 9.2, while some early reports cited 10.0 (itsecurityguru.org). The 9.2 rating is the more authoritative and is also used by digit.fyi and olhardigital.com.br (citing CERT-EU). The NSA warning is described as file system or database exposure (techcrunch.com), while CERT-EU describes unauthenticated remote code execution (olhardigital.com.br). The latter is more specific and more recent.
- Leak and disclosure timing: The leak is dated October 20, two days before a planned disclosure that was moved up to October 21 (itsecurityguru.org). Another source gives only the October 21, 10 am ET disclosure (wired.com), and digit.fyi says only that fixes were released on Wednesday ahead of schedule. The Register says that, after leaks and a purported exploit video, the Foundation warned of emergency patches on Monday (October 19) and released them on Wednesday, which does not match an October 20 leak date. The leak date remains unconfirmed.
- Number of flaws: Most sources speak of two flaws. The Register (October 26) reports a third vulnerability disclosed later, which will be fixed in a regular release and is considered much harder to exploit. Wiz is also said to have found multiple PHP vulnerabilities (techcrunch.com).
- Patch scope: One source cites a single release, PHP 8.5.12 (techcrunch.com), while another says four updates were released on October 21 (olhardigital.com.br). These may be compatible, since the updates likely cover several branches, but the sources do not say.
- Scope and versions: Wiz's estimate that about 70 percent of the Internet is affected (digit.fyi) is an unverified estimate. The Register frames the 70% figure as the share of websites running PHP, TechCrunch as the share of web-facing servers. The Register says every release back to roughly 5.4 is reportedly affected, while olhardigital.com.br says all versions since PHP 5 (2004). There is no official version list or CVEs in the sources, so the actual exposed share is unconfirmed.
- Discovery and attribution of the flaws: Wired.com reports that Calif found them in early August and that OpenAI and Wiz found them independently. The folded reporting describes one flaw reported by Calif with OpenAI (11 August) and a second by Wiz (24 August) (techcrunch.com, olhardigital.com.br). The newest olhardigital.com.br articles say Calif and OpenAI found the flaws independently in August and that Wiz later found additional exploitation paths. The sources do not reconcile whether Wiz found a separate flaw or further exploitation paths for the same ones. TechCrunch (October 27) credits Calif with exploiting the dormant bugs using an AI model.
- Origin and age of the flaws: Commentators say the flaws sit in code dating back to 2004 (itsecurityguru.org). Some underlying issues were known in 2024 but judged too complex to exploit (wired.com). The PHP Foundation traces the first signs to a 2024 Quarkslab audit (olhardigital.com.br), and TechCrunch says they were identified about two years ago and judged practically unexploitable. Only olhardigital.com.br names the auditor.
Linha do tempo
-
CISA confirms active exploitation of critical PHP2Shell PHP flaws; federal agencies ordered to patch within three days
ler a análise desta fonte ocultar
CISA has confirmed in-the-wild exploitation of two critical remote code execution vulnerabilities in PHP, collectively called PHP2Shell, and ordered civilian federal agencies to remediate within three days. The flaws allow unauthenticated remote compromise over the internet. Patches were released less than a week before the warning. PHP reportedly runs on about 70% of web-facing servers, and some governments and large companies took sites offline to patch. CISA gave no details on who is exploiting the bugs, but its statement followed researcher reports of exploitation evidence.
Technical Details
- The two vulnerabilities went undiscovered for years. They were first identified about two years ago, but the maintaining foundation judged them practically unexploitable and left them unpatched.
- Researchers at Calif later found a way to exploit the two dormant bugs with the help of an AI model, including chaining the flaws together.
- Building a working exploit was initially difficult but not impossible.
- The article does not give CVE identifiers, IOCs or attribution.
Remediation
- Apply the latest PHP patches immediately. CISA set a three-day deadline for federal civilian agencies.
- Prioritize internet-facing PHP servers, and monitor them for compromise, disruption and data theft.
-
PHP2Shell memory-corruption flaws in PHP core expose ~70% of websites to RCE; exploitation attempts seen, no confirmed success
ler a análise desta fonte ocultar
PHP2Shell is the collective name for two related memory-safety vulnerabilities in the PHP core (a third was disclosed later) that can lead to remote code execution. They reportedly affect every PHP release of the past decade, back to roughly 5.4, and PHP is estimated to run about 70% of websites. The flaws were found in August by Calif (using OpenAI models) and independently by Wiz, and were handled under coordinated disclosure with the PHP Foundation. After leaks and a purported exploit video, the Foundation warned of emergency patches on Monday and released them on Wednesday. AWS reported exploitation attempts starting at least four days before the patches. As of Friday, the Foundation and the experts the outlet spoke to had seen no evidence of successful exploitation, despite many proof-of-concept exploits (some working, some not). Legacy servers on old PHP versions have no direct update path and remain exposed.
Technical Details
- Both flaws are memory-corruption bugs that let an attacker shape memory PHP trusts and hijack execution, leading to RCE, site takeover, data theft and malware delivery to visitors.
- The attack chain is complex, which is why working exploits have been slow to appear.
- VulnCheck developed an exploit showing that ASLR can be bypassed on x86 machines, so configuration-based mitigations are not a guarantee.
- A third vulnerability was disclosed later. The PHP Foundation says it requires "a very unlikely combination of configuration and specific PHP application code" and will ship in the next regular release, not an emergency one. It can be mitigated with a configuration setting.
- The patch for the current version is reportedly about 13,000 lines of C.
- Downstream impact is wide: WordPress and most major CMSes are built on PHP, and some services (e.g. the Final Fantasy XIV login server) went offline during patching.
Remediation
- Apply the PHP Foundation emergency patches released Wednesday.
- Systems on old PHP versions with no update path need to be taken down and rebuilt or reinstalled.
- Apply the configuration-based mitigation for the third issue, bearing in mind the ASLR-bypass research.
- Inventory indirect PHP dependencies, such as CMSes and applications built on PHP.
-
PHP2Shell PHP remote command execution flaw exploited four days before public disclosure, AWS honeypots show
ler a análise desta fonte ocultar
Security Affairs founder Pierluigi Paganini answers questions on PHP2Shell, a PHP flaw that can let attackers execute commands remotely on vulnerable servers. AWS researchers detected exploitation attempts against their honeypot network four days before public disclosure and patch release. The article notes that detected attempts do not automatically mean successful compromises. The piece stresses that attackers can weaponize flaws within hours, possibly with AI assistance, and that patch-only vulnerability management is not enough for internet-facing systems.
Technical Details
- The flaw allows remote command execution on vulnerable PHP servers. Impact depends on server configuration and privileges.
- Possible consequences include data theft, malware installation, website defacement and use of the server as a pivot to attack other systems.
- AWS honeypots recorded the first exploitation attempts around October 17, ahead of public disclosure.
- AI-assisted analysis may shorten the time from technical information to a working exploit, but the article cautions against assuming every attack is AI-driven.
- The article gives no CVE identifier, affected versions or IOCs.
Remediation
- Apply the relevant PHP security updates and check whether PHP applications are vulnerable.
- Review logs for suspicious activity dating back to October 17, and investigate signs of compromise.
- Restrict access to vulnerable services and apply temporary mitigations where patches are not available.
- Maintain an inventory of internet-facing assets, continuous monitoring and tested incident response plans.
- Users should use unique passwords and multi-factor authentication.
-
PHP2Shell: Millions of exploitation attempts hit newly disclosed PHP flaws within hours, with no confirmed successful compromise
ler a análise desta fonte ocultar
Criminal actors began probing the newly public PHP2Shell vulnerabilities in PHP within hours of disclosure, and Cloudflare observed millions of exploitation attempts. The PHP Foundation says it has seen no successful exploitation so far, and many publicly circulating exploits appear ineffective. Calif researchers first identified the flaws and built exploits in early August, partly using frontier AI tools through OpenAI's Patch the Planet program alongside expert human knowledge. Researchers at OpenAI and Wiz later found PHP2Shell flaws independently with AI tools. Some of the underlying issues were known in 2024 but were judged too complex to exploit, so AI effectively turned a long-dormant weakness into a present-day crisis.
Technical Details
- The flaws are collectively named PHP2Shell. The article gives no CVE identifiers, affected versions or technical mechanics.
- Cloudflare's analysis shows mass opportunistic scanning and exploitation attempts shortly after publication.
- Experts say AI is magnifying existing vulnerability discovery and disclosure dynamics rather than doing anything new. Sophisticated AI-aided finds still require skilled humans, and Calif's Ryan Austin wrote that no model can find "all bugs" with simple prompts.
- Jake Williams (Hunter Strategy) expects more long-lived bugs to surface, but notes that few targets match PHP's internet-facing reach.
Remediation
- Apply the PHP patches for the PHP2Shell flaws as soon as they are available, prioritising internet-facing PHP applications.
- Monitor web logs and WAF telemetry for exploitation attempts, and treat publicly circulating proof-of-concept code with caution.
-
PHP2Shell: two critical PHP flaws exploited since Oct 17 after leak; emergency patches released, government sites taken offline
The article adds the operational impact: government services in France, Denmark, Finland and the Netherlands, UK council sites, NHS Scotland and Wikipedia went offline. It also adds reports of millions of exploitation attempts with unconfirmed success, CERT-EU's isolation advice, OpenAI's statement that its model helped find the flaws, and post-exposure account hygiene guidance.
ler a análise desta fonte ocultar
The article adds the operational impact: government services in France, Denmark, Finland and the Netherlands, UK council sites, NHS Scotland and Wikipedia went offline. It also adds reports of millions of exploitation attempts with unconfirmed success, CERT-EU's isolation advice, OpenAI's statement that its model helped find the flaws, and post-exposure account hygiene guidance.
Two critical vulnerabilities in the PHP engine, dubbed PHP2Shell, were fixed in emergency releases on 21 October 2026 after details leaked on 20 October, ahead of the scheduled disclosure. Amazon's threat intelligence team reported exploitation attempts dating back to 17 October, and the UK NCSC urged organisations to identify affected systems and update immediately. The article reports that government services in France, Denmark, Finland and the Netherlands, as well as UK council sites, NHS Scotland and Wikipedia, went offline, and that there were reports of millions of exploitation attempts. It is unclear whether any attempts succeeded. PHP rates both flaws 9.2, not the 10.0 some early reports cited. The flaws reportedly sat in code dating back to 2004 and were found with AI assistance, and OpenAI says its model helped find them.
Technical Details
- The flaws are in the PHP runtime itself, so updating WordPress or a framework does not fix them. PHP must be updated and the services that run it restarted.
- The leak surfaced on Hacker News. PHP moved its security release up a day, the source of the leak is unknown, and there is no evidence a threat group was involved. No exploit code was released, but a screenshot of the researchers' exploit tool circulated.
- Experts warn that AI lowers the effort needed for variant analysis and exploit development, and that attackers could distribute malicious "patches" while no official fix is available.
- Forgotten PHP is likely in vendor appliances (NAS, firewalls, VPN portals, printers, building-management panels, OT/ICS HMIs), bundled runtimes in control panels, containers on old base images, marketing microsites, admin tools (phpMyAdmin, Roundcube, Nextcloud) and origins reachable directly behind a CDN or WAF. Older and EOL versions may never receive an official fix.
- Hunting indicators: php-fpm, httpd or apache2 spawning sh, bash, curl, wget or base64; new or modified .php files in web roots; outbound connections from web worker processes.
Remediation
- Inventory all PHP use, prioritising internet-facing systems and not limiting the search to outdated versions. Combine external exposure scanning, host package managers, container/SBOM scans and network behaviour analysis.
- Update to the latest PHP version, apply the PHP Team's hardening steps and restart services. Use the PHP Team's temporary mitigations if updating is not possible. If those cannot be applied quickly, the NCSC suggests taking the system offline. CERT-EU advised isolating internet-facing applications.
- If compromise is suspected, rotate secrets on exposed hosts, filter multipart requests at the edge, restrict egress from web workers and isolate EOL hosts.
- Review who has access to exposed sites and internal systems, remove unused admin accounts, apply least privilege, replace passwords and enforce MFA on every login.
- Monitor third-party vendors for updates to products that bundle PHP, and validate patches against official release notes.
-
Critical PHP flaws, first seen in a 2024 audit, trigger preventive takedowns of sites and Brazil's SEI
ler a análise desta fonte ocultar
The PHP Foundation confirmed that the first signs of the PHP vulnerabilities behind preventive site takedowns in several countries date to a 2024 external audit by Quarkslab. The fix stalled and restarted repeatedly because of limited resources and decision-making difficulties in the security team. The flaws could let attackers take control of servers without authentication. A security update is available. Since 21 October, governments and companies have temporarily suspended access to sites, including Anatel and Brazilian courts, which took down the Sistema Eletrônico de Informações (SEI). The article reports no major impact in Brazil so far.
Technical Details
- The vulnerabilities were identified in a 2024 audit by Quarkslab. The delay was attributed to the lack of a known remote command execution method at the time, difficulties in security team decisions, and a shortage of experts who know PHP internals.
- Calif and OpenAI independently found and reported the flaws in August 2026. Wiz later found additional ways to exploit them, using Google's Gemini 4 Argon and OpenAI's GPT-5.6 Sol models.
- Wiz called the findings among the most severe ever discovered in PHP and possibly in software history, saying they went unanswered for over a decade because discovery and exploitation were complex without AI.
- The PHP Foundation uses AI tools such as Claude Mythos to analyze code and triage reports, but says they cannot determine the best way to structure a fix.
- PHP is reportedly used by about 70% of sites, including platforms such as WordPress, Moodle and Magento.
- The article gives no CVE identifiers, affected versions or evidence of active exploitation.
Remediation
- Apply the available PHP security update to affected systems; the article presents this as the main measure to reduce the risk of intrusion.
- Preventive suspension of exposed services was used as a temporary measure by public bodies and companies.
-
PHP2Shell critical PHP flaws leaked, pre-disclosure exploitation surges to millions of attempts, Cloudflare reports
ler a análise desta fonte ocultar
Cloudflare reported millions of exploitation attempts after the disclosure of PHP2Shell, a set of critical PHP vulnerabilities rated 9.2 out of 10.0 that can theoretically allow unauthenticated remote code execution. CEO Matthew Prince said thousands of attempts were seen in the 10 hours before public disclosure, rising to tens of thousands within 45 minutes of the announcement and hundreds of thousands within 2 hours. The flaws leaked before the planned coordinated disclosure, so the PHP project released fixes on Wednesday ahead of schedule. Wiz researchers say AI helped turn long-dormant code weaknesses, present for over a decade, into a working exploit. Some high-profile sites were reportedly taken offline temporarily as a precaution.
Technical Details
- Wiz estimates that about 70 percent of the Internet is affected, given PHP's prevalence in platforms such as WordPress.
- The vulnerable code is more than a decade old. Frontier AI models allowed a single researcher to build a working exploit in a week, work that previously would have taken a world-class team months.
- Exploitation attempts began before the public disclosure, which indicates the leak was exploited ahead of patching.
- PHP is often embedded in third-party products and applications, so asset discovery is difficult. The article does not give CVE identifiers, affected PHP versions or IOCs.
Remediation
- Apply the updates the PHP project has released across all PHP deployments.
- Where patching is not yet possible, virtually patch with a Web Application Firewall (Akamai's recommendation) and consider taking exposed sites offline until patched.
- Inventory where PHP runs, including embedded and third-party components (Accenture).
- Maintain a 'lifeboat' continuity plan for critical operations while vulnerabilities are addressed (Accenture).
-
Two critical unauthenticated PHP flaws force Anatel and Brazilian courts to suspend systems including SEI
ler a análise desta fonte ocultar
Two serious PHP vulnerabilities, reportedly allowing attackers to take control of servers without credentials, led Anatel and Brazilian courts to temporarily suspend access to their systems, including the Sistema Eletrônico de Informações (SEI). The Brazilian federal government issued an alert urging public bodies to update the software. According to the report, the flaws were identified with the help of artificial intelligence, and a fix is already available. The article is a newsletter digest and gives no CVE identifiers, exploitation evidence or technical details.
Technical Details
- Two flaws in PHP that may enable unauthenticated takeover of servers (no login or password required).
- Discovery reportedly assisted by AI.
- No CVEs, affected versions or evidence of in-the-wild exploitation are stated.
Remediation
- Apply the available PHP fix, as urged by the federal government's alert to public bodies.
- Keep exposed systems suspended or restricted until patched, as Anatel and the courts did.
-
Critical unauthenticated PHP flaws force Anatel and Brazilian federal courts to take SEI and other systems offline
ler a análise desta fonte ocultar
Two critical PHP vulnerabilities, reportedly allowing unauthenticated attackers to take control of servers, led Anatel and several Brazilian federal courts (TRF1, TRF4, TRF6) to suspend external access to systems including the Sistema Eletrônico de Informações (SEI). The Brazilian CISC issued an alert on 21 October 2026. GSI instructed federal IT managers to update immediately, and MGI told federal teams to temporarily take PHP-based systems offline. The flaws were found independently by Calif and OpenAI in August, and Wiz later found additional ways to exploit them using AI models. Wiz called them among the most serious PHP vulnerabilities ever found. The NSA recommended patching immediately or shutting down vulnerable systems, and French local government sites, including that of the Paris Police Prefecture, were made inaccessible at the request of the national cybersecurity agency. The article does not give CVE identifiers, affected versions or evidence of in-the-wild exploitation.
Technical Details
- Two vulnerabilities in PHP allow actions on servers without credentials, which can lead to full server takeover.
- Discovered independently by Calif and OpenAI in August 2026. Wiz later found further exploitation paths using the Gemini 4 Argon (Google) and GPT-5.6 Sol (OpenAI) models.
- PHP is cited as used by roughly 70% of websites and underpins platforms such as WordPress, Moodle and Magento.
- Impact in Brazil: Anatel suspended external access to SEI and other internet-facing services. TRF1, TRF4 and TRF6 took similar measures, and TRF6 also suspended its electronic judicial process system. TSE was asked whether its systems are affected.
Remediation
- Update PHP to the fixed release, which is available.
- Where immediate patching is not possible, take vulnerable systems offline or suspend external access.
- Technically assess all services that use PHP.
-
PHP2Shell: two critical PHP engine flaws leaked before disclosure, emergency patches issued, exploitation attempts since 17 October
ler a análise desta fonte ocultar
Two critical vulnerabilities in the PHP engine, collectively referred to as PHP2Shell, prompted emergency security releases on 21 October 2026 after details leaked on 20 October ahead of the planned disclosure. The UK NCSC urged organisations to identify affected systems and update immediately, and Amazon threat intelligence reported exploitation attempts dating back to 17 October. PHP rates both flaws 9.2 (some early reports cited 10.0). Commentators say the flaws could allow remote access without authentication, sit in code dating back to 2004, and were found with AI assistance. The article does not give CVE identifiers.
Technical Details
- The flaws are in the PHP runtime itself, so updating WordPress or other frameworks does not fix them. PHP must be updated and the services that run it restarted.
- The leak surfaced on Hacker News two days before the planned disclosure, and PHP moved its security release up by a day. The source of the leak is unknown, no threat group involvement is evidenced, and no exploit code was released, although a screenshot of the researchers' exploit tool circulated.
- Forgotten PHP is likely to be found in vendor appliances (NAS, firewalls, VPN portals, printers, building-management panels, OT/ICS HMIs), bundled runtimes in control panels, containers on old base images, marketing microsites and acquired-company infrastructure, admin tools (phpMyAdmin, Roundcube, Nextcloud), and origins reachable directly behind a CDN or WAF.
- Older and end-of-life PHP versions in appliances and containers may never receive an official fix.
- Hunting indicators suggested by commentators: php-fpm, httpd or apache2 spawning sh, bash, curl, wget or base64; new or modified .php files in web roots; outbound connections from web worker processes.
- One expert warns that attackers could distribute malicious "patches" exploiting the rush to fix.
Remediation
- Inventory all PHP use, prioritising internet-facing systems and not limiting the search to outdated versions. Combine external exposure scanning, host package managers, container/SBOM scans and network behaviour analysis.
- Update to the latest PHP version, apply the PHP Team's additional hardening steps, and restart services. Use the PHP Team's temporary mitigations if updating is not possible.
- If mitigations cannot be applied quickly, the NCSC suggests taking the system offline until it is fully mitigated.
- Monitor third-party vendors for updates to products that bundle PHP, and validate patches against official release notes.
- If compromise is suspected, rotate secrets on exposed hosts, filter multipart requests at the edge, restrict egress from web workers, and isolate EOL hosts.
-
PHP2Shell critical PHP flaws exploited in the wild from Oct 17, four days before public disclosure and patches
ler a análise desta fonte ocultar
Amazon Web Services (AWS) research indicates that at least one actor began attempting to exploit two critical PHP vulnerabilities, collectively called PHP2Shell, on October 17. That was four days before the fixes and public disclosure on Wednesday, October 21, at 10 am ET. AWS found the activity by reviewing honeypot logs. The attempt came from an IP address linked to a "commercial anonymization network" (a proxy network or VPN). It remains unclear how the actor obtained exploit information while the flaws were under embargo. AWS CISO CJ Moses stated that "PHP2Shell is now being broadly exploited in the wild."
Technical Details
- Proof-of-concept exploits began appearing publicly on October 21, some apparently hours before the 10 am ET disclosure.
- In the hours before disclosure, AWS honeypots received multiple exploitation attempts. They were anonymized through various providers and had seemingly different attack goals, likely indicating multiple actors.
- AWS says technical analysis is ongoing and it is unclear whether any of these attempts would have succeeded against real-world systems.
- Attackers also reverse engineered the fixes to learn how to exploit the flaws.
- AWS argues that AI-assisted analysis makes it cheaper to turn a patch diff or advisory into a working exploit, so more actors can act within hours or minutes of disclosure.
- The article does not give CVE identifiers, affected versions or IOCs.
Remediation
- Administrators raced to identify and patch vulnerable websites and apps after the fixes were released on October 21.
- Some governments and other groups proactively took key websites offline ahead of the release so they would not be targeted before patching.
- AWS says coordinated vulnerability disclosure must evolve and speed up.
-
Two critical unauthenticated PHP RCE flaws (CVSS 9.2) patched; early exploitation reported, sites taken offline
ler a análise desta fonte ocultar
Two critical vulnerabilities in the PHP language, each rated CVSS 9.2, allow unauthenticated remote code execution on vulnerable servers. Governments and companies reportedly took sites offline as a precaution. The PHP Foundation released four updates on 21 October 2026. One flaw was first identified by a researcher in 2024, and the Foundation described the delay in fixing it as a structural process failure. Cloudflare's CTO reported the first attacks, while CISA said it had not observed exploitation at the time of the Reuters report. Experts warn that attackers may reverse-engineer the patches to target unpatched systems.
Technical Details
- CERT-EU confirms the flaws permit remote code execution without authentication or an account.
- Possible impact: data theft, web page defacement, or full system compromise. The NSA warned they could expose server files and databases.
- PHP runs on nearly 70% of sites whose server-side language can be identified, and underpins WordPress (about 40% of websites). Many organizations may not know they are running PHP in third-party products, CMSs or hosting platforms.
- Discovery: Calif, working with OpenAI and using the GPT-5.6 Sol model, reported one flaw on 11 August. Wiz independently found the issues and reported a second flaw on 24 August, also using AI models.
- Don Smith (Sophos CTU) warns that exploitation could be prioritized within hours or days of patch release, including with AI assistance.
Remediation
- Apply the PHP updates immediately; if not possible, isolate or take vulnerable systems offline (NSA, CISA).
- Inventory all PHP instances, including third-party components, and prioritize internet-facing services.
- Confirm affected versions against official advisories.
- Monitor PHP and web server processes for unusual activity or unexpected command execution, and check whether compromise occurred before patching.
-
Two critical PHP vulnerabilities ('PHP2Shell') under exploitation attempts after leaked details
ler a análise desta fonte ocultar
Security teams are racing to remediate two critical PHP vulnerabilities, collectively dubbed PHP2Shell, after technical details leaked and exploitation attempts were reported. The article excerpt gives no CVE identifiers, affected versions or attribution. The name suggests a path to shell access, but the excerpt does not confirm the impact. Given PHP's footprint across web infrastructure, defenders should treat this as urgent while waiting for vendor guidance.
Technical Details
The available text states only that there are two critical vulnerabilities, that details have leaked, and that exploitation attempts have been reported. It gives no root cause, affected versions or indicators.
Remediation
The excerpt gives no specific remediation steps. Monitor PHP project advisories and apply patches once available.
-
NSA warns of two critical decades-old PHP flaws found with AI by OpenAI/Calif and Wiz; patched in PHP 8.5.12
ler a análise desta fonte ocultar
The NSA issued a warning on Wednesday about two critical vulnerabilities in PHP, present in all versions since PHP 5 (2004). The PHP Foundation has already released patches. The flaws let attackers reach the server file system or the database used by the web application. The NSA advises that any site that cannot be patched immediately should be taken offline until it can be. No in-the-wild exploitation has been reported so far, but public disclosure starts a race to patch tens of millions of sites.
Technical Details
- One flaw was found by the security firm Calif, working under OpenAI's "Patch the Planet" initiative with GPT-5.6 Sol. Separately, Wiz found multiple PHP vulnerabilities using its Wiz Atlas AI agent, Gemini 4 Argon and GPT-5.6 Sol.
- The flaws were previously considered too hard to exploit. Wiz warns that LLMs can now help adapt exploits to different PHP versions, configurations and memory layouts, which lowers that barrier.
- PHP is reported to power roughly 70% of websites (over 33 million sites) and underpins CMSes such as WordPress. WordPress says the issue is not WordPress-specific.
- The article gives no CVE identifiers or technical root causes.
Remediation
- Update to the current PHP release, 8.5.12. PHP 8.1 and older are unsupported, so those sites need to be upgraded.
- Pull sites offline if they cannot be patched immediately (NSA guidance).
- WordPress site owners must patch their servers and work with their hosting providers.
- Wiz thanked Automattic, Akamai, AWS, Cloudflare, Fastly, Imperva, Microsoft and Vercel for helping validate the patches.