Skip to main content
Mallory
CriticalPublic exploitVulnerability

CVE-2026-121529: Fortinet2Shell

Published Nov 10, 2026 · Updated 1 day ago
Trending this week#60of 428,812 vulnerabilities
Mention activity · 30 days25 mentions · peak 15 in a dayToday
CVSS9.6Critical · v3.1
EPSS0.22%12th percentile
CISA KEVNoNot listed
In the wildNonePublic exploit, no exploitation reported
Vendor
Fortinet
IdentifiersCVE-2026-121529CWE-346 · Origin Validation Error

Fortinet2Shell is an origin-validation vulnerability in the local API endpoint of Fortinet FortiClient for Windows and macOS, including FortiClient Windows PAM. Improper origin validation, associated with a CORS misconfiguration, permits an attacker-controlled webpage to send requests to a local FortiClient service and cause execution of unauthorized code or commands. Exploitation requires victim interaction, such as visiting a malicious website, but does not require attacker authentication. Affected releases include FortiClient Mac 7.4.0–7.4.7, FortiClient Windows 7.4.0–7.4.7 and 7.2.0–7.2.14, and FortiClient Windows PAM 7.4.7 and 7.2.14.

Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Analyst brief

Impact, mitigation & remediation

What it means and what to do now: patch path, interim mitigations, and the assume-compromise checklist.

Impact

What an attacker gets, and what they’ve been doing with it.

Successful exploitation enables unauthorized code or command execution on the affected endpoint and can result in privilege escalation. Depending on the execution privileges obtained, an attacker could install malware, access or modify sensitive information, disrupt endpoint operation, or use the compromised endpoint to access other network systems. The exact execution security context is not specified.

Mitigation

If you can’t patch tonight, do this now.

No vendor workaround or virtual patch is available. Pending updates, reduce exposure to untrusted webpages through web filtering and user warnings, limit unnecessary administrative privileges, and strengthen endpoint monitoring. Review unexpected processes or commands originating from FortiClient, unusual local API activity, privilege changes, and anomalous outbound connections. These measures reduce risk or support detection but do not remediate the vulnerability.

Remediation

Patch, then assume compromise.

Upgrade affected FortiClient Mac 7.4 installations to 7.4.8 or later. Upgrade affected FortiClient Windows and FortiClient Windows PAM 7.4 installations to 7.4.8 or later, and affected 7.2 installations to 7.2.15 or later. The 8.0 branches and FortiClient Mac 7.2 are listed as unaffected. Prioritize privileged-user endpoints and Windows PAM installations, and verify installed versions after deployment.

Public exploits

Exploits

4 exploits validated by Mallory's analysis. Hidden: 2 rejected by Mallory's analysis as not a working exploit.

View more in app →
CVE-2026-121529This is a standalone, containerized exploit demonstration and mitigation-testing laboratory attributed by its authors to CVE-2026-121529/CWE-346. It contains 26 files, including three Python programs, a Bash driver, a Makefile, two Containerfiles, documentation, and 14 captured evidence files. Embedded HTML/JavaScript supplies the browser attack. Seven files are counted as executable code or build/orchestration definitions. The repository URL comes from citation metadata; no analyzed git reference or archive byte size was supplied, so those fields use empty/zero placeholders. The central limitation is explicit: lab/forticlient_api.py is an intentionally vulnerable custom API model, not FortiClient code. Its command routes, WebSocket implementation, allowlist, shared secret, version banners, and hypothetical patch behaviors do not establish equivalent implementation details in the vendor product. The claimed CVE, advisory, affected ranges, and fixed releases were not independently verified. Several supplied files are truncated, limiting full implementation review. lab/attacker_server.py serves a drive-by page and collects reports. Its JavaScript submits attacker-controlled commands using preflighted JSON POST, simple text/plain POST, a legacy route, an image GET without Origin, and a WebSocket. It then loads a command-created PNG as a blind confirmation oracle and posts results to /report. poc/poc.py drives Chromium with Playwright, correlates execution proofs, and offers check, exploit, mitigation, origin-matrix, raw HTTP, WebSocket, and simulated rebinding modes. Even check mode executes the marker command; this is not a passive detector. FCT_MODE selects four model behaviors. Captured results show execution and readable responses in vuln; continued execution and readable WebSocket responses in fixed-cors; blind image-GET execution in fixed-origin-only; and no execution in fixed-strict, which combines origin checks with a shared-secret header. The evidence records uid 0 in a Linux container, not demonstrated Windows SYSTEM access. The version 7.4.8 emitted by hypothetical remediation models is not evidence of a real vendor patch bypass or new product zero-day. Makefile and run-lab.sh build and run the target, attacker, and browser containers on a Podman network, with the browser sharing each target's network namespace. Container definitions use Python 3.12 and Playwright 1.56.0. Evidence includes JSON verdicts, raw HTTP responses, attacker reports, a four-way comparison, and command output. Browser flags disable sandboxing and some isolation features; origin-matrix tests additionally tamper with browser requests through Playwright, and rebinding is emulated. Consequently, those tests are not all naturally achievable by an ordinary web page. Result interpretation also requires care: strict-model records still show marker_present and exfiltrated as true despite zero execution proofs. Shared marker artifacts can survive prior runs, and successful delivery of an attacker report is not itself exfiltration of target data. Execution proofs and readable response bodies are stronger indicators than those booleans. Raw vulnerable probes list execution hits while returning exit code 0, so exit status alone is insufficient. No unrelated malicious operator payload is apparent; rm -rf occurrences are scoped to lab results or temporary cleanup.googlesky · Disclosed Nov 11, 2026 · python · javascript · browser · web · localOperationalFortinet2ShellThis standalone repository is an executable, containerized demonstration of browser-to-loopback origin-validation failures, attributed by its documentation to CVE-2026-121529 / FG-IR-26-194. It explicitly contains no proprietary FortiClient implementation or binaries. Its API routes, WebSocket surface, and remediation variants are synthetic; the supplied evidence does not establish a working exploit against the actual product or a vendor patch bypass. The advisory, CVE validity, affected-version claims, and claimed fixed versions were not independently verified. The 26-file inventory contains three Python programs, one Bash driver, a Makefile, two container build files, six metadata/documentation files, and thirteen evidence files. JavaScript is embedded in lab/attacker_server.py. lab/forticlient_api.py implements the command-executing model; lab/attacker_server.py serves a wallpaper-themed malicious page and collects reports; poc/poc.py drives Chromium through Playwright and evaluates execution proofs. run-lab.sh and Makefile orchestrate resource-limited Podman containers. The target image uses Python 3.12-slim; the runner uses the Playwright 1.56.0 image and matching Python package. Documentation describes scope, claimed vulnerability details, and security reporting. Several supplied source files and logs are truncated, limiting complete implementation review. The page tries JSON POST, simple text/plain POST, a legacy POST endpoint, an image GET carrying a command, and a WebSocket message. It then loads a generated PNG as a blind-execution oracle and submits attempt metadata and readable responses to its own /report endpoint. The default shell payload writes identity output and a PWNED marker rather than installing a shell or persistence. Browser-controlled command and endpoint parameters support customization. Four FCT_MODE policies demonstrate distinct results. Supplied logs show arbitrary-origin reflection plus credentialed CORS in vuln mode, yielding execution and readable responses. fixed-cors restricts response headers but leaves actions reachable; WebSocket responses remain readable because WebSocket handshakes require their own server-side origin checks. fixed-origin-only rejects untrusted explicit origins but accepts a missing Origin, allowing command execution through an image GET while keeping the command response unreadable. fixed-strict combines origin validation with the x-fortipam-client shared-secret header and blocks the supplied attack. Its default secret is hardcoded for the model and should not be interpreted as production authentication. The models' 7.4.7/7.4.8 banners are configured labels, not evidence about vendor binaries. The runner additionally supports a documented 17-candidate origin matrix with and without the known lab token, nine raw HTTP probes, and rebinding-style request rewriting. Header manipulation and destination rewriting use Playwright interception, so these tests are not all capabilities available to ordinary hostile webpage JavaScript. Supplied strict-model raw and rebinding logs show no execution. Documentation reports a trailing-whitespace parser discrepancy, but it does not establish a readable real-product bypass. Evidence interpretation requires care: fixed-model logs can still contain marker_present=true and exfiltrated=true despite executed=false because all models share an artifact volume and the page reports failed attempts too. These flags alone do not prove fresh execution or sensitive-data theft. Raw-probe mode also returns exit code 0 even when raw_hits contains successful executions; consumers must inspect the result fields. Scoped rm -rf commands clean generated results or temporary files and are not evidence of a malicious fake exploit. Repository URL comes from citation metadata; the analyzed git reference and archive size were unavailable, with size_bytes=0 representing unknown.googlesky · Disclosed Nov 11, 2026 · python · javascript · browser · web · localOperationalCVE-2026-121529-ToolkitStandalone Python toolkit claiming exploitation of CVE-2026-121529, an Origin-validation weakness in the FortiClient/FortiPAM privileged-session API. The six-file repository contains two Python entry points, a README, an MIT license, a PySocks dependency file, and ignore rules for logs, reports, and development artifacts. No established exploit framework is identified. exploit.py visibly defines the privileged-session endpoint, command defaults, Windows enumeration commands, randomized User-Agent selection, configurable request jitter, SOCKS proxy handling, rotating logs, concurrent target processing, report generation calls, and dispatch to browser-server and interactive-shell functions. Its central implementation is truncated. The README describes POST-based command execution, a malicious-page server, browser-side output exfiltration, command chaining, file transfers, retries, and JSON reporting; those capabilities cannot all be verified from the visible code. No fixed exfiltration destination is shown. verifier.py is the companion detection utility. Visible code provides TCP reachability checks, HTTP/HTTPS requests, a synthetic attacker Origin, root and privileged-session probe paths, concurrent scanning, and reporting calls. HTTPS certificate verification is disabled. The README describes OPTIONS-based CORS checks, WAF heuristics, and JSON/CSV reporting without execution payloads, but the core CORS-check logic is also truncated. CORS exposure alone would not establish arbitrary command execution or SYSTEM privileges, and the listed HTTP error codes are not reliable proof of a WAF. The source is exploit-oriented rather than README-only or detection-only, but successful exploitation is not demonstrated. The CVE assignment, affected versions, severity, patch versions, macOS applicability, and WebSocket claims remain unverified assertions. Browser restrictions may prevent loopback access independently of CORS. There is no visible evidence of a deliberately malicious fake exploit, but the incomplete excerpts prevent full validation. The repository URL is taken from the README; no analyzed git reference or archive size was supplied, so ref is unknown and size_bytes uses 0 as an unavailable-value placeholder. Listed file sizes total 52,338 bytes, which is not an archive-size measurement.tc4dy · Disclosed Nov 13, 2026 · python · browser · local · networkOperationalfortinet-exploitsThe supplied repository listing contains 22 files: a root README and license, plus five CVE-named directories, each containing README.md, LICENSE, poc.py, and exploit.py. The ten standalone Python entry points use primarily standard-library HTTP, TLS, socket, argument-parsing, JSON, and concurrency facilities. PHP and Bash appear as embedded payload languages. No recognized exploitation-framework integration, dependency manifest, or test suite is listed. Many source bodies are explicitly truncated, limiting validation of protocol handling and payload construction. The original repository URL, git reference, and archive size were not supplied; empty strings and zero represent unavailable metadata, not a measured archive size. The FortiSIEM pair claims unauthenticated gRPC filename traversal on TCP/27918. Visible constants specify a text-file write test and a PHP webshell at ../../web/ng/wp-cron.php, accessed through /ng/wp-cron.php. The exploit tail supports command execution and a Bash reverse shell and invokes escalation and persistence helpers. Their implementations and the actual gRPC request construction are omitted. The FortiSandbox authentication-bypass PoC defines privileged REST paths, encoded and duplicate-slash variants, and localhost/proxy-header spoofing combinations. Its companion advertises administrator creation and exposes command-execution, reverse-shell, and post-exploitation call sites, but the complete privileged requests are not visible. The FortiClient pair checks service exposure and CORS behavior on TCP/9191, supports threaded target lists and JSON output, and advertises an HTTP server delivering a cross-origin payload at /payload, normally on port 8080. Permissive CORS alone does not establish operating-system command execution; the supplied excerpts omit the privileged API calls and browser payload needed to substantiate that claim. The FortiManager/FortiAnalyzer pair targets /jsonrpc, normally over HTTPS/443, with scanning, information collection, configurable commands, and threaded execution attempts. The README illustrates command substitution in an SMTP-test server field, but the actual injection builder is truncated. An accessible JSON-RPC endpoint is not by itself evidence of unauthenticated RCE. The most completely visible attack routine is the FortiSandbox command-injection PoC: it sends a semicolon-containing jid value, redirects output to /tmp/out_39808.txt, and then retrieves /ng/test.txt. No connection between those files is shown. It treats any nonempty HTTP 200 response as success without validating command-specific output, creating a substantial false-positive risk. Documentation instead describes pipe injection and /web/ng/test.txt. Additional inconsistencies include command-injection modules labeled as path traversal, conflicting affected-version text, incompatible README command-line examples, and differing advisory identifiers. The collection contains exploit-intended code and concrete payloads rather than only detection scripts, but operational maturity here describes implementation intent, not demonstrated success. There is insufficient evidence to label it deliberately fake; equally, none of the claimed CVE assignments, advisories, patch versions, root-execution assertions, CISA KEV inclusion, or in-the-wild exploitation claims is independently established by the supplied material. No fixed external exfiltration or callback infrastructure is visible.Ghost-Bx1 · Disclosed Nov 16, 2026 · python · php · network · web · browserOperational
Exposure surface

Affected products & vendors

Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.

VendorProductType
FortinetForticlientmacapplication
FortinetForticlientwindowsapplication
FortinetForticlientwindowspamapplication
FAQ

Frequently asked questions

Is CVE-2026-121529 exploited in the wild?
As of Nov 16, 2026, Mallory has no reports of CVE-2026-121529 being exploited in the wild. It is not listed in CISA KEV.
Is there a public exploit for CVE-2026-121529?
Yes. Mallory tracks 4 public exploits or proofs of concept for CVE-2026-121529, the earliest disclosed Nov 11, 2026.
Is there a patch for CVE-2026-121529?
Upgrade affected FortiClient Mac 7.4 installations to 7.4.8 or later. Upgrade affected FortiClient Windows and FortiClient Windows PAM 7.4 installations to 7.4.8 or later, and affected 7.2 installations to 7.2.15 or later.
What products are affected by CVE-2026-121529?
Fortinet Forticlientmac, Fortinet Forticlientwindows, and Fortinet Forticlientwindowspam.
How severe is CVE-2026-121529?
CVE-2026-121529 has a CVSS v3.1 score of 9.6 (Critical) and an EPSS score of 0.22% (12th percentile) as of Nov 16, 2026.
Activity feed

Recent activity

26 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.

View more in app →
csirt bjNews · Nov 13, 2026Exécution de code arbitraire à distance affectant Fortinet FortiClient - bjCSIRTThe advisory describes a critical Fortinet FortiClient vulnerability caused by insufficient origin validation of requests to its local API. An unauthenticated attacker could induce a user to visit a malicious webpage, send commands to that API, and execute arbitrary code on the endpoint, compromising confidentiality and integrity. The reported CVSS v3.1 score is 9.6. The advisory recommends upgrading to versions 7.4.8, 7.2.15, 8.0.0, or later.Read more ↗brucertNews · Nov 12, 2026Critical Remote Code Execution Vulnerability in Fortinet FortiClient (Forti2Shell) CVE-2026-121529 | BruCERT - Brunei Computer Emergency Response TeamA critical improper-origin-validation vulnerability in a FortiClient local API endpoint, rated CVSS 9.1. An unauthenticated remote attacker could execute unauthorized code or commands, potentially gaining elevated privileges. Exploitation requires user interaction, such as visiting a malicious webpage. Affected versions are FortiClient Windows and Windows PAM 7.4.0–7.4.7 and 7.2.0–7.2.14, and FortiClient Mac 7.4.0–7.4.7. Fixed releases are 7.4.8 or later and, for affected Windows 7.2 branches, 7.2.15 or later. FortiClient 8.0, FortiClient Mac 7.2, and FortiClient Windows Free VPN are explicitly unaffected.Read more ↗cryptika blogNews · Nov 11, 2026Critical Fortinet2Shell Vulnerability Enables RCE Attacks in FortiClient with FortiPAM Enabled | Cryptika CybersecurityAn origin-validation error caused by a CORS misconfiguration in the FortiClient API component allows an attacker-controlled website to send requests to a local service and execute attacker-supplied code. No authentication is required, but the victim must visit a malicious page. Affected versions are FortiClient Mac 7.4.0–7.4.7 and FortiClient Windows and Windows PAM 7.4.0–7.4.7 and 7.2.0–7.2.14. Upgrade affected 7.4 installations to 7.4.8 or later and affected 7.2 installations to 7.2.15 or later. The 8.0 branch and FortiClient Mac 7.2 are unaffected.Read more ↗cyber security newsNews · Nov 11, 2026Critical Fortinet2Shell Vulnerability Enables RCE Attacks in FortiClient with FortiPAM EnabledA critical, unauthenticated remote code execution vulnerability in FortiClient's API component, caused by improper origin validation and a CORS misconfiguration. An attacker-controlled website can send requests to a local service and execute attacker-supplied code after the victim visits the page. The researcher confirmed reliable code execution. Affected versions include FortiClient Mac 7.4.0–7.4.7 and FortiClient Windows and Windows PAM 7.4.0–7.4.7 and 7.2.0–7.2.14. The 8.0 branch and FortiClient Mac 7.2 are unaffected. Fixed versions are 7.4.8 or later and, for affected 7.2 installations, 7.2.15 or later. Fortinet reports no known exploitation. No virtual patch is available; user warnings and web-filtering reviews are recommended during patch deployment.Read more ↗
+3 more in the timeline. Sign up to see them
Do you run this?See affected versions in your environment and whether adversaries are exploiting this vulnerability against assets like yours.
Check your exposure
Mallory platform

The version that knows your environment.

This page is what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which adversaries are exploiting it right now, and which detections to deploy tonight.
Exposure mappingQuery your assets running an affected version, and investigate the blast radius.
Threat actor evidenceEvery report linking this CVE to a named adversary.
Associated malwareMalware families linked to this CVE, with evidence and IOCs.
Detection signaturesYARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Vendor-by-vendor mappingCross-references every affected SKU, including bundled OEM variants.
Social activity17Community discussion across Reddit, Mastodon, and other social sources.