Aug 5, 2026·6 min read·11 visits
A validation flaw in Electron allowed sandboxed iframes to bypass security restrictions and launch OS-registered external protocol handlers without authorization.
Improper access control in Electron versions prior to 39.8.8, 40.9.0, 41.2.1, and 42.0.0-beta.3 allowed sandboxed iframes to bypass sandbox restrictions and trigger external application protocols on the host operating system. The application's custom permission handler was also not provided with the frame's sandbox state, preventing effective validation of the request context.
Electron is a widely adopted framework designed for building desktop applications using Chromium and Node.js. A critical component of its security architecture is the multi-process model, where web pages run within isolated renderer processes, and privileged operations are delegated to the main (browser) process. To securely isolate third-party or untrusted content, developers routinely load arbitrary web resources inside an <iframe> configured with the HTML5 sandbox attribute.
In standard Chromium architectures, a sandboxed iframe is restricted from performing actions that cross security boundaries, such as navigating the top-level window to custom external protocols (e.g., skype:, zoommtg:, mailto:, or custom URI handlers). These restrictions prevent untrusted code from executing arbitrary commands or calling third-party system applications that might contain vulnerability surface areas of their own.
CVE-2026-70612 represents an access control breakdown where Electron overridden handlers failed to propagate the sandbox state from the Blink renderer layer to the host application's decision engine. Consequently, a nested sandboxed subframe could execute a navigation request to arbitrary custom schemes, forcing the operating system to launch associated software binary packages outside the Electron sandbox boundary.
The root cause of this vulnerability lies in Electron's custom implementation of the Chromium browser-side interface HandleExternalProtocol. Chromium implements defensive checks inside its navigation pipeline to prevent sandboxed documents from launching external applications. These validations assess the initiating document's WebSandboxFlags to ensure that custom protocol navigations are explicitly permitted by the sandbox configuration (such as having the allow-top-navigation-to-custom-protocols or allow-top-navigation tokens).
Electron overrides ContentBrowserClient::HandleExternalProtocol inside its browser core (shell/browser/electron_browser_client.cc) to hook navigation requests and direct them through Electron's JS-facing shell.openExternal permission system. However, the overridden signature of HandleExternalProtocol did not preserve or process the sandbox_flags and is_primary_main_frame boolean arguments that Chromium's content layer provides to indicate the structural origin and isolation constraints of the requesting frame.
Because these critical parameters were omitted when transitioning execution to the UI thread via HandleExternalProtocolInUI, the application's internal handler operated under the false assumption that all requests were initiated by a trusted, top-level frame. Furthermore, because the frame's sandbox state was never populated, any user-configured setPermissionRequestHandler callbacks executed without vital contextual information, causing default-allow handlers to authorize the launch of external software packages.
The security patch introduced across the active branches of Electron directly resolves the missing state propagation and adds the required validation steps inside the core browser client.
// Vulnerable method signature in shell/browser/electron_browser_client.cc
void HandleExternalProtocolInUI(
const GURL& url,
content::WeakDocumentPtr document_ptr,
content::WebContents::OnceGetter web_contents_getter,
bool has_user_gesture) {
// Lack of is_primary_main_frame and sandbox_flags allowed processing of sandboxed iframe requests
}To remediate this, the signature was updated to include the network mojom sandbox flags and structural boolean values:
// Patched method signature and logic in shell/browser/electron_browser_client.cc
void HandleExternalProtocolInUI(
const GURL& url,
content::WeakDocumentPtr document_ptr,
content::WebContents::OnceGetter web_contents_getter,
bool has_user_gesture,
bool is_primary_main_frame,
network::mojom::WebSandboxFlags sandbox_flags) {
content::WebContents* web_contents = std::move(web_contents_getter).Run();
if (!web_contents)
return;
content::RenderFrameHost* rfh = nullptr;
if (document_ptr) {
rfh = document_ptr.GetAsRenderFrameHost();
}
if (!rfh) {
rfh = web_contents->GetPrimaryMainFrame();
}
// Sandboxed iframes without one of the appropriate sandbox-escape tokens
// must not be able to launch external protocol handlers.
if (!is_primary_main_frame) {
using SandboxFlags = network::mojom::WebSandboxFlags;
auto allow = [sandbox_flags](SandboxFlags flag) {
return (sandbox_flags & flag) == SandboxFlags::kNone;
};
const bool allowed = allow(SandboxFlags::kTopNavigationToCustomProtocols) ||
(allow(SandboxFlags::kTopNavigationByUserActivation) &&
has_user_gesture);
if (!allowed) {
rfh->AddMessageToConsole(
blink::mojom::ConsoleMessageLevel::kError,
"Navigation to external protocol blocked by sandbox, because it "
"doesn't contain any of: "
"'allow-top-navigation-to-custom-protocols', "
"'allow-top-navigation-by-user-activation', "
"'allow-top-navigation', or 'allow-popups'.");
return;
}
}
GURL escaped_url(base::EscapeExternalHandlerValue(url.spec()));
auto callback = base::BindOnce(&OnOpenExternal, escaped_url);
permission_helper->RequestOpenExternalPermission(rfh, std::move(callback), has_user_gesture);
}The fix successfully aligns Electron's navigation pathway with standard Chromium enforcement structures (crbug.com/1148777). By evaluating the bitwise intersection of sandbox_flags, the application drops unauthorized protocol navigation attempts on subframes immediately, outputting an explicit warning to the developer console.
To exploit this vulnerability, an attacker must have the ability to execute untrusted code (such as via Cross-Site Scripting or hosting a malicious third-party site) inside an iframe of an affected Electron application. The hosting window would load the attacker-controlled URL with standard sandbox restrictions:
<iframe sandbox="allow-scripts" src="https://attacker.com/payload.html"></iframe>Once loaded inside the sandboxed iframe, the attacker executes JavaScript to trigger a navigation event targeting a specific protocol registered on the victim's host operating system. Because the sandbox attributes are bypassed during external protocol evaluation in vulnerable Electron versions, this command is forwarded directly to the operating system:
// Inside payload.html
window.location.href = "ms-settings:workplace";
// Or other application-defined protocols known to accept argument parametersThe operating system then resolves the custom scheme handler and attempts to execute the target binary. Depending on the nature of the registered protocol handlers on the user's workstation, this can lead to arguments being injected into command-line utilities, potentially achieving remote code execution (RCE) on the host computer system.
The security threat posed by this vulnerability varies from system to system, depending heavily on the external protocol handlers configured on the underlying operating system. The base vulnerability allows a sandboxed subframe context to achieve sandbox escape by shifting execution flow to processes external to Electron.
In scenarios where the host system has vulnerable handlers registered (for example, utility handlers that perform insecure file actions or allow command execution through argument manipulation), an unauthenticated attacker could transition from a low-privilege script execution environment within an iframe to full remote code execution inside the operating system. This represents a significant breakdown of the browser-host containment boundaries.
Because the scope of the containment has been altered (S:C), and no user interaction is required for the nested iframe to trigger the navigation scheme, this vulnerability represents a medium-severity issue with a CVSS base rating of 5.4. Applications that run general-purpose or untrusted web clients within sandboxed frames are at the highest level of risk.
The ultimate remediation is to upgrade to a version of Electron that contains the official validation fix. These patches are backported to the primary support branches:
If upgrading the primary Electron package is not immediately feasible, developers can employ defensive mitigations. Specifically, applications must explicitly intercept permission requests and drop external protocol attempts by overriding the session's permission request handler:
const { session } = require('electron');
session.defaultSession.setPermissionRequestHandler((webContents, permission, callback, details) => {
if (permission === 'openExternal') {
// Implement strong origin-based structural validation
const url = details.externalURL;
if (isTrustedProtocolAndOrigin(url, webContents.getURL())) {
return callback(true);
}
// Default to denying untrusted requests
return callback(false);
}
callback(true);
});Implementing strict Content Security Policies (CSP) within nested iframes can also prevent unintended navigation redirections and execution vectors.
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:L/I:L/A:N| Product | Affected Versions | Fixed Version |
|---|---|---|
electron Electron | < 39.8.8 | 39.8.8 |
electron Electron | >= 40.0.0-alpha.1 < 40.9.0 | 40.9.0 |
electron Electron | >= 41.0.0-alpha.1 < 41.2.1 | 41.2.1 |
electron Electron | >= 42.0.0-alpha.1 < 42.0.0-beta.3 | 42.0.0-beta.3 |
| Attribute | Detail |
|---|---|
| CWE ID | CWE-284 (Improper Access Control) |
| Attack Vector | Network |
| CVSS v3.1 Score | 5.4 |
| EPSS Score | Not indexed |
| Impact | Medium (Scope Change, potential host-level command execution) |
| Exploit Status | Proof of Concept (PoC) available |
| KEV Status | Not listed in CISA KEV |
The product does not restrict or incorrectly restricts access to a resource from an unauthorized actor.
CVE-2026-12243 is a path traversal vulnerability in the Natural Language Toolkit (NLTK) version 3.9.4. The flaw exists because the input validation routine fails to account for percent-encoded directory traversal sequences like '..%2f' before passing them to urllib.request.url2pathname(), which decodes them into active traversal sequences.
CVE-2026-73654 is a high-severity prototype pollution vulnerability in Trigger.dev. The flaw occurs during the handling of run-metadata updates through the PUT /api/v1/runs/:runId/metadata endpoint. Because user-supplied keys are parsed directly by the @jsonhero/path library without sanitization, an authenticated attacker with low privileges can pollute the global Object.prototype. This causes database queries via Prisma ORM to fail validation and induces unhandled exceptions in the Prometheus metrics client, resulting in a process-wide denial of service.
CVE-2026-73559 is an uncontrolled resource consumption vulnerability in the vLLM engine, specifically within the /v1/completions API endpoint, allowing authenticated attackers to cause application-level denial of service via unbounded prompt arrays.
A critical security bypass vulnerability in Argo Workflows allows authenticated attackers with workflow submission privileges to bypass 'Strict' or 'Secure' template referencing restrictions. By injecting unvalidated fields into the nested ArtifactGC configuration, attackers can execute arbitrary pod patches, leading to host namespace escape and cluster-wide privilege escalation.
A Server-Side Request Forgery (SSRF) / Confused Deputy vulnerability has been identified in Pydantic AI UI Adapters (such as VercelAIAdapter). Under certain conditions, a malicious client can supply manipulated message history with provider metadata that forces the server to resolve files within privileged cloud environments (AWS S3, Google Cloud Storage) or model providers. This occurs because the adapters deserialize client-provided metadata structures directly into UploadedFile instances without validation, which are subsequently fetched using high-privilege server credentials.
A path traversal vulnerability in the optional dashboard server of atomic-agents-stack before version 1.1.0 allows unauthenticated remote attackers to read arbitrary files from the host filesystem.