To protect our users, TAG routinely hunts for 0-day vulnerabilities exploited in-the-wild. In late August 2021, TAG discovered watering hole attacks targeting visitors to Hong Kong websites for a media outlet and a prominent pro-democracy labor and political group. The watering hole served an XNU privilege escalation vulnerability (CVE-2021-30869) unpatched in macOS Catalina, which led to the installation of a previously unreported backdoor.

As is our policy, we quickly reported this 0-day to the vendor (Apple) and a patch was released to protect users from these attacks.

Based on our findings, we believe this threat actor to be a well-resourced group, likely state backed, with access to their own software engineering team based on the quality of the payload code.

In this blog we analyze the technical details of the exploit chain and share IOCs to help teams defend against similar style attacks.

Watering Hole

The websites leveraged for the attacks contained two iframes which served exploits from an attacker-controlled server—one for iOS and the other for macOS.

iOS Exploits

The iOS exploit chain used a framework based on Ironsquirrel to encrypt exploits delivered to the victim's browser. We did not manage to get a complete iOS chain this time, just a partial one where CVE-2019-8506 was used to get code execution in Safari.

macOS Exploits

The macOS exploits did not use the same framework as iOS ones. The landing page contained a simple HTML page loading two scripts—one for Capstone.js and another for the exploit chain.

The parameter rid is a global counter which records the number of exploitation attempts. This number was in the 200s when we obtained the exploit chain.

While the javascript starting the exploit chain checks whether visitors were running macOS Mojave (10.14) or Catalina (10.15) before proceeding to run the exploits, we only observed remnants of an exploit when visiting the site with Mojave but received the full non-encrypted exploit chain when browsing the site with Catalina.

The exploit chain combined an RCE in WebKit exploiting CVE-2021-1789 which was patched on Jan 5, 2021 before discovery of this campaign and a 0-day local privilege escalation in XNU (CVE-2021-30869) patched on Sept 23, 2021.

Remote Code Execution (RCE)

Loading a page with the WebKit RCE on the latest version of Safari (14.1), we learned the RCE was an n-day since it did not successfully trigger the exploit. To verify this hypothesis, we ran git bisect and determined it was fixed in this commit.

Sandbox Escape and Local Privilege Escalation (LPE)

Capstone.js

It was interesting to see the use of Capstone.js, a port of the Capstone disassembly framework, in an exploit chain as Capstone is typically used for binary analysis. The exploit authors primarily used it to search for the addresses of dlopen and dlsym in memory. Once the embedded Mach-O is loaded, the dlopen and dlsym addresses found using Capstone.js are used to patch the Mach-O loaded in memory.

With the Capstone.js configured for X86-64 and not ARM, we can also derive the target hardware is Intel-based Macs.

Embedded Mach-O

After the WebKit RCE succeeds, an embedded Mach-O binary is loaded into memory, patched, and run. Upon analysis, we realized this binary contained code which could escape the Safari sandbox, elevate privileges, and download a second stage from the C2.

Analyzing the Mach-O was reminiscent of a CTF reverse engineering challenge. It had to be extracted and converted into binary from a Uint32Array.

Then the extracted binary was heavily obfuscated with a relatively tedious encoding mechanism--each string is XOR encoded with a different key. Fully decoding the Mach-O was necessary to obtain all the strings representing the dynamically loaded functions used in the binary. There were a lot of strings and decoding them manually would have taken a long time so we wrote a short Python script to make quick work of the obfuscation. The script parsed the Mach-O at each section where the strings were located, then decoded the strings with their respective XOR keys, and patched the binary with the resulting strings.

Once we had all of the strings decoded, it was time to figure out what capabilities the binary had. There was code to download a file from a C2 but we did not come across any URL strings in the Mach-O so we checked the javascript and saw there were two arguments passed when the binary is run–the url for the payload and its size.

After downloading the payload, it removes the quarantine attribute of the file to bypass Gatekeeper. It then elevated privileges to install the payload.

N-day or 0-day?

Before further analyzing how the exploit elevated privileges, we needed to figure out if we were dealing with an N-day or a 0-day vulnerability. An N-day is a known vulnerability with a publicly available patch. Threat actors have used N-days shortly after a patch is released to capitalize on the patching delay of their targets. In contrast, a 0-day is a vulnerability with no available patch which makes it harder to defend against.

Despite the exploit being an executable instead of shellcode, it was not a standalone binary we could run in our virtual environment. It needed the address of dlopen and dlsym patched after the binary was loaded into memory. These two functions are used in conjunction to dynamically load a shared object into memory and retrieve the address of a symbol from it. They are the equivalent of LoadLibrary and GetProcAddress in Windows.

To run the exploit in our virtual environment, we decided to write a loader in Python which did the following:

• load the Mach-O in memory
• find the address of dlopen and dlsym
• patch the loaded Mach-O in memory with the address of dlopen and dlsym
• pass our payload url as a parameter when running the Mach-O

For our payload, we wrote a simple bash script which runs id and pipes the result to a file in /tmp. The result of the id command would tell us whether our script was run as a regular user or as root.

Having a loader and a payload ready, we set out to test the exploit on a fresh install of Catalina (10.15) since it was the version in which we were served the full exploit chain. The exploit worked and ran our bash script as root. We updated our operating system with the latest patch at the time (2021-004) and tried the exploit again. It still worked. We then decided to try it on Big Sur (11.4) where it crashed and gave us the following exception.

The exception indicates that Apple added generic protections in Big Sur which rendered this exploit useless. Since Apple still supports Catalina and pushes security updates for it, we decided to take a deeper look into this exploit.

Elevating Privileges to Root

The Mach-O was calling a lot of undocumented functions as well as XPC calls to mach_msg with a MACH_SEND_SYNC_OVERRIDE flag. This looked similar to an earlier in-the-wild iOS vulnerability analyzed by Ian Beer of Google Project Zero. Beer was able to quickly recognize this exploit as a variant of an earlier port type confusion vulnerability he analyzed in the XNU kernel (CVE-2020-27932). Furthermore, it seems this exact exploit was presented by Pangu Lab in a public talk at zer0con21 in April 2021 and Mobile Security Conference (MOSEC) in July 2021.

In exploiting this port type confusion vulnerability, the exploit authors were able to change the mach port type from IKOT_NAMED_ENTRY to a more privileged port type like IKOT_HOST_SECURITY allowing them to forge their own sec_token and audit_token, and IKOT_HOST_PRIV enabling them to spoof messages to kuncd.

MACMA Payload

After gaining root, the downloaded payload is loaded and run in the background on the victim's machine via launchtl. The payload seems to be a product of extensive software engineering. It uses a publish-subscribe model via a Data Distribution Service (DDS) framework for communicating with the C2. It also has several components, some of which appear to be configured as modules. For example, the payload we obtained contained a kernel module for capturing keystrokes. There are also other functionalities built-in to the components which were not directly accessed from the binaries included in the payload but may be used by additional stages which can be downloaded onto the victim's machine.

Notable features for this backdoor include:

• victim device fingerprinting
• screen capture
• file download/upload
• executing terminal commands
• audio recording
• keylogging

Conclusion

Our team is constantly working to secure our users and keep them safe from targeted attacks like this one. We continue to collaborate with internal teams like Google Safe Browsing to block domains and IPs used for exploit delivery and industry partners like Apple to mitigate vulnerabilities. We are appreciative of Apple’s quick response and patching of this critical vulnerability.

For those interested in following our in-the-wild work, we will soon publish details surrounding another, unrelated campaign we discovered using two Chrome 0-days (CVE-2021-37973 and CVE-2021-37976). That campaign is not connected to the one described in today’s post.

Related IOCs

Delivery URLs

• http://103[.]255[.]44[.]56:8372/6nE5dJzUM2wV.html
• http://103[.]255[.]44[.]56:8371/00AnW8Lt0NEM.html
• http://103[.]255[.]44[.]56:8371/SxYm5vpo2mGJ?rid=<redacted>
• http://103[.]255[.]44[.]56:8371/iWBveXrdvQYQ?rid=?rid=<redacted>
• https://appleid-server[.]com/EvgSOu39KPfT.html
• https://www[.]apple-webservice[.]com/7pvWM74VUSn2.html
• https://appleid-server[.]com/server.enc
• https://amnestyhk[.]org/ss/defaultaa.html
• https://amnestyhk[.]org/ss/4ba29d5b72266b28.html
• https://amnestyhk[.]org/ss/mac.js

Javascript

• cbbfd767774de9fecc4f8d2bdc4c23595c804113a3f6246ec4dfe2b47cb4d34c (capstone.js)
• bc6e488e297241864417ada3c2ab9e21539161b03391fc567b3f1e47eb5cfef9 (mac.js)
• 9d9695f5bb10a11056bf143ab79b496b1a138fbeb56db30f14636eed62e766f8

Sandbox escape / LPE

• 8fae0d5860aa44b5c7260ef7a0b277bcddae8c02cea7d3a9c19f1a40388c223f
• df5b588f555cccdf4bbf695158b10b5d3a5f463da7e36d26bdf8b7ba0f8ed144

Backdoor

• cf5edcff4053e29cb236d3ed1fe06ca93ae6f64f26e25117d68ee130b9bc60c8 (2021 sample)
• f0b12413c9d291e3b9edd1ed1496af7712184a63c066e1d5b2bb528376d66ebc (2019 sample)

C2

• 123.1.170.152
• 207.148.102.208

Google’s Threat Analysis Group tracks actors involved in disinformation campaigns, government backed hacking, and financially motivated abuse. Since late 2019, our team has disrupted financially motivated phishing campaigns targeting YouTubers with Cookie Theft malware.

The actors behind this campaign, which we attribute to a group of hackers recruited in a Russian-speaking forum, lure their target with fake collaboration opportunities (typically a demo for anti-virus software, VPN, music players, photo editing or online games), hijack their channel, then either sell it to the highest bidder or use it to broadcast cryptocurrency scams.

In collaboration with YouTube, Gmail, Trust & Safety, CyberCrime Investigation Group and Safe Browsing teams, our protections have decreased the volume of related phishing emails on Gmail by 99.6% since May 2021. We blocked 1.6M messages to targets, displayed ~62K Safe Browsing phishing page warnings, blocked 2.4K files, and successfully restored ~4K accounts. With increased detection efforts, we’ve observed attackers shifting away from Gmail to other email providers (mostly email.cz, seznam.cz, post.cz and aol.com). Moreover, to protect our users, we have referred the below activity to the FBI for further investigation.

In this blog, we share examples of the specific tactics, techniques and procedures (TTPs) used to lure victims, as well as some guidance on how users can further protect themselves.

Tactics, techniques and procedures

Cookie Theft, also known as “pass-the-cookie attack,” is a session hijacking technique that enables access to user accounts with session cookies stored in the browser. While the technique has been around for decades, its resurgence as a top security risk could be due to a wider adoption of multi-factor authentication (MFA) making it difficult to conduct abuse, and shifting attacker focus to social engineering tactics.

Social engineering YouTubers with advertisement offer

Many YouTube creators provide an email address on their channel for business opportunities. In this case, the attackers sent forged business emails impersonating an existing company requesting a video advertisement collaboration.

The phishing typically started with a customized email introducing the company and its products. Once the target agreed to the deal, a malware landing page disguised as a software download URL was sent via email or a PDF on Google Drive, and in a few cases, Google documents containing the phishing links. Around 15,000 actor accounts were identified, most of which were created for this campaign specifically.

Fake software landing pages and social media accounts

The attackers registered various domains associated with forged companies and built multiple websites for malware delivery. To date, we’ve identified at least 1,011 domains created solely for this purpose. Some of the websites impersonated legitimate software sites, such as Luminar, Cisco VPN, games on Steam, and some were generated using online templates. During the pandemic, we also uncovered attackers posing as news providers with a “Covid19 news software.”

In one case, we observed a fake social media page copying content from an existing software company. The following screenshot is an example of a fake page where the original URL is replaced with one leading to a cookie theft malware download.

Because Google actively detects and disrupts phishing links sent via Gmail, the actors were observed driving targets to messaging apps like WhatsApp, Telegram or Discord.

Delivering cookie theft malware

Once the target runs the fake software, a cookie stealing malware executes, taking browser cookies from the victim’s machine and uploading them to the actor's command & control servers. Although this type of malware can be configured to be persistent on the victim's machine, these actors are running all malware in non-persistent mode as a smash-and-grab technique. This is because if the malicious file is not detected when executed, there are less artifacts on an infected host and therefore security products fail to notify the user of a past compromise.

We have observed that actors use various types of malware based on personal preference, most of which are easily available on Github. Some commodity malware used included RedLine, Vidar, Predator The Thief, Nexus stealer, Azorult, Raccoon, Grand Stealer, Vikro Stealer, Masad (Google’s naming), and Kantal (Google’s naming) which shares code similarity with Vidar. Open source malware like Sorano and AdamantiumThief were also observed.Related hashes are listed in the Technical Details section, at the end of this report.

Most of the observed malware was capable of stealing both user passwords and cookies. Some of the samples employed several anti-sandboxing techniques including enlarged files, encrypted archive and download IP cloaking. A few were observed displaying a fake error message requiring user click-through to continue execution.

Cryptocurrency scams and channel selling

A large number of hijacked channels were rebranded for cryptocurrency scam live-streaming. The channel name, profile picture and content were all replaced with cryptocurrency branding to impersonate large tech or cryptocurrency exchange firms. The attacker live-streamed videos promising cryptocurrency giveaways in exchange for an initial contribution.

On account-trading markets, hijacked channels ranged from $3 USD to $4,000 USD depending on the number of subscribers.

Hack-for-Hire attackers

These campaigns were carried out by a number of hack-for-hire actors recruited on Russian-speaking forums via the following job description, offering two types of work:

This recruitment model explains the highly customized social engineering, as well as the varied malware types given each actor's choice of preferred malware.

Protecting our users from attacks

We are continuously improving our detection methods and investing in new tools and features that automatically identify and stop threats like this one. Some of these improvements include:

• Additional heuristic rules to detect and block phishing & social engineering emails, cookie theft hijacking and crypto-scam livestreams.
• Safe Browsing is further detecting and blocking malware landing pages and downloads.
• YouTube has hardened channel transfer workflows, detected and auto-recovered over 99% of hijacked channels.
• Account Security has hardened authentication workflows to block and notify the user on potential sensitive actions.

It is also important that users remain aware of these types of threats and take appropriate action to further protect themselves. Our recommendations:

• Take Safe Browsing warnings seriously. To avoid malware triggering antivirus detections, threat actors social engineer users into turning off or ignoring warnings.
• Before running software, perform virus scanning using an antivirus or online virus scanning tool like VirusTotal to verify file legitimacy.
• Enable the “Enhanced Safe Browsing Protection” mode in your Chrome browser, a feature that increases warnings on potentially suspicious web pages & files.
• Be aware of encrypted archives which are often bypassing antivirus detection scans, increasing the risk of running malicious files.
• Protect your account with 2-Step-verification (multi-factor authentication) which provides an extra layer of security to your account in case your password is stolen. Starting November 1, monetizing YouTube creators must turn on 2-Step Verification on the Google Account used for their YouTube channel to access YouTube Studio or YouTube Studio Content Manager.

Additional resources: Avoid & Report Phishing Emails.

Technical Details

Related Malware hashes:

• RedLine (commodity)
  • c8b42437ffd8cfbbe568013eaaa707c212a2628232c01d809a3cf864fe24afa8
  • 501fe2509581d43288664f0d2825a6a47102cd614f676bf39f0f80ab2fd43f2c
  • c8b42437ffd8cfbbe568013eaaa707c212a2628232c01d809a3cf864fe24afa8
• Vidar (commodity)
  • 9afc029ac5aa525e6fdcedf1e93a64980751eeeae3cf073fcbd1d223ab5c96d6
• Kantal (share code similarity with Vidar)
  • F59534e6d9e0559d99d2b3a630672a514dbd105b0d6fc9447d573ebd0053caba (zip archive)
  • Edea528804e505d202351eda0c186d7c200c854c41049d7b06d1971591142358 (unpacked sample)
• Predator The Thief (commodity)
  • 0d8cfa02515d504ca34273d8cfbe9d1d0f223e5d2cece00533c48a990fd8ce72 (zip archive)
• Sorano (open source)
  • c7c8466a66187f78d953c64cbbd2be916328085aa3c5e48fde6767bc9890516b
• Nexus stealer (commodity)
  • ed8b2af133b4144bef2b89dbec1526bf80cc06fe053ece1fa873f6bd1e99f0be
  • efc88a933a8baa6e7521c8d0cf78c52b0e3feb22985de3d35316a8b00c5073b3
• Azorult (commodity)
  • 8cafd480ac2a6018a4e716a4f9fd1254c4e93501a84ee1731ed7b98b67ab15dd
• Raccoon (commodity)
  • 85066962ba1e8a0a8d6989fffe38ff564a6cf6f8a07782b3fbc0dcb19d2497cb
• Grand Stealer (commodity)
  • 6359d5fa7437164b300abc69c8366f9481cb91b7558d68c9e3b0c2a535ddc243
• Vikro Stealer (commodity)
  • 04deb8d8aee87b24c7ba0db55610bb12f7d8ec1e75765650e5b2b4f933b18f6d
• Masad (commodity)
  • 6235573d8d178341dbfbead7c18a2f419808dc8c7c302ac61e4f9645d024ed85
• AdamantiumThief (open source)
  • Db45bb99c44a96118bc5673a7ad65dc2a451ea70d4066715006107f65d906715

Top Phishing Domains:

• pro-swapper[.]com
• downloadnature[.]space
• downloadnature[.]com
• fast-redirect[.]host
• bragi-studio[.]com
• plplme[.]site
• fenzor[.]com
• universe-photo[.]com
• rainway-gaming[.]com
• awaken1337[.]xyz
• pixelka[.]fun
• vortex-cloudgaming[.]com
• vontex[.]tech
• user52406.majorcore[.]space
• voneditor[.]tech
• spaceditor[.]space
• roudar[.]com
• peoplep[.]site
• anypon[.]online
• zeneditor[.]tech
• yourworld[.]site
• playerupbo[.]xyz
• dizzify[.]me

Google’s Threat Analysis Group tracks actors involved in disinformation campaigns, government backed hacking, and financially motivated abuse. We have a long-standing policy to send you a warning if we detect that your account is a target of government-backed phishing or malware attempts. So far in 2021, we’ve sent over 50,000 warnings, a nearly 33% increase from this time in 2020. This spike is largely due to blocking an unusually large campaign from a Russian actor known as APT28 or Fancy Bear.

We intentionally send these warnings in batches to all users who may be at risk, rather than at the moment we detect the threat itself, so that attackers cannot track our defense strategies. On any given day, TAG is tracking more than 270 targeted or government-backed attacker groups from more than 50 countries. This means that there is typically more than one threat actor behind the warnings.

In this blog, we explore some of the most notable campaigns we’ve disrupted this year from a different government-backed attacker: APT35, an Iranian group, which regularly conducts phishing campaigns targeting high risk users. This is the one of the groups we disrupted during the 2020 US election cycle for its targeting of campaign staffers. For years, this group has hijacked accounts, deployed malware, and used novel techniques to conduct espionage aligned with the interests of the Iranian government.

Hijacked websites used for credential phishing attacks

In early 2021, APT35 compromised a website affiliated with a UK university to host a phishing kit. Attackers sent email messages with links to this website to harvest credentials for platforms such as Gmail, Hotmail, and Yahoo. Users were instructed to activate an invitation to a (fake) webinar by logging in. The phishing kit will also ask for second-factor authentication codes sent to devices.

APT35 has relied on this technique since 2017 — targeting high-value accounts in government, academia, journalism, NGOs, foreign policy, and national security. Credential phishing through a compromised website demonstrates these attackers will go to great lengths to appear legitimate – as they know it's difficult for users to detect this kind of attack.

Utilization of Spyware Apps

In May 2020, we discovered that APT35 attempted to upload spyware to the Google Play Store. The app was disguised as VPN software that, if installed, could steal sensitive information such as call logs, text messages, contacts, and location data from devices. Google detected the app quickly and removed it from the Play Store before any users had a chance to install it. Although Play Store users were protected, we are highlighting the app here as TAG has seen APT35 attempt to distribute this spyware on other platforms as recently as July 2021.

Conference-themed phishing emails

One of the most notable characteristics of APT35 is their impersonation of conference officials to conduct phishing attacks. Attackers used the Munich Security and the Think-20 (T20) Italy conferences as lures in non-malicious first contact email messages to get users to respond. When they did, attackers sent them phishing links in follow-on correspondence.

Targets typically had to navigate through at least one redirect before landing on a phishing domain. Link shorteners and click trackers are heavily used for this purpose, and are oftentimes embedded within PDF files. We’ve disrupted attacks using Google Drive, App Scripts, and Sites pages in these campaigns as APT35 tries to get around our defenses. Services from Dropbox and Microsoft are also abused.

Telegram for threat actor notifications

One of APT35’s novel techniques involves using Telegram for operator notifications. The attackers embed javascript into phishing pages that notify them when the page has been loaded. To send the notification, they use the Telegram API sendMessage function, which lets anyone use a Telegram bot to send a message to a public channel. The attackers use this function to relay device-based data to the channel, so they can see details such as the IP, useragent, and locales of visitors to their phishing sites in real-time. We reported the bot to Telegram and they have taken action to remove it.

How we keep users safe from these threats

We warn users when we suspect a government-backed threat like APT35 is targeting them. Thousands of these warnings are sent every month, even in cases where the corresponding attack is blocked. If you receive a warning it does not mean your account has been compromised, it means you have been identified as a target.

Workspace administrators are also notified regarding targeted accounts in their domain. Users are encouraged to take these warnings seriously and consider enrolling in the Advanced Protection Program or enabling two-factor authentication if they haven't already.

We also block malicious domains using Google Safe Browsing – a service that Google's security team built to identify unsafe websites across the web and notify users and website owners of potential harm. When a user of a Safe Browsing-enabled browser or app attempts to access unsafe content on the web, they’ll see a warning page explaining that the content they’re trying to access may be harmful. When a site identified by Safe Browsing as harmful appears in Google Search results, we show a warning next to it in the results.

Threat Analysis Group will continue to identify bad actors and share relevant information with others in the industry, with the goal of bringing awareness to these issues, protecting you and fighting bad actors to prevent future attacks.

Technical Details

Indicators from APT28 phishing campaign:

service-reset-password-moderate-digital.rf[.]gd

reset-service-identity-mail.42web[.]io

digital-email-software.great-site[.]net

Indicators from APT35 campaigns:

Abused Google Properties:

https://sites.google[.]com/view/ty85yt8tg8-download-rtih4ithr/

https://sites.google[.]com/view/user-id-568245/

https://sites.google[.]com/view/hhbejfdwdhwuhscbsb-xscvhdvbc/

Abused Dropbox Properties:

https://www.dropbox[.]com/s/68y4vpfu8pc3imf/Iraq&Jewish.pdf

Phishing Domains:

nco2[.]live

summit-files[.]com

filetransfer[.]club

continuetogo[.]me

accessverification[.]online

customers-verification-identifier[.]site

service-activity-session[.]online

identifier-service-review[.]site

recovery-activity-identification[.]site

review-session-confirmation[.]site

recovery-service-activity[.]site

verify-service-activity[.]site

service-manager-notifications[.]info

Android App:

https://www.virustotal.com/gui/file/5d3ff202f20af915863eee45916412a271bae1ea3a0e20988309c16723ce4da5/detection

Android App C2:

communication-shield[.]site

cdsa[.]xyz

Introduction

Google’s Threat Analysis Group tracks actors involved in disinformation campaigns, government backed hacking, and financially motivated abuse. Understanding the techniques used by attackers helps us counter these threats effectively. This blog post is intended to highlight a new evasion technique we identified, which is currently being used by a financially motivated threat actor to avoid detection.

Attackers often rely on varying behaviors between different systems to gain access. For instance, attacker’s may bypass filtering by convincing a mail gateway that a document is benign so the computer treats it as an executable program. In the case of the attack outlined below, we see that attackers created malformed code signatures that are treated as valid by Windows but are not able to be decoded or checked by OpenSSL code — which is used in a number of security scanning products. We believe this is a technique the attacker is using to evade detection rules.

Technical Details

Code signatures on Windows executables provide guarantees about the integrity of a signed executable, as well as information about the identity of the signer. Attackers who are able to obscure their identity in signatures without affecting the integrity of the signature can avoid detection longer and extend the lifetime of their code-signing certificates to infect more systems.

OpenSUpdater, a known family of unwanted software which violates our policies and is harmful to the user experience, is used to download and install other suspicious programs.The actor behind OpenSUpdater tries to infect as many users as possible and while they do not have specific targeting, most targets appear to be within the United States and prone to downloading game cracks and grey-area software.

Groups of OpenSUpdater samples are often signed with the same code-signing certificate, obtained from a legitimate certificate authority. Since mid-August, OpenSUpdater samples have carried an invalid signature, and further investigation showed this was a deliberate attempt to evade detection. In these new samples, the signature was edited such that an End of Content (EOC) marker replaced a NULL tag for the 'parameters' element of the SignatureAlgorithm signing the leaf X.509 certificate.

EOC markers terminate indefinite-length encodings, but in this case an EOC is used within a definite-length encoding (l= 13). 

Bytes: 30 0D 06 09 2A 86 48 86  F7 0D 01 01 0B 00 00 

Decodes to the following elements:

SEQUENCE (2 elem)

OBJECT IDENTIFIER 1.2.840.113549.1.1.11 sha256WithRSAEncryption (PKCS #1)

EOC

Security products using OpenSSL to extract signature information will reject this encoding as invalid. However, to a parser that permits these encodings, the digital signature of the binary will otherwise appear legitimate and valid. This is the first time TAG has observed actors using this technique to evade detection while preserving a valid digital signature on PE files. 

As shown in the following screenshot, the signature is considered to be valid by the Windows operating system. This issue has been reported to Microsoft.

Since first discovering this activity, OpenSUpdater's authors have tried other variations on invalid encodings to further evade detection.

The following are samples using this evasion:

https://www.virustotal.com/gui/file/5094028a0afb4d4a3d8fa82b613c0e59d31450d6c75ed96ded02be1e9db8104f/detection

New variant:

https://www.virustotal.com/gui/file/5c0ff7b23457078c9d0cbe186f1d05bfd573eb555baa1bf4a45e1b79c8c575db/detection

Our team is working in collaboration with Google Safe Browsing to protect users from downloading and executing this family of unwanted software. Users are encouraged to only download and install software from reputable and trustworthy sources.

This bulletin includes coordinated influence operation campaigns terminated on our platforms in Q3 2021. It was last updated on August 31, 2021.

July 

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Ukraine. This campaign uploaded content in Ukrainian and Russian that was supportive of Russia’s government and critical of the Ukrainian military. We received leads from FireEye that supported us in this investigation.

• We blocked 10 domains from eligibility to appear on Google News surfaces and Discover as part of our investigation into coordinated influence operations linked to Russia. This campaign uploaded content in Russian that was critical of Ukraine’s government and supportive of Russia.

• We terminated 2 YouTube channels as part of our investigation into coordinated influence operations linked to Iraq. This campaign uploaded content in Arabic that was supportive of Iran-backed militias and critical of the U.S. and its allies. Our findings are similar to findings reported by Facebook.

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Jordan. This campaign uploaded content in Arabic that was supportive of the Jordanian government and critical of its opposition. Our findings are similar to findings reported by Facebook.

• We terminated 15 YouTube channels as part of our investigation into coordinated influence operations linked to Algeria. This campaign uploaded content in Arabic that was supportive of the Algerian government and its military. Our findings are similar to findings reported by Facebook. We received leads from Graphika that supported us in this investigation.

• We terminated 6 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was critical of certain local politicians in Campeche, Mexico. Our findings are similar to findings reported by Facebook.

• We terminated 4 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was supportive of a member of the National Action Party). Our findings are similar to findings reported by Facebook.

• We terminated 16 YouTube channels and 1 ads account as part of our investigation into coordinated influence operations linked to Sudan. This campaign uploaded content in Arabic that was supportive of the Muslim Brotherhood and critical of the current Sudanese government. Our findings are similar to findings reported by Facebook.

• We terminated 850 YouTube channels as part of our ongoing investigation into coordinated influence operations linked to China. These channels mostly uploaded spammy content in Chinese about music, entertainment, and lifestyle. A very small subset uploaded content in Chinese and English about China’s COVID-19 vaccine efforts and social issues in the U.S. These findings are consistent with our previous reports.

This bulletin includes coordinated influence operation campaigns terminated on our platforms in Q3 2021. It was last updated on August 31, 2021.

July 

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Ukraine. This campaign uploaded content in Ukrainian and Russian that was supportive of Russia’s government and critical of the Ukrainian military. We received leads from FireEye that supported us in this investigation.

• We blocked 10 domains from eligibility to appear on Google News surfaces and Discover as part of our investigation into coordinated influence operations linked to Russia. This campaign uploaded content in Russian that was critical of Ukraine’s government and supportive of Russia.

• We terminated 2 YouTube channels as part of our investigation into coordinated influence operations linked to Iraq. This campaign uploaded content in Arabic that was supportive of Iran-backed militias and critical of the U.S. and its allies. Our findings are similar to findings reported by Facebook.

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Jordan. This campaign uploaded content in Arabic that was supportive of the Jordanian government and critical of its opposition. Our findings are similar to findings reported by Facebook.

• We terminated 15 YouTube channels as part of our investigation into coordinated influence operations linked to Algeria. This campaign uploaded content in Arabic that was supportive of the Algerian government and its military. Our findings are similar to findings reported by Facebook. We received leads from Graphika that supported us in this investigation.

• We terminated 6 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was critical of certain local politicians in Campeche, Mexico. Our findings are similar to findings reported by Facebook.

• We terminated 4 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was supportive of a member of the National Action Party). Our findings are similar to findings reported by Facebook.

• We terminated 16 YouTube channels and 1 ads account as part of our investigation into coordinated influence operations linked to Sudan. This campaign uploaded content in Arabic that was supportive of the Muslim Brotherhood and critical of the current Sudanese government. Our findings are similar to findings reported by Facebook.

• We terminated 850 YouTube channels as part of our ongoing investigation into coordinated influence operations linked to China. These channels mostly uploaded spammy content in Chinese about music, entertainment, and lifestyle. A very small subset uploaded content in Chinese and English about China’s COVID-19 vaccine efforts and social issues in the U.S. These findings are consistent with our previous reports.

This bulletin includes coordinated influence operation campaigns terminated on our platforms in Q3 2021. It was last updated on August 31, 2021.

July 

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Ukraine. This campaign uploaded content in Ukrainian and Russian that was supportive of Russia’s government and critical of the Ukrainian military. We received leads from FireEye that supported us in this investigation.

• We blocked 10 domains from eligibility to appear on Google News surfaces and Discover as part of our investigation into coordinated influence operations linked to Russia. This campaign uploaded content in Russian that was critical of Ukraine’s government and supportive of Russia.

• We terminated 2 YouTube channels as part of our investigation into coordinated influence operations linked to Iraq. This campaign uploaded content in Arabic that was supportive of Iran-backed militias and critical of the U.S. and its allies. Our findings are similar to findings reported by Facebook.

• We terminated 7 YouTube channels as part of our investigation into coordinated influence operations linked to Jordan. This campaign uploaded content in Arabic that was supportive of the Jordanian government and critical of its opposition. Our findings are similar to findings reported by Facebook.

• We terminated 15 YouTube channels as part of our investigation into coordinated influence operations linked to Algeria. This campaign uploaded content in Arabic that was supportive of the Algerian government and its military. Our findings are similar to findings reported by Facebook. We received leads from Graphika that supported us in this investigation.

• We terminated 6 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was critical of certain local politicians in Campeche, Mexico. Our findings are similar to findings reported by Facebook.

• We terminated 4 YouTube channels as part of our investigation into coordinated influence operations linked to Mexico. This campaign uploaded content in Spanish that was supportive of a member of the National Action Party). Our findings are similar to findings reported by Facebook.

• We terminated 16 YouTube channels and 1 ads account as part of our investigation into coordinated influence operations linked to Sudan. This campaign uploaded content in Arabic that was supportive of the Muslim Brotherhood and critical of the current Sudanese government. Our findings are similar to findings reported by Facebook.

• We terminated 850 YouTube channels as part of our ongoing investigation into coordinated influence operations linked to China. These channels mostly uploaded spammy content in Chinese about music, entertainment, and lifestyle. A very small subset uploaded content in Chinese and English about China’s COVID-19 vaccine efforts and social issues in the U.S. These findings are consistent with our previous reports.

Zero-day vulnerabilities are unknown software flaws. Until they’re identified and fixed, they can be exploited by attackers. Google’s Threat Analysis Group (TAG) actively works to detect hacking attempts and influence operations to protect users from digital attacks, this includes hunting for these types of vulnerabilities because they can be particularly dangerous when exploited and have a high rate of success.

In this blog, we’re sharing details about four in-the-wild 0-day campaigns targeting four separate vulnerabilities we’ve discovered so far this year: 

• CVE-2021-21166 and CVE-2021-30551 in Chrome,

• CVE-2021-33742 in Internet Explorer, and

• CVE-2021-1879 in WebKit (Safari).

The four exploits were used as a part of three different campaigns. As is our policy, after discovering these 0-days, we quickly reported to the vendor and patches were released to users to protect them from these attacks. We assess three of these exploits were developed by the same commercial surveillance company that sold these capabilities to two different government-backed actors. Google has also published root cause analyses (RCAs) on each of the 0-days.

In addition to the technical details, we’ll also provide our take on the large uptick of in-the-wild 0-day attacks the industry is seeing this year. Halfway into 2021, there have been 33 0-day exploits used in attacks that have been publicly disclosed this year — 11 more than the total number from 2020. While there is an increase in the number of 0-day exploits being used, we believe greater detection and disclosure efforts are also contributing to the upward trend.

Chrome: CVE-2021-21166 and CVE-2021-30551

Over the past several months, we have discovered two Chrome renderer remote code execution 0-day exploits, CVE-2021-21166 and ​​CVE-2021-30551, which we believe to be used by the same actor. CVE-2021-21166 was discovered in February 2021 while running Chrome 88.0.4323.182 and CVE-2021-30551 was discovered in June 2021 while running Chrome 91.0.4472.77.

Both of these 0-days were delivered as one-time links sent by email to the targets, all of whom we believe were in Armenia. The links led to attacker-controlled domains that mimicked legitimate websites related to the targeted users. When a target clicked the link, they were redirected to a webpage that would fingerprint their device, collect system information about the client and generate ECDH keys to encrypt the exploits, and then send this data back to the exploit server. The information collected from the fingerprinting phase included screen resolution, timezone, languages, browser plugins, and available MIME types. This information was collected by the attackers to decide whether or not an exploit should be delivered to the target. Using appropriate configurations, we were able to recover two 0-day exploits (CVE-2021-21166 & CVE-2021-30551), which were targeting the latest versions of Chrome on Windows at the time of delivery.

After the renderer is compromised, an intermediary stage is executed to gather more information about the infected device including OS build version, CPU, firmware and BIOS information. This is likely collected in an attempt to detect virtual machines and deliver a tailored sandbox escape to the target. In our environment, we did not receive any payloads past this stage.

While analyzing CVE-2021-21166 we realized the vulnerability was also in code shared with WebKit and therefore Safari was also vulnerable. Apple fixed the issue as CVE-2021-1844. We do not have any evidence that this vulnerability was used to target Safari users.

Related IOCs

• lragir[.]org

• armradio[.]org

• asbares[.]com

• armtimes[.]net

• armlur[.]org

• armenpress[.]org

• hraparak[.]org

• armtimes[.]org

• hetq[.]org

Internet Explorer: CVE-2021-33742

Despite Microsoft announcing the retirement of Internet Explorer 11, planned for June 2022, attackers continue to develop creative ways to load malicious content inside Internet Explorer engines to exploit vulnerabilities. For example, earlier this year, North Korean attackers distributed MHT files embedding an exploit for CVE-2021-26411. These files are automatically opened in Internet Explorer when they are double clicked by the user.

In April 2021, TAG discovered a campaign targeting Armenian users with malicious Office documents that loaded web content within Internet Explorer. This happened by either embedding a remote ActiveX object using a Shell.Explorer.1 OLE object or by spawning an Internet Explorer process via VBA macros to navigate to a web page. At the time, we were unable to recover the next stage payload, but successfully recovered the exploit after an early June campaign from the same actors. After a fingerprinting phase, similar to the one used with the Chrome exploit above, users were served an Internet Explorer 0-day. This vulnerability was assigned CVE-2021-33742 and fixed by Microsoft in June 2021.

The exploit loaded an intermediary stage similar to the one used in the Chrome exploits. We did not recover additional payloads in our environment.

During our investigation we discovered several documents uploaded to VirusTotal.

Based on our analysis, we assess that the Chrome and Internet Explorer exploits described here were developed and sold by the same vendor providing surveillance capabilities to customers around the world.

Related IOCs

Examples of related Office documents uploaded to VirusTotal:

• https://www.virustotal.com/gui/file/656d19186795280a068fcb97e7ef821b55ad3d620771d42ed98d22ee3c635e67/detection

• https://www.virustotal.com/gui/file/851bf4ab807fc9b29c9f6468c8c89a82b8f94e40474c6669f105bce91f278fdb/detection

Unique URLs serving ​​CVE-2021-33742 Internet Explorer exploit:

• http://lioiamcount[.]com/IsnoMLgankYg6/EjlYIy7cdFZFeyFqE4IURS1

• http://db-control-uplink[.]com/eFe1J00hISDe9Zw/gzHvIOlHpIXB

• http://kidone[.]xyz/VvE0yYArmvhyTl/GzV

Word documents with the following classid:

• {EAB22AC3-30C1-11CF-A7EB-0000C05BAE0B}

Related infrastructure:

• workaj[.]com

• wordzmncount[.]com

WebKit (Safari): CVE-​2021-1879

Not all attacks require chaining multiple 0-day exploits to be successful. A recent example is CVE-​2021-1879 that was discovered by TAG on March 19, 2021, and used by a likely Russian government-backed actor. (NOTE: This exploit is not connected to the other three we’ve discussed above.)

In this campaign, attackers used LinkedIn Messaging to target government officials from western European countries by sending them malicious links. If the target visited the link from an iOS device, they would be redirected to an attacker-controlled domain that served the next stage payloads. The campaign targeting iOS devices coincided with campaigns from the same actor targeting users on Windows devices to deliver Cobalt Strike, one of which was previously described by Volexity.

After several validation checks to ensure the device being exploited was a real device, the final payload would be served to exploit CVE-​2021-1879. This exploit would turn off Same-Origin-Policy protections in order to collect authentication cookies from several popular websites, including Google, Microsoft, LinkedIn, Facebook and Yahoo and send them via WebSocket to an attacker-controlled IP. The victim would need to have a session open on these websites from Safari for cookies to be successfully exfiltrated. There was no sandbox escape or implant delivered via this exploit. The exploit targeted iOS versions 12.4 through 13.7. This type of attack, described by Amy Burnett in Forget the Sandbox Escape: Abusing Browsers from Code Execution, are mitigated in browsers with Site Isolation enabled such as Chrome or Firefox. 

Related IOCs

• supportcdn.web[.]app

• vegmobile[.]com

• 111.90.146[.]198

Why So Many 0-days?

There is not a one-to-one relationship between the number of 0-days being used in-the-wild and the number of 0-days being detected and disclosed as in-the-wild. The attackers behind 0-day exploits generally want their 0-days to stay hidden and unknown because that’s how they’re most useful. 

Based on this, there are multiple factors that could be contributing to the uptick in the number of 0-days that are disclosed as in-the-wild:

Increase in detection & disclosure

This year, Apple began annotating vulnerabilities in their security bulletins to include notes if there is reason to believe that a vulnerability may be exploited in-the-wild and Google added these annotations to their Android bulletins. When vendors don’t include these annotations, the only way the public can learn of the in-the-wild exploitation is if the researcher or group who knows of the exploitation publishes the information themselves. 

In addition to beginning to disclose when 0-days are believed to be exploited in-the-wild, it wouldn’t be surprising if there are more 0-day detection efforts, and successes, occurring as a result. It’s also possible that more people are focusing on discovering 0-days in-the-wild and/or reporting the 0-days that they found in the wild.

Increased Utilization

There is also the possibility that attackers are using more 0-day exploits. There are a few reasons why this is likely:

• The increase and maturation of security technologies and features mean that the same capability requires more 0-day vulnerabilities for the functional chains. For example, as the Android application sandbox has been further locked down by limiting what syscalls an application can call, an additional 0-day is necessary to escape the sandbox. 
• The growth of mobile platforms has resulted in an increase in the number of products that actors want capabilities for. 
• There are more commercial vendors selling access to 0-days than in the early 2010s.
• Maturing of security postures increases the need for attackers to use 0-day exploits rather than other less sophisticated means, such as convincing people to install malware. Due to advancements in security, these actors now more often have to use 0-day exploits to accomplish their goals. 

Conclusion

Over the last decade, we believe there has been an increase in attackers using 0-day exploits. Attackers needing more 0-day exploits to maintain their capabilities is a good thing — and it  reflects increased cost to the attackers from security measures that close known vulnerabilities. However, the increasing demand for these capabilities and the ecosystem that supplies them is more of a challenge. 0-day capabilities used to be only the tools of select nation states who had the technical expertise to find 0-day vulnerabilities, develop them into exploits, and then strategically operationalize their use. In the mid-to-late 2010s, more private companies have joined the marketplace selling these 0-day capabilities. No longer do groups need to have the technical expertise, now they just need resources. Three of the four 0-days that TAG has discovered in 2021 fall into this category: developed by commercial providers and sold to and used by government-backed actors.

Meanwhile, improvements in detection and a growing culture of disclosure likely contribute to the significant uptick in 0-days detected in 2021 compared to 2020, but reflect more positive trends. Those of us working on protecting users from 0-day attacks have long suspected that overall, the industry detects only a small percentage of the 0-days actually being used. Increasing our detection of 0-day exploits is a good thing — it allows us to get those vulnerabilities fixed and protect users, and gives us a fuller picture of the exploitation that is actually happening so we can make more informed decisions on how to prevent and fight it.

We’d be remiss if we did not acknowledge the quick response and patching of these vulnerabilities by the Apple, Google, and Microsoft teams. 

This bulletin includes coordinated influence operation campaigns terminated on our platforms in Q2 2021. It was last updated on May 26, 2021.

April

• We terminated 3 YouTube channels as part of our investigation into coordinated influence operations linked to El Salvador. This campaign uploaded content in Spanish focusing on a mayoral race in the Santa Tecla municipality. Our findings are similar to findings reported by Facebook.

• We terminated 43 YouTube channels as part of our investigation into coordinated influence operations linked to Albania. This campaign uploaded content in Farsi that was critical of Iran’s government and supportive of Mojahedin-e Khalq. Our findings are similar to findings reported by Facebook.

• We terminated 728 YouTube channels as part of our ongoing investigation into coordinated influence operations linked to China. These channels mostly uploaded spammy content in Chinese about music, entertainment, and lifestyle. A very small subset uploaded content in Chinese and English about protests in Hong Kong and criticism of the U.S. response to the COVID-19 pandemic. These findings are consistent with our previous reports.

In January, the Threat Analysis Group documented a hacking campaign, which we were able to attribute to a North Korean government-backed entity, targeting security researchers. On March 17th, the same actors behind those attacks set up a new website with associated social media profiles for a fake company called “SecuriElite.”

The new website claims the company is an offensive security company located in Turkey that offers pentests, software security assessments and exploits. Like previous websites we’ve seen set up by this actor, this website has a link to their PGP public key at the bottom of the page. In January, targeted researchers reported that the PGP key hosted on the attacker’s blog acted as the lure to visit the site where a browser exploit was waiting to be triggered.

The attacker’s latest batch of social media profiles continue the trend of posing as fellow security researchers interested in exploitation and offensive security. On LinkedIn, we identified two accounts impersonating recruiters for antivirus and security companies. We have reported all identified social media profiles to the platforms to allow them to take appropriate action. 

At this time, we have not observed the new attacker website serve malicious content, but we have added it to Google Safebrowsing as a precaution.

Following our January blog post, security researchers successfully identified these actors using an Internet Explorer 0-day. Based on their activity, we continue to believe that these actors are dangerous, and likely have more 0-days. We encourage anyone who discovers a Chrome vulnerability to report that activity through the Chrome Vulnerabilities Rewards Program submission process.

Actor controlled sites and accounts

Fake Security Company Website:

• www.securielite[.]com
  

• https://twitter.com/alexjoe9983
• https://twitter.com/BenH3mmings
• https://twitter.com/chape2002
• https://twitter.com/julia0235
• https://twitter.com/lookworld0821
• https://twitter.com/osm4nd
• https://twitter.com/seb_lazar
• https://twitter.com/securielite

LinkedIn Profiles:

• SecuriElite - https://www.linkedin.com/company/securielite/
• Carter Edwards, HR Director @ Trend Macro - https://www.linkedin.com/in/carter-edwards-a99138204/
• Colton Perry, Security Researcher - https://www.linkedin.com/in/colton-perry-6a8059204/
• Evely Burton, Technical Recruiter @ Malwarebytes - https://www.linkedin.com/in/evely-burton-204b29207/
• Osman Demir, CEO @ SecuriElite - https://www.linkedin.com/in/osman-demir-307520209/
• Piper Webster, Security Researcher - https://www.linkedin.com/in/piper-webster-192676203/
• Sebastian Lazarescue, Security Researcher @ SecuriElite - https://www.linkedin.com/in/sebastian-lazarescue-456840209/

Email:

• [email protected]
• [email protected]
• [email protected]

Attacker Owned Domains:

• bestwing[.]org
• codebiogblog[.]com
• coldpacific[.]com
• cutesaucepuppy[.]com
• devguardmap[.]org
• hireproplus[.]com
• hotelboard[.]org
• mediterraneanroom[.]org
• redeastbay[.]com
• regclassboard[.]com
• securielite[.]com
• spotchannel02[.]com
• wileprefgurad[.]net