We have started posting sample preprocessing analyses of threat reports from our Report Hub, showcasing results from one of the first stages of our multi-stage engine. If you have any suggestions for tweet format improvement, please send us a message
#threatreport#MediumCompleteness
Beyond Optimizely: Malicious Scripts Traced Across Four Web Services | 17-11-2026
Source: infostealers.com/article/bey…
Key details below ↓
🧑💻Actors/Campaigns:
Optimjack
💀Threats:
Coruna_tool, Darksword_tool,
🎯Victims: Media, Financial services, Information technology, Digital marketing, Web analytics
🏭Industry: Financial, Aerospace
🌐Geo: Usa
📚TTPs:
⚔️Tactics: 1
🛠️Technics: 0
🤖LLM extracted TTPs:`
T1005, T1027.009, T1059.007, T1105, T1189, T1190, T1195.002, T1552.004, T1634.001
🧨IOCs:
- Url: 8
- File: 4
- Hash: 8
- Domain: 7
💽Software: Zoom, Salesforce, Chrome
🔢Algorithms: sha256
📜Programming Languages: javascript
💻Platforms: intel, apple
#threatreport:
Optimizely confirmed that an unauthorized party exploited a vulnerability in its Experimentation account signup flow and injected malicious JavaScript into customer projects. Because the code was served through legitimate Optimizely snippets, websites loading those snippets could execute attacker-controlled code in visitors’ browsers. Report URI identified 188 poisoned snippets associated with 178 organizations. Hudson Rock directly recovered eight malicious Optimizely assets; its investigation also documented poisoned code delivered through Kissmetrics, StorifyMe and minute-ly. November 13 records show earlier observations, but do not establish continuous malicious delivery before the larger November 16 activity.
The malicious code included a financial-theft payload. In later BBC and HSBC UAE Optimizely files, researchers found the same 24,392-byte payload embedded as a fallback. The loader attempted to execute it if a remote payload failed or did not arrive within roughly 2.5 seconds, subject to browser execution policies. Consequently, blocking the external malware host alone would not remove all risk: remediation needed to remove the poisoned project code and account for copies already delivered to browsers.
The reporting also describes mobile-focused stages associated with Coruna and DarkSword. Ohey researcher Kirk reported retrieving later DarkSword stages, including `post_exploit.js`, which describes device registration and uploading databases such as `keychain-2.db`; other reported panel features target seed phrases and private keys. These findings describe intended collection behavior, not confirmed victim outcomes. Hudson Rock did not independently reproduce the later stages or demonstrate a successful compromise of a real iPhone. An iPhone user-agent string in a scan is not evidence of Safari exploitation.
The article distinguishes exposure to compromised third-party scripts from compromise of a website’s backend. It reports company responses and does not assert a campaign-wide victim count or successful real-device exploitation.
#threatreport#HighCompleteness
Exposing tcherber: VPN Access Brokerage, Custom Malware, and an Exposed C2 Panel | 17-11-2026
Source: cyderes.com/howler-cell/expo…
Key details below ↓
🧑💻Actors/Campaigns:
Tcherber (🧠motivation: cyber_criminal, information_theft)
💀Threats:
Uac_bypass_technique, Hvnc_tool, Enigma_ransomware, Zer0day_tool, Anydesk_tool, Shadow_copies_delete_technique, Openconnect_tool, Crackmapexec_tool, Netexec_tool,
🎯Victims: Enterprises
🌐Geo: Usa
📚TTPs:
⚔️Tactics: 7
🛠️Technics: 28
🧨IOCs:
- IP: 6
- Domain: 2
- File: 47
- Command: 4
- Registry: 7
- Email: 1
- Url: 2
- Path: 1
- Hash: 11
💽Software: FortiGate, SonicWall, Active Directory, Telegram, Windows Security, Linux, qemu, xen, virtualbox, chrome, ...
🔢Algorithms: aes-256-gcm, xor, zip
🔠Functions: GetConsoleWindow, Get-WmiObject, Get-VM
🗂️Win API: VirtualAlloc, LoadLibraryA, GetProcAddress, CreateThread, CreateProcessA, AmsiScanBuffer, EtwEventWrite, GlobalMemoryStatusEx, GetDiskFreeSpaceExA, GetTickCount, ...
⚙️Win Services: sqlservr, sqlwriter, msmdsrv, sqlbrowser, sqlagent, sqladhlp, ocssd, acronisagent, cvd, powerpnt, ...
📜Programming Languages: powershell, rust, golang
💻Platforms: x64, x86, apple
#threatreport:
The activity attributed to the handle **tcherber** spans malware development, enterprise VPN access, ransomware, and botnet operations. Rust binaries contained the build path `/home/tcherber/`, linking the loader, stealer, HVNC implant, and Enigma ransomware; Go tooling contained the same username in its build paths. The recovered Go toolkit supports FortiGate and SonicWall VPN access, internal network discovery, Active Directory enumeration, SMB reconnaissance, and credential validation. Its database held more than 500 FortiGate VPN credential entries, and associated logs recorded successful, multi-hour VPN tunnels into multiple victim environments during May and June 2026.
A fake Windows Update installer drops PowerShell and batch files into `C:\Users\Public`, creates a temporary scheduled task, adds that directory to Microsoft Defender exclusions, and retrieves a Rust loader from `93.152.223.30:5000`. The loader downloads and launches stealer and HVNC payloads, checks for sandbox indicators, and can establish user-level persistence through the HKCU Run key. The stealer collects browser credentials and cookies, application data, selected files, system details, and screenshots. It packages stolen data in ZIP archives and exfiltrates them through HTTP to the campaign server and through Telegram.
The HVNC implant uses an AES-256-GCM-encrypted custom C2 protocol. It creates a hidden desktop, launches applications such as browsers within the victim’s user context, and captures desktop frames on demand. It also checks for virtualized environments and patches AMSI and ETW functions. The Enigma ransomware uses ChaCha12 encryption with ECIES-wrapped key material. Its capabilities include terminating processes and services, deleting shadow copies, impairing defenses, persistence, UAC bypass, and network-share propagation; analysis found no data-exfiltration mechanism.
Underground advertisements associated with tcherber offered VPN access, custom C2 and loader development, and crypting services. A separate unauthenticated STANDALONE C2 Panel v3 at `158.94.209.138` exposed builder, cryptor, and delivery functions. The panel showed 178 active bots and agent builds dated September 2026, indicating continued operations. Together, the credential inventory, VPN logs, malware, advertisements, and panel indicate activity consistent with both initial-access brokerage and malware-as-a-service operations.
#threatreport#HighCompleteness
OPTIMJACK: Optimizely-Hosted Scripts Loaded a Crypto Clipper and Card Skimmer on 186 Confirmed Sites | 16-11-2026
Source: stepsecurity.io/blog/optimja…
Key details below ↓
🧑💻Actors/Campaigns:
Optimjack
💀Threats:
Darksword_tool, Coruna_tool,
🎯Victims: News media, Software, Airlines, Retail, Postal and shipping services, Insurance, Housing and property management, Telecommunications, Banking and financial services, Nonprofit and humanitarian organizations, ...
🏭Industry: Aerospace, Financial, Retail
🌐Geo: Hong kong, Singapore, Usa
📚TTPs:
⚔️Tactics: 1
🛠️Technics: 0
🤖LLM extracted TTPs:`
T1056, T1059.007, T1082, T1105, T1113, T1115, T1119, T1189, T1195.002, T1539, ...
🧨IOCs:
- Domain: 8
- Url: 1
- File: 5
- IP: 3
- Hash: 2
- Coin: 8
💽Software: Zoom, Salesforce, Cloudflare, Trongrid
📲Wallets: tron
🪙Crypto: coingecko, bitcoin
🔢Algorithms: sha256, ed25519
📜Programming Languages: javascript, php
💻Platforms: apple
#threatreport:
On November 16, 2026, a compromised third-party JavaScript supply chain served a malicious payload from `1688360.cc`. A one-line loader injected `track.js` into affected pages, giving the attacker’s code the same browser privileges as the site’s own scripts. Public scans recorded requests from pages on 30 domains; copies of the payload were confirmed on 17, including news, software, airline, retail, insurance, and property-management sites. The exposure lasted at least 13 hours, beginning with a scan at 08:28 UTC and continuing after the final version appeared at 17:56 UTC. The report traced 12 of 15 example pages to scripts hosted on `cdn.optimizely.com`.
The roughly 30 KB JavaScript payload contained a cryptocurrency address swapper, a payment-card skimmer, and an iOS-specific branch. The swapper altered wallet addresses in page text and form fields, links and images, clipboard and copy operations, sharing, and fetch/XHR data. It also hooked `Function.prototype.call` and `apply`. After a visible address was changed, the script loaded `html2canvas`, captured the page as a JPEG, and sent it with the URL, up to 8,000 characters of accessible cookies, and up to 20 original addresses. This collection occurred only after a swap; HttpOnly cookies were inaccessible.
The skimmer inspected form submissions and network requests for Luhn-valid card numbers, expiry dates, and CVVs, then sent matches to `/card.php` using `sendBeacon`. It did not collect names or billing addresses and could not read fields inside cross-origin payment iframes. Captured card data was also stored in `sessionStorage`.
On iPhones, iPads, and iPods, the script loaded additional `boot.js` files from `.icu` domains. For devices running iOS 18.6 or earlier, it also created a hidden iframe. Although a researcher reported possible DarkSword or Coruna exploits, the second-stage files were not analyzed, so the exploit chain and affected versions remain unconfirmed. The report recommends investigating affected pages and logs, blocking the identified infrastructure, reviewing vendor-hosted scripts, and treating exposed payment and wallet data as compromised.
#threatreport#HighCompleteness
Poisoned Scripts on BBC, Zoom and Other Major Websites Linked to iPhone Exploit Loaders | 16-11-2026
Source: infostealers.com/article/poi…
Key details below ↓
🧑💻Actors/Campaigns:
Optimjack
💀Threats:
Coruna_tool, Darksword_tool,
🎯Victims: News media, Banking, Insurance, Communications, Software, Business services, Marketing technology
🏭Industry: Financial
🌐Geo: Singapore
📚TTPs:
⚔️Tactics: 2
🛠️Technics: 0
🤖LLM extracted TTPs:`
T1027, T1027.009, T1056, T1056.003, T1059.007, T1070, T1071.001, T1082, T1105, T1140, ...
🧨IOCs:
- Url: 18
- File: 8
- Domain: 11
- Hash: 33
- IP: 1
- Coin: 8
💽Software: Zoom, salesforce, Google Chrome, Chrome, Cloudflare
📲Wallets: tron
🪙Crypto: bitcoin, solana, litecoin, dogecoin
🔢Algorithms: base64, sha256, xor
🔠Functions: IP-sync
📜Programming Languages: javascript
💻Platforms: apple, intel
#threatreport:
Malicious JavaScript was found in responses from four third-party delivery providers: Optimizely, Kissmetrics, StorifyMe and minute-ly. Evidence includes 15 archived website captures containing 16 poisoned vendor responses, with hashes independently corroborated; researchers directly recovered nine malicious files. The findings identify 21 web properties, including BBC, Zoom, GitLab, CNN and Fox News, but do not establish how many visitors received the code or suffered harm.
The financial-theft payload, served through poisoned Optimizely project configurations, can replace cryptocurrency addresses in page content, form fields, links and clipboard operations. It also targets payment-card data and conditionally collects page and cookie information. The BBC and HSBC UAE samples contain a complete Base64-embedded payload, so blocking attacker domains alone may not prevent locally embedded code from running. The initial compromise method and any confirmed theft totals remain unknown.
An iOS-directed branch loads a router associated by researchers with Coruna and DarkSword infrastructure. Observed routes include A8, associated with Coruna, and D11, labelled DarkSword; C10 was configured but not verified as available. Two scans recorded successful responses from router and stage URLs while presenting iOS Safari user-agent strings. However, the scans used headless Chrome, not iOS Safari or WebKit, and therefore show delivery—not successful exploitation or compromise of an iPhone.
Defenders should preserve response bodies, hashes, timestamps and publication records, then remove poisoned code and verify the bytes delivered after remediation. They should block observed delivery and downstream hosts, while recognizing that domain blocking may not disable embedded payloads. Investigations should distinguish financial-theft activity from mobile-stage delivery and separate requested stages from executed exploits or confirmed data loss. The report does not establish the attacker’s identity or how the affected publishing systems were accessed.
#threatreport#MediumCompleteness
N-able N-central Exploit Chains Give Attackers Root and SYSTEM Access | 13-11-2026
Source: hivepro.com/threat-advisory/…
Key details below ↓
💀Threats:
4everable2shell_vuln, Agent2shell_vuln, Xmrig_miner, Supply_chain_technique, N-able_tool,
🎯Victims: Managed service providers
🌐Geo: Germany, United states, Poland
🔓CVEs: CVE-2026-125883 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *9.3*,
- Vulners: Exploitation: True
CVE-2026-90436 \[[Vulners](vulners.com/cve/CVE-2026-904…)]
- CVSS V3.1: *Unknown*,
- Vulners: Exploitation: Unknown
CVE-2026-125881 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *9.3*,
- Vulners: Exploitation: Unknown
CVE-2026-125880 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *8.3*,
- Vulners: Exploitation: Unknown
CVE-2026-125879 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *8.6*,
- Vulners: Exploitation: Unknown
CVE-2026-125878 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *8.7*,
- Vulners: Exploitation: Unknown
CVE-2026-91000 \[[Vulners](vulners.com/cve/CVE-2026-910…)]
- CVSS V3.1: *8.4*,
- Vulners: Exploitation: Unknown
CVE-2026-125882 \[[Vulners](vulners.com/cve/CVE-2026-125…)]
- CVSS V3.1: *8.6*,
- Vulners: Exploitation: Unknown
📚TTPs:
⚔️Tactics: 1
🛠️Technics: 0
🤖LLM extracted TTPs:`
T1005, T1018, T1059, T1059.007, T1068, T1072, T1078, T1098, T1105, T1136, ...
🧨IOCs:
- File: 3
- IP: 9
- Url: 7
- Hash: 1
- Coin: 1
🪙Crypto: monero
🔢Algorithms: sha256, zip
#threatreport:
N-able N-central is affected by several public exploit chains that can compromise the management server and, through it, managed Windows endpoints. Hotfix 2, build 2026.4.1.26, released November 10, fixes seven vulnerabilities, including CVE-2026-125883, an XXE flaw reported as exploited in the wild. However, the 4EverAble2Shell and Agent2Shell chains were verified against that patched build and have no N-central fix described in the report.
Able2Shell exploits unauthenticated XXE and SSRF flaws, including CVE-2026-125883 and CVE-2026-125881, to access administrative functions, steal a local session token, reset an administrator password, execute a planted JSP as the nable account, and create an administrator account. Other flaws enable stored XSS, OGNL command injection, path traversal, and abuse of root-running helpers to escalate from nable to root. CVE-2026-125879 involves improperly permissioned syslog-ng configuration consumed by a root process; CVE-2026-91000 is a local privilege-escalation flaw in versions before 2026.4.1.
4EverAble2Shell exploits CVE-2026-90436 in the bundled, end-of-life Apache Axis 1.x library. An unauthenticated SOAP request can trigger unsafe reflection while Axis resolves a service, before service-level authentication, enabling code execution on the N-central server. On November 12, an attacker reportedly used this chain to install XMRig on 21 servers, including 19 running Hotfix 2.
Agent2Shell targets managed Windows systems. An unauthenticated XMPP message can cause an online agent to run commands as SYSTEM; an alternate route stages a ZIP that can replace a Probe’s wsp.exe. Both routes require the appliance ID, which may be exposed through other vulnerabilities or a compromised managed device. A compromised N-central server can provide access to managed systems, credentials, and administrative functions. The report also notes exploitation attempts against CVE-2026-125883 and additional indicators of compromise reported on November 13.
#threatreport#MediumCompleteness
CVE-2026-84869 in ScreenConnect Exploited to Run Malicious Files on MSP Technician Accounts | 16-11-2026
Source: huntress.com/blog/cve-2026-8…
Key details below ↓
💀Threats:
Screenconnect_tool, Sckit_tool,
🎯Victims: Managed service providers
🔓CVEs: CVE-2026-84869 \[[Vulners](vulners.com/cve/CVE-2026-848…)]
- CVSS V3.1: *9.9*,
- Vulners: Exploitation: True
Soft:
- connectwise screenconnect (<26.6.5.9742)
📚TTPs:
⚔️Tactics: 8
🛠️Technics: 18
🧨IOCs:
- Domain: 1
- File: 15
- Url: 1
- Command: 1
- IP: 33
- Path: 1
- Hash: 1
💽Software: Telegram, curl, Windows Service, Microsoft 365
🔢Algorithms: base64
📜Programming Languages: vbscript, powershell
#threatreport:
CVE-2026-84869 is a ScreenConnect client vulnerability disclosed on September 8, 2026, with a CVSS score of 9.9. It allows a guest in Support or Access sessions to trigger file-transfer actions that the technician’s host client processes without authorization or a confirmation prompt, including executing files with elevated rights. The vulnerable component is the technician’s client, so updating the ScreenConnect server alone does not protect technicians. ConnectWise fixed the flaw in host client and Access agent version 26.6.5; removing the TransferFiles permission is an interim mitigation.
On November 12, investigators observed a technician join a blank-named Access session and have the client execute an attacker-pushed batch file. Within about nine seconds, the script reported to a Telegram bot, downloaded an installer from masterdarky.com, and launched a rogue ScreenConnect client configured to connect to the attacker’s relay. A second technician’s client ran an attacker-supplied canary file on November 13. Earlier activity included guest-pushed files sent to at least four additional technicians, including an attempted path traversal to place Sync.bat in a Startup folder.
From October 20 through November 16, attackers sent approximately 688,000 file-execution requests for update.bat and update_<8 hex>.bat files to 42 MSP ScreenConnect servers, using 29 source IP addresses. Repeated requests appear designed to wait for a technician to join a session. Matching payload names and timing link the spraying activity to the observed compromise. A separate campaign involving unauthorized guest sessions with cryptocurrency-exchange names also targeted MSP servers, although its relationship to this exploitation was not established.
Hunt for ScreenConnect client processes launching scripts or text files from Temp or other unexpected locations, rogue clients such as WindowsUpdate.WindowsClient.exe, suspicious wclient_* directories, wupdate_*.msi files, and clients configured to use unrecognized relays. A relevant detection is curl downloading an MSI from a script launched by the ScreenConnect host client. Update technician clients and Access agents to version 26.6.5 or later, restrict guest access and permissions, and investigate affected workstations for credential exposure.
#threatreport#LowCompleteness
CookUp follow-up: nine signing companies and 25 C2 domains behind the “useful app” installers | 11-11-2026
Source: rogueauthority.org/cookup-cl…
Key details below ↓
💀Threats:
Tamperedchef, Opensupdater, Projextor, Catchlight,
🌐Geo: Ukraine, United kingdom, Ukrainian, Mexico
🤖LLM extracted TTPs:`
T1027, T1027.001, T1027.009, T1036, T1036.005, T1090.002, T1129, T1140, T1204.002, T1480, ...
🧨IOCs:
- File: 34
- Domain: 26
- Hash: 65
💽Software: CookUp, Claude, Opus 5, Claude Code, Cloudflare, Electron
🔢Algorithms: ecc
🗂️Win API: RUN
#threatreport:
The report links installers signed by nine different companies to a single campaign associated with CookUp and publicly detected as TamperedChef. The installers use varied formats, including Go programs, .NET applications, NSIS and Inno Setup packages, and fake copies of 7-Zip’s `7z.dll`. Despite their different appearances, the confirmed samples share the same payload-unpacking scrambler and a distinctive 32-byte constant. The cluster also overlaps with OpenSUpdater and installer activity previously associated with Catchlight.
The installers use extended, junk-padded signature areas that can cause analysis tools to miss or misreport their EV code signatures, although Windows accepts them. The nine signers include five US companies, three Kyiv-based Ukrainian companies, and one UK company. Their installers contact 25 command-and-control domains, all hosted behind Cloudflare; the report notes that Cloudflare IP addresses are not useful indicators for blocking. One identified domain, `codeonicinc.com`, is linked to samples analyzed by G DATA.
The installers are disguised as recipe applications, PDF tools, note-taking software, game cheats, and generic download installers. Some deliver working applications as cover. As in the earlier CookUp analysis, the C2 servers appear to decide whether to provide the final payload based on the victim, and observed systems were not selected. Consequently, the final-stage payload remains unknown. Some communications are encrypted, limiting visibility into the requests made by the installers.
#threatreport#MediumCompleteness
Lotuslite V2: The Evolution of Mustang Panda’s .NET Backdoor | 15-11-2026
Source: antares-research.com/reports…
Key details below ↓
🧑💻Actors/Campaigns:
Red_delta (🧠motivation: cyber_espionage)
💀Threats:
Lotuslite, Dll_sideloading_technique,
🎯Victims: Thailand
🌐Geo: Thailand, China
📚TTPs:
⚔️Tactics: 4
🛠️Technics: 0
🧨IOCs:
- Url: 1
- File: 9
- IP: 2
- Registry: 1
- Path: 2
- Hash: 3
💽Software: Windows registry, winhttp, NET Framework
🔢Algorithms: aes-256-gcm, aes-gcm, xor, zip
🗂️Win API: LdrLoadDll
#threatreport:
The .NET implant Lotuslite V2 was delivered through a public Zoho WorkDrive link as an assessment-themed ZIP archive. The archive contained an executable and `CommonLib.dll`; the executable masqueraded as a signed PDFgear program and side-loaded the malicious DLL. The malware used Windows registry modifications for persistence: it wrote to the system-wide Run key when executing as SYSTEM, or configured the current user’s Windows Load value otherwise. It also used a PDFgear file-association hijack. Memory analysis identified the named Windows Mutant `PDFgear_DeviceSync_Mutex`, an artifact useful for host detection and scoping.
The implant stores its command-and-control configuration in XOR-encoded byte arrays, decoded at runtime with key `0x55`. It contains two hardcoded IP addresses, `31.59.184.185` and `103.175.50.32`, and is configured for UDP port 53 and WSS over ports 443, 80, 22, 21, 8080, and 1080. Failed connections are retried with progressively longer delays of 1, 3, 5, 10, and 30 seconds.
Lotuslite V2 uses the protocol marker `0x50524F31` (“PRO1”), replacing identifiers reported in earlier LOTUSLITE versions, and protects its formalized binary protocol with application-layer AES-256-GCM encryption. This change makes detections that depend on the older protocol markers less effective. The report assesses with high confidence that the implant is a modernized LOTUSLITE successor associated with Mustang Panda (Earth Preta), based on shared command-and-control conventions, execution and persistence techniques, and regional victimology.
#threatreport#MediumCompleteness
Unpacking a Multi-Stage Loader: From Embedded Payloads to Cobalt Strike Beacon | 10-11-2026
Source: threatcat.ch/blog/unpacking-…
Key details below ↓
💀Threats:
Cobalt_strike_tool, Prependloader, Process_injection_technique, Domain_fronting_technique,
📚TTPs:
⚔️Tactics: 1
🛠️Technics: 0
🤖LLM extracted TTPs:`
T1027, T1027.002, T1027.007, T1027.009, T1027.010, T1027.013, T1027.015, T1055, T1059.001, T1059.003, ...
🧨IOCs:
- File: 32
- Domain: 2
- Command: 4
- Path: 2
- Url: 3
- Hash: 3
💽Software: curl, Chrome
🔢Algorithms: md5, rc4, lznt1, sha256, zip, xor, gzip, base64, sha1
🔠Functions: GetType, GetMethod, GetField, GetValue, GetProperty, CreateInstance, Get-Content, GetExecutingAssembly, GetManifestResourceStream, GetParameters, ...
🗂️Win API: EXE, Decompress, VirtualProtect, VirtualAllocEx, CreateThread, SetThreadContext
📜Programming Languages: powershell
💻Platforms: x64
#threatreport:
The infection chain begins with a ZIP archive containing a Windows shortcut that carries additional data after its normal file structure. The shortcut’s script locates itself, searches for a `REM` marker, and uses `curl` with a `file://` URL and `--continue-at` to extract the bytes following the marker. It saves the extracted content as a tar archive and unpacks a batch/PowerShell polyglot dropper.
The roughly 700 KB dropper contains Base64-encoded content for a decoy PDF and a gzip-compressed .NET assembly. It extracts and opens the PDF, which displays a luxury-apartment brochure consistent with a real-estate lure. It then pipes its own source into a second PowerShell process; batch/PowerShell syntax and split strings help conceal the PowerShell commands. The script restores the deliberately altered GZip header, decompresses the assembly in memory, and loads and invokes it reflectively without writing the assembly to disk. The batch portion also deletes the dropper. The report does not establish the .NET payload’s purpose at this stage.
The assembly extracts an embedded x64 shellcode resource, allocates unmanaged memory, copies the shellcode into it, changes the memory to executable and writable with `VirtualProtect`, and invokes it through a delegate. The shellcode is identified as Cobalt Strike’s PrependLoader, using nested XOR, RC4, and LZNT1 transformations, EAF bypass and indirect syscalls, as well as reflective-loader drip loading.
The final payload is a 64-bit Cobalt Strike Beacon. Its configuration indicates version 4.13 or later, including a 15-second check-in delay; process-injection drip-loading settings provide a lower bound of version 4.12. Beacon is configured for HTTPS command-and-control at `d3qx7quq68rdce.cloudfront.ne…`, Base64-decodes task data from HTTP response bodies, and uses process injection configured with `VirtualAllocEx`, writable-to-executable memory protection, and a 110 ms drip-load delay.
#threatreport#HighCompleteness
QUIC and Covert: Novel HTTP/3-Powered RATs Target Southeast Asian Governments | 16-11-2026
Source: unit42.paloaltonetworks.com/…
Key details below ↓
🧑💻Actors/Campaigns:
Earth_alux
💀Threats:
Credential_harvesting_technique, Dll_sideloading_technique, Squidoor, Finaldraft, Perfectpioneer, Kukurat, Aspxspy_shell, Godzilla_webshell, Suo5_tool, Process_injection_technique, Credential_stealing_technique, Dcsync_technique, Credential_dumping_technique, Mimikatz_tool, Procdump_tool,
🎯Victims: Government
🏭Industry: Government
🌐Geo: America, Asia, Myanmar, Asian
📚TTPs:
⚔️Tactics: 6
🛠️Technics: 0
🧨IOCs:
- File: 11
- Hash: 70
- Domain: 9
- IP: 2
- Command: 2
💽Software: WildFire, Linux, Microsoft Exchange, GlobalProtect, PsExec, Windows service, Google Chrome, Chrome, Active Directory, Local Security Authority, ...
🔢Algorithms: sha256, xor, base64
🔠Functions: Get-ManagementRoleAssignment
🗂️Win API: VirtualAllocEx, WriteProcessMemory, CreateRemoteThread, PsExecService, CreateProcessA
📜Programming Languages: powershell, golang
💻Platforms: x86
#threatreport:
TGR-STA-0049, also tracked as REF7707, Earth Alux and Ink Dragon, has targeted government organizations and other high-value entities since at least 2023. The campaign described began by mid-2025 and combines exploitation of vulnerable Microsoft Exchange and web servers with phishing. Initial access included web-shell deployment, including ASPXSpy, Godzilla, Bysunue.aspx and Suo5. Phishing used lookalike government domains and executables disguised as PDFs that opened decoy documents.
The actor deployed PerfectPioneer, a custom Go-based RAT, and later KUKURAT, a modular C++ backdoor. Both communicate with command-and-control infrastructure over HTTP/3 using QUIC over UDP; some variants fall back to or use HTTP/2. PerfectPioneer supports file operations, shell commands, process management, keylogging, screenshots and SOCKS5 proxying. It also includes reconnection and cleanup routines, and an “isDangerousCommand” check that blocks commands likely to cause disruption or attract security attention. The malware has been delivered through DLL side-loading, decoy installers and phishing.
KUKURAT retrieves capability modules from its C2 server at runtime, enabling functions including file collection, command execution, process injection, network scanning, PsExec-based lateral movement, SMB credential brute-forcing and SOCKS tunneling. Its camouflage engine generates decoy web requests and temporary files and imitates Chrome-related registry activity. In some intrusions, a watchdog relaunched KUKURAT if it stopped running.
After compromising a domain controller, the attackers used logon scripts to distribute malware and also moved laterally through PsExec, WMI and WinRM. Credential access included LSASS dumping, Mimikatz, ProcDump, SharpDPAPI, SAM/SYSTEM hive collection and DCSync. The campaign targeted documents related to political affairs and international relations, and used Exchange impersonation privileges to access mailboxes and collect email data.
#threatreport#LowCompleteness
Typosquatters Copy Big Brands’ New Domains Within a Day of Registration | 16-11-2026
Source: security.com/threat-intellig…
Key details below ↓
💀Threats:
Typosquatting_technique, Credential_harvesting_technique,
🎯Victims: Pharmaceuticals, Medical devices, Automotive retail, Food services, Consumer goods
🌐Geo: China
🤖LLM extracted TTPs:`
T1583.001, T1656
🧨IOCs:
- Domain: 27
💻Platforms: intel
#threatreport:
The report describes a typosquatting campaign targeting newly registered domains associated with major brands, including Abbott Laboratories, Bayer, CarMax, McDonald’s, Procter & Gamble, and Sanofi. The authentic domains were registered through brand-protection registrars CSC or MarkMonitor, often for campaign sites such as patient-savings programs, sweepstakes, claims portals, or legal settlements. Within the next working day, attackers registered closely resembling domains through a China-based registrar. The report identifies 25 lookalike domains and describes the pattern recurring across 27 instances.
The lookalikes used subtle one-character changes that preserved the names’ readability and plausibility. The reported examples included shortened versions of CarMax domains, such as registering `carmaxclaim.com` after `carmaxclaims.com`. The actors did not rely on common techniques such as keyboard-adjacent substitutions, homoglyphs, or hyphen insertion. The timing suggests they monitored new registrations made through the two brand-protection registrars. The activity continued through at least November 11, 2026.
When accessed, the lookalike domains served parked pages with pay-per-click advertising, deceptive search portals, or brand-themed blogs containing ads that directed visitors to scam sites. The report notes that these domains could also be repurposed through DNS or content changes to host malicious pages, including credential-harvesting or fake claims forms; it does not report that such pages were observed in this campaign.
#threatreport#LowCompleteness
A USB worm’s second life: Retadup spreading CoffeeLoader | 16-11-2026
Source: securelist.com/retadup-deliv…
Key details below ↓
💀Threats:
Retadup, Coffeeloader, Armouryloader, Smokeloader,
🌐Geo: Peru, Russia, Bolivia, Angola
🤖LLM extracted TTPs:`
T1027.002, T1027.009, T1027.013, T1036.005, T1055, T1055.002, T1059.010, T1071.001, T1082, T1091, ...
🧨IOCs:
- File: 8
- Hash: 6
- Domain: 9
- Url: 2
💽Software: OpenGL, Firefox, cloudflare, AutoIt, AutoHotkey, WinHTTP, nginx
🔢Algorithms: md5, xor, base64
💻Platforms: x64
#threatreport:
In August 2026, a new CoffeeLoader variant was observed being distributed by the Retadup worm. Its NSIS installer drops a modified `7z.dll` into a temporary directory and executes it with `rundll32.exe`. The library contains self-decrypting shellcode using a new version of the Armoury packer. Unlike earlier versions, which used OpenCL for GPU-assisted decryption, this version uses OpenGL.
The decrypted shellcode acts as a CoffeeLoader downloader, combining installation, persistence, process injection, and payload retrieval. It copies `7z.dll` to a subfolder under `%PROGRAMDATA%` or `%LOCALAPPDATA%`, with names derived from system volume information, and uses a Run registry key for autorun. It also creates a registry key intended to prevent users from disabling autorun through Task Manager. The downloader injects itself into a process such as `dllhost.exe`, `WerFault.exe`, `svchost.exe`, `OpenWith.exe`, or `AtBroker.exe`, then retrieves CoffeeLoader’s main payload from a hardcoded C2 server over HTTPS.
The downloaded payload is XOR-encrypted, with the first 16 bytes serving as the key. The downloader verifies a response signature and checksum before decrypting and executing the shellcode. The main CoffeeLoader module retains its previous payload-execution options, but uses updated encryption keys and a new protocol magic value.
Retadup spreads through removable drives by copying hidden files and LNK shortcuts; when a user opens an infected shortcut, the worm can infect another computer. Although its botnet was dismantled in 2019, former C2 domains were re-registered and used to distribute other malware, including CoffeeLoader. The reported campaign blocked infection attempts on about 15,000 systems, mainly in Peru, Angola, Russia, and Bolivia.
#threatreport#LowCompleteness
WPPConnect Compromised: Malicious Releases Steal Secrets and WhatsApp Sessions Through the Project's Own Pipeline | 16-11-2026
Source: stepsecurity.io/blog/wppconn…
Key details below ↓
💀Threats:
Mitm_technique, Typosquatting_technique,
🎯Victims: Information technology
🤖LLM extracted TTPs:`
T1005, T1016.001, T1020, T1027, T1033, T1059.007, T1082, T1083, T1119, T1195.002, ...
🧨IOCs:
- File: 9
- IP: 1
- Hash: 3
💽Software: WhatsApp, Telegram, Docker, arrayref
🔢Algorithms: sha1, base64, sha256
🗂️Win API: lockfile
📜Programming Languages: rust
#threatreport:
On November 15–16, 2026, malicious releases of `@wppconnect-team/wppconnect` 2.4.1 and `@wppconnect/server` 2.10.82 were published through the project’s GitHub Actions pipeline using valid trusted-publisher credentials. The library’s malicious commit was also associated with a SLSA provenance attestation; the attestation identifies the build source but does not establish that the code was safe. Pipeline logs and Docker Hub records indicate that 2.4.3 and 2.10.83 may also have contained the payload, although those versions are not listed on npm.
The payload runs when the library is imported or when the server starts. It collects host details, public IP, environment variables, portions of selected `.env` and configuration files, and up to five WhatsApp session-token files. The server variant also checks for CI and analysis environments and delays execution; the library variant has no such checks. Collected data is sent as a text file through Telegram’s Bot API. WhatsApp session tokens contain credentials for the web interface and should be treated as compromised if accessed by the payload.
The server version depends on the WPPConnect library, so affected server installations could load the malicious library transitively. Docker images built from payload-bearing commits were also affected. The report identifies images tagged 2.10.82 and 2.10.83, and says floating tags such as `latest`, `2`, and `2.10` pointed to the 2.10.83 image until 03:54 UTC on November 16. The state of the `main` tag before 07:31 UTC and the cleanliness of replacement images were not verified.