October 3, 2026
Why this day matters
- GitLab fixes critical AI Gateway flaw that could let authenticated Duo users escape a prompt sandbox and execute commands on self-hosted gateways.
- doxx.net’s new ADN platform prevents agentic misadventure while the agent is operating under the user’s authority. The post doxx.net Raises $38 Million to Prevent AI Agent-on-the-Internet Misadventures appeared first on SecurityWeek .
Enriched source records
Luna-generated claims retained by the evidence and claim-validation gates. Automated validation does not imply human review. A record does not need a CVE to be included.
Luna-enriched source article · securityaffairsAntino Backdoor Lets China-Linked UAT-11587 Turn Microsoft 365 Into a C2 Channel
Cisco Talos tracks UAT-11587, which it assesses with high confidence as China-nexus; by July 2026, the group had targeted at least 16 government and policy organizations across eight Asian countries.
Antino Backdoor Lets China-Linked UAT-11587 Turn Microsoft 365 Into a C2 Channel
Cisco Talos tracks UAT-11587, which it assesses with high confidence as China-nexus; by July 2026, the group had targeted at least 16 government and policy organizations across eight Asian countries.
Source published Oct 3, 2026, 9:26 AM UTC · Evidence retrieved Oct 3, 2026, 2:51 PM UTC
What happened
Cisco Talos tracks UAT-11587, which it assesses with high confidence as China-nexus; by July 2026, the group had targeted at least 16 government and policy organizations across eight Asian countries. [1] [2]
Antino is a Rust-compiled Windows backdoor supporting reconnaissance, shell and PowerShell execution, file transfer, in-memory shellcode loading and persistence on both 32-bit and 64-bit Windows. [3] [4] [5]
Antino uses Microsoft Graph with Outlook and OneDrive for command-and-control: it reads commands from an Outlook mailbox, returns results through email, and transfers stolen files through OneDrive. [6] [7] [8] [9]
The intrusion began with targeted phishing using researched decoy documents, including material copied from a Taiwan Ministry of Finance ruling and a reused Associated Press story. [12] [13] [14] [15]
The delivery chain used a Gmail-like attachment preview linking to an attacker-controlled page, followed by HTA, Windows Script Host and .NET deserialization stages; Antino was sideloaded through a signed Microsoft diagnostic binary. [21] [22] [23] [24]
Why it matters
Using Microsoft Graph allows Antino traffic to blend with normal Microsoft 365 activity; the implant polls its Outlook mailbox for commands every ten seconds. [7] [8] [10] [11]
A spoofed message passed SPF but failed DMARC alignment; because the impersonated domain used monitoring rather than rejection, the email still reached the inbox. [16] [17] [18] [19] [20]
Known limitations
Talos based its China-nexus assessment on development, preparation-environment and targeting indicators, including Simplified Chinese metadata, UTC+8 timestamps and a mainland-China Rust package mirror; the assessment was not attributed to a single indicator. [2] [25] [26] [27]
Cited source evidence
These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.
- [1]
Cisco Talos has been tracking a cluster of espionage activity since September 2025 that it calls UAT-11587, and by July 2026 the group had hit at least 16 government and policy organizations across eight Asian countries.
- [2]
“Based on the development, preparation-environment, and targeting indicators detailed in this report, Talos assesses with high confidence that UAT-11587 is China-nexus.” the report concludes.
- [3]
Antino is written in Rust and works on both 32-bit and 64-bit versions of Windows.
- [4]
It supports the usual backdoor features, including shell and PowerShell access, file transfers, in-memory shellcode execution and persistence.
- [5]
“Antino is a Rust-compiled Windows backdoor that supports host reconnaissance, shell and PowerShell execution, file transfer, in-memory shellcode loading and persistence.
- [6]
Its native command-and-control channel operates exclusively through Microsoft 365, using Microsoft Graph to interact with Outlook and OneDrive.” reads the report published by Talos.
- [7]
What makes Antino different is how it communicates: instead of using its own command-and-control server, it uses Microsoft Graph to read commands from an Outlook mailbox and send stolen files to OneDrive.
- [8]
Once Antino is running, it checks its Outlook mailbox for new commands every ten seconds.
- [9]
“Command messages are identified by the subject prefix command_req_[session_id] and responses by command_res_[session_id], as indicated in the HTTP GET request sent by Antino: “ A separate system is used to upload stolen files to one OneDrive folder and download attacker tools from another.
- [10]
This allows its traffic to blend in with normal Microsoft 365 activity.
- [11]
The implant actively pulls commands from the threat actor’s Outlook mailbox folder every 10 seconds.
- [12]
The attack starts in a more traditional way, with a convincing phishing email.
- [13]
UAT-11587 created highly targeted documents that suggest the attackers had researched their victims, including a fake workshop document about Taiwan’s information warfare and a document that closely copied a real Taiwan Ministry of Finance ruling about tax treatment for legislators.
- [14]
“The document exactly reproduces a public Taiwan Ministry of Finance ruling to make the decoy appear credible.” conctinues the report.
- [15]
“Its subject strongly suggests that it was prepared for members of Taiwan’s public sector.” In another case, the attackers reused a real Associated Press story about alleged Russian offers to the US over Venezuela.
- [16]
The delivery trick is a simple but effective email spoofing technique that takes advantage of a gap many people overlook.
- [17]
The attackers sent the email through a legitimate provider, using one domain as the technical sender, while the visible From address showed the organization they were impersonating.
- [18]
SPF passed because the real sending domain was authorized, but DMARC detected the mismatch and failed.
- [19]
That still wasn’t enough to block the message.
- [20]
The impersonated domain had a DMARC policy set to monitoring rather than rejection, so the failed email still reached the inbox.
- [21]
The attackers recreated Gmail’s normal attachment preview card almost pixel by pixel using images embedded in the email’s HTML.
- [22]
They then made the fake preview a link to a page controlled by the attackers.
- [23]
From that click, a five-stage chain kicks in, leaning on legitimate-looking HTA files, Windows Script Host, and a scripted .NET deserialization trick that abuses a known gadget chain to load malicious code inside a trusted process.
- [24]
The final stage sideloads Antino through a signed Microsoft diagnostic binary, meaning the thing dropping the backdoor onto disk is a tool Windows itself trusts by default.
- [25]
The attribution case leans on accumulated small details rather than any single smoking gun, which is honestly how these assessments should work.
- [26]
Decoy document metadata carries Simplified Chinese language tags and a UTC+8 timestamp, a combination more consistent with mainland China than Taiwan or Hong Kong, where Traditional Chinese dominates.
- [27]
Separately, ten different Antino builds reference a Rust package mirror built specifically to speed up dependency downloads inside mainland China, the kind of tooling choice a developer picks for convenience, not for disguise.
Luna-enriched source article · the hacker newsWarlock Exploits SharePoint Flaws to Disable Security Tools and Deploy Ransomware
The suspected China-linked threat actor Warlock is continuing to weaponize Microsoft SharePoint vulnerabilities, likely including both older and newer flaws, against organizations in Portuguese- and Spanish-speaking countries.
Warlock Exploits SharePoint Flaws to Disable Security Tools and Deploy Ransomware
The suspected China-linked threat actor Warlock is continuing to weaponize Microsoft SharePoint vulnerabilities, likely including both older and newer flaws, against organizations in Portuguese- and Spanish-speaking countries.
Source published Oct 3, 2026, 2:36 PM UTC · Evidence retrieved Oct 4, 2026, 1:23 AM UTC
What happened
The suspected China-linked threat actor Warlock is continuing to weaponize Microsoft SharePoint vulnerabilities, likely including both older and newer flaws, against organizations in Portuguese- and Spanish-speaking countries. [1]
Observed activity has targeted critical infrastructure, government, and education organizations, according to the Symantec and Carbon Black Threat Hunter Team. [2]
Cited source evidence
These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.
- [1]
The suspected China-linked threat actor known as Warlock is still continuing to weaponize Microsoft SharePoint vulnerabilities, likely both old and new, in attacks targeting organizations in Portuguese- and Spanish-speaking countries.
- [2]
The activity, observed by the Symantec and Carbon Black Threat Hunter Team, has hit critical infrastructure, government, and education organizations.
Luna-enriched source article · securityaffairsFake Zoom installer hides macOS backdoor CloudSyncD
Jamf Threat Labs identified CloudSyncD as a fake macOS Zoom installer that uses invisible zero-width Unicode characters to conceal a phished password.
Fake Zoom installer hides macOS backdoor CloudSyncD
Jamf Threat Labs identified CloudSyncD as a fake macOS Zoom installer that uses invisible zero-width Unicode characters to conceal a phished password.
Source published Oct 3, 2026, 3:11 PM UTC · Evidence retrieved Oct 3, 2026, 8:51 PM UTC
What happened
Jamf Threat Labs identified CloudSyncD as a fake macOS Zoom installer that uses invisible zero-width Unicode characters to conceal a phished password. [1] [2] [3] [4]
The disk image presents a Zoom-branded volume and instructs users to bypass Gatekeeper because the app is only ad-hoc signed. [5] [6] [7]
The installer repeatedly prompts for the local account password, validates it with dscl, and proceeds only after the validation succeeds. [8] [9] [10]
After validation, the malware encodes the password, adds variable filler, and stores it in a field resembling an application cache value in data.json. [11] [12] [13] [14]
The dropper contains a complete universal Mach-O payload and first attempts execution from memory; after that method failed in testing because of System Integrity Protection, it writes a temporary file and runs it with sudo using the stolen password. [15] [16] [17] [18] [19] [20]
Why it matters
The second-stage implant checks in every 8 to 16 seconds with a hardware identifier and can receive either a compressed archive or a complete executable to unpack or run. [21] [22] [23]
By publication, samples were communicating with two live domains; neither domain was detected as malicious at that time, and all samples used the same encryption key and initialization vector. [24] [25] [26] [27]
Known limitations
In recovered builds, the cleanup mechanism was present but did not activate because its intended replacement bundle was absent. [28] [29] [30]
Cited source evidence
These excerpts support the numbered claims above. They come from one source article and do not establish independent corroboration.
- [1]
Jamf Threat Labs details CloudSyncD, a fake macOS Zoom installer that hides a phished password using invisible zero-width Unicode characters.
- [2]
Jamf Threat Labs found CloudSyncD while doing routine scanning on VirusTotal, buried inside a disguised Zoom client.
- [3]
The trick used to find the hidden password is also where the file gets its name.
- [4]
When decoded, they tell the malware where the real password starts and how long it is inside the padded text.
- [5]
The delivery method is a classic macOS trick, made to look like a normal installer.
- [6]
The disk image opens as a volume called Zoom, with an app icon and an Applications shortcut.
- [7]
The background image then guides the victim step by step on how to bypass Gatekeeper, because the app is only ad-hoc signed and macOS would normally block it.
- [8]
After the app starts, a fake dialog asks for the user’s password, claiming it is needed to continue the installation.
- [9]
The prompt keeps appearing until the password is correct.
- [10]
“Once a password is entered, the dropper validates it against the local account using dscl and does not continue until the check succeeds.” reads the analysis published by Jamf Threat Labs.
- [11]
“After the password has been validated a fake progress window reading Downloading Zoom...
- [12]
appears.” The captured password doesn’t get sent anywhere immediately.
- [13]
It gets base64 encoded, padded with a random amount of filler text, and buried inside a field that looks like an innocent cache value in a fake settings file named data.json.
- [14]
The amount of filler changes each time, so the password is hidden in a different position on every run, while the technique stays the same.
- [15]
The malware does not download the payload separately.
- [16]
The full payload is already inside the dropper as a universal Mach-O file.
- [17]
It first tries to run the payload directly from memory through an anonymous file descriptor, avoiding writing it to disk.
- [18]
On most Macs, however, System Integrity Protection blocks this method.
- [19]
Jamf also saw the attempt fail during testing and recorded the exact error.
- [20]
When the fileless method fails, the malware writes the payload to a temporary file and runs it with sudo , using the password it just stole to get the privileges it needs.
- [21]
The second stage, which Jamf calls cloudsyncd after its disguised daemon name, is a fairly restrained implant once it’s actually running.
- [22]
It just builds a working directory, logs its activity with encrypted records, and checks in with its server every 8 to 16 seconds carrying nothing but a hardware identifier.
- [23]
The server can send back either a compressed archive to unpack or a complete executable to run.
- [24]
By the time Jamf published its research, the malware had moved beyond testing and was communicating with two live domains.
- [25]
Both were registered through the same registrar in 2011 and were protected by Cloudflare.
- [26]
At the time, neither domain was detected as malicious.
- [27]
All the samples also used the same encryption key and initialization vector.
- [28]
Upon failure, it uses mkstemp to write the file temporarily to disk, then executes it using sudo along with the collected user’s password.” A cleanup script follows, assembled at runtime from scattered obfuscated fragments and designed to swap in a replacement app bundle before deleting itself either way.
- [29]
In the builds Jamf actually recovered, though, the replacement bundle simply doesn’t exist, so this entire swap mechanism never fires.
- [30]
It’s a feature built and wired up for a payload that hadn’t shipped yet.
Additional source records
Material developmentsFortra Patches Critical Vulnerabilities in BoKS
Securityweek published a source item for review.
Fortra Patches Critical Vulnerabilities in BoKS
Securityweek published a source item for review.
What happened
Securityweek published a source item for review.
Why it matters
A reviewed impact interpretation has not been published for this record.
Reviewed next steps
- Review the exact source item and determine whether it changes exposure or monitoring priorities.
Evidence
- Fortra Patches Critical Vulnerabilities in BoKS Securityweek · Published 2026-10-03T11:34:00Z · Retrieved Oct 3, 2026, 1:23 PM UTC
Known limitation
This item is supported by one source record and has not been independently corroborated here.
Threat and risk signalsCVE-2026-90970: Critical GitLab AI Gateway Flaw Fixed
Securityaffairs published details for CVE-2026-90970.
CVE-2026-90970: Critical GitLab AI Gateway Flaw Fixed
Securityaffairs published details for CVE-2026-90970.
What happened
Securityaffairs published details for CVE-2026-90970.
Why it matters
A reviewed impact interpretation has not been published for this record.
Structured associations
Reviewed next steps
- Check asset inventory and patch status for CVE-2026-90970.
Evidence
- CVE-2026-90970: Critical GitLab AI Gateway Flaw Fixed Securityaffairs · Published 2026-10-03T10:43:39Z · Retrieved Oct 3, 2026, 2:51 PM UTC
Known limitation
This item is supported by one source record and has not been independently corroborated here.
Cloud and infrastructureThe State of Cybersecurity in 2026: Key Segments, Insights, and Innovations
The Hacker News published a source item for review.
The State of Cybersecurity in 2026: Key Segments, Insights, and Innovations
The Hacker News published a source item for review.
What happened
The Hacker News published a source item for review.
Why it matters
A reviewed impact interpretation has not been published for this record.
Reviewed next steps
- Review the exact source item and determine whether it changes exposure or monitoring priorities.
Evidence
- The State of Cybersecurity in 2026: Key Segments, Insights, and Innovations The Hacker News · Published 2026-10-03T11:00:00Z · Retrieved Oct 3, 2026, 1:23 PM UTC
Known limitation
This item is supported by one source record and has not been independently corroborated here.
AI and model realitydoxx.net Raises $38 Million to Prevent AI Agent-on-the-Internet Misadventures
Securityweek published a source item for review.
doxx.net Raises $38 Million to Prevent AI Agent-on-the-Internet Misadventures
Securityweek published a source item for review.
What happened
Securityweek published a source item for review.
Why it matters
A reviewed impact interpretation has not been published for this record.
Reviewed next steps
- Review the exact source item and determine whether it changes exposure or monitoring priorities.
Evidence
- doxx.net Raises $38 Million to Prevent AI Agent-on-the-Internet Misadventures Securityweek · Published 2026-10-03T11:45:00Z · Retrieved Oct 3, 2026, 1:23 PM UTC
Known limitation
This item is supported by one source record and has not been independently corroborated here.