Weekly Threat Intel: July 24–31, 2026 — The Indicator

29 Reports analyzed

656 IOCs extracted

9 Reports from one vendor across two batch windows

97 IOCs in the week's largest single report

4 Reports on Claude-brand abuse or AI-agent-targeted attacks

Key Takeaways for Security Professionals

Patch Check Point SmartConsole and Arista VeloCloud Orchestrator immediately — both face active exploitation.

CVE-2026-16232 (Check Point) grants unauthenticated attackers full administrator rights; CVE-2026-16812 (Arista) is an unauthenticated OS command injection zero-day. Both are already being exploited in the wild.
Hunt CVE-2026-12569 across any PTC Windchill/FlexPLM deployment — two vendors confirm active Cl0p exploitation.

HivePro and Censys independently corroborated Cl0p-affiliated actors exploiting this RCE to deploy JSP web shells and steal engineering IP for extortion — treat matching independent reports as higher-confidence than any single source.
Add cloudlanecdn[.]com and logAzure.txt to watchlists — a third vendor now documents the same M365 calendar C2.

Following Group-IB and Securelist's independent reports last week, HivePro's advisory this week adds Israeli-entity targeting details to the same HOLLOWGRAPH campaign, reinforcing it as an active, multi-vendor-tracked threat.
Treat AI Skills/MCP GitHub repos as untrusted until verified — FakeGit now baits AI coding agents directly.

The AgentBaiting variant of FakeGit specifically targets AI coding assistants that autonomously browse and fetch tools, not just human developers, delivering SmartLoader and StealC malware through convincing but fake repositories.
Only install Claude Desktop from claude.ai/download — fake install guides and malvertising continue across platforms.

This week alone saw a fake Claude install guide deploying MacSync's six-stage macOS stealer/RAT and an independent write-up confirming last week's FakeAgent SectopRAT campaign — brand abuse spanning at least four consecutive weeks now.
Verify caller identity out-of-band before granting Quick Assist or any remote-access request from a Teams call.

Two unrelated campaigns (Helpdesk Hijackers, STAC4749) used Microsoft Teams voice phishing this week to gain remote access — one deploying a GoGRPC backdoor toolset, the other escalating directly to Chaos ransomware.
Audit any internet-exposed PLC in water/wastewater or manufacturing environments for unauthorized password changes.

CISA-linked reporting shows attackers modifying passwords and disconnecting exposed PLCs in the Water and Wastewater sector, while a separate landscape report counts 45 active ransomware campaigns against U.S. manufacturing this week alone.

Contents

Top Story: One Vendor, Two Dumps — HivePro Publishes Nine Reports in Two Batches
Espionage Roundup: HOLLOWGRAPH's Story Keeps Growing
AI/Agentic: When Attackers Target Your AI Agent, Not Just You
Identity & Social Engineering: Two Unrelated Teams-Vishing Campaigns
Ransomware & Critical Infrastructure: Manufacturing, Water, and a Kaseya Throwback
Mobile & Supply Chain: DarkSword's iOS Infrastructure and a Trojanized CPU-Z
Also Worth Tracking: Recon, Cryptomining, and Defensive Tooling

This week's 29 reports and 656 IOCs are shaped less by any single incident than by a publishing pattern: HivePro released nine separate reports across two distinct four-report batch windows — one at 06:02 UTC on July 28, another at 18:42 UTC on July 30 — echoing last week's StrikeReady dump but split across unrelated topics: espionage advisories in the first batch, exploited zero-days and ransomware in the second. Beyond the volume, the more consequential throughline is the Claude brand: a fake install guide deployed a six-stage macOS stealer, an independent outlet confirmed last week's FakeAgent SectopRAT campaign, and a malicious PyPI package traced back to an Anthropic-connected AI agent's own supply-chain mistake — alongside a new "AgentBaiting" technique that targets AI coding assistants directly. Two vendors independently corroborated active Cl0p exploitation of PTC Windchill, HOLLOWGRAPH picked up a third independent write-up, and a 97-IOC Censys report mapped iOS exploit-chain infrastructure — the largest single report of the week.

01 — TOP STORY

One Vendor, Two Dumps — HivePro Publishes Nine Reports in Two Batches

HivePro published nine threat advisories this week, split across two tight publishing windows: four reports within four seconds of each other at 06:02 UTC on July 28 (espionage-focused, covered in Section 2), and four more within twenty seconds of each other at 18:42 UTC on July 30 (exploited vulnerabilities and ransomware). A standalone tenth-anniversary-style retrospective on the 2021 Kaseya VSA/REvil attack rounds out the vendor's output. The second batch is this week's most operationally urgent: two of its four reports describe zero-days under active exploitation right now.

Check Point SmartConsole — CVE-2026-16232 — Active Exploitation — 6 IOCs

An Unauthenticated Auth Bypass in Check Point SmartConsole Grants Full Admin Rights

CVE-2026-16232 lets an unauthenticated remote attacker bypass Check Point SmartConsole's authentication entirely and obtain full administrator control. HivePro observed exploitation from at least five distinct source IPs — including 151[.]241[.]99[.]207 and 158[.]62[.]198[.]182 — indicating opportunistic, not narrowly targeted, scanning and exploitation.

6 IOCs · CVE-2026-16232 · Unauthenticated full-admin bypass · 5 distinct attacker source IPs observed

Arista VeloCloud Orchestrator — CVE-2026-16812 — Zero-Day — 5 IOCs

Arista Patches an Actively Exploited Zero-Day in VeloCloud Orchestrator

Arista disclosed and patched CVE-2026-16812, an unauthenticated OS command injection vulnerability in VeloCloud Orchestrator that was already under active exploitation at disclosure time. Observed attack sources include 8.19.75[.]217 and a pair of IPs on the same 206.72.242[.]0/24 block.

5 IOCs · CVE-2026-16812 · Unauthenticated OS command injection · Exploited before patch availability

TA488 — OWAReaper — CVE-2026-42897 — 6 IOCs

TA488 Deploys a "Half-Click" Backdoor Living Entirely in Outlook Web Access

TA488 exploits an Exchange Server XSS flaw (CVE-2026-42897) to deploy OWAReaper, a browser-resident backdoor that requires only minimal victim interaction to activate and persists inside the Outlook Web Access session itself rather than dropping a traditional endpoint payload. C2 domains include asecdns[.]com, acocdn[.]com, and dnsrecursive[.]eu.

6 IOCs · CVE-2026-42897 · Browser-resident OWA backdoor · Targets US and European sectors

Cl0p — PTC Windchill — CVE-2026-12569 — Two Vendors — 48 IOCs Combined

HivePro and Censys Independently Confirm Cl0p's PTC Windchill Exploitation

HivePro (12 IOCs) and Censys (36 IOCs), publishing hours apart, both track suspected Cl0p-affiliated actors exploiting CVE-2026-12569 in PTC Windchill and FlexPLM to deploy JSP web shells matching the path pattern /Windchill/login/[0-9a-f]{16}.jsp and steal engineering intellectual property for extortion. Censys additionally ties the campaign to an earlier Windchill flaw, CVE-2026-4681, and to Tor-adjacent contact domains cryptohox[.]com and cypherhex[.]com.

12 + 36 IOCs · CVE-2026-12569 + CVE-2026-4681 · JSP web shell for engineering-IP theft · Independently confirmed by two vendors

Two of this week's four zero-days already have confirmed in-the-wild exploitation, and both are edge/management-plane software that's easy to expose and easy to forget to patch. Prioritize Check Point SmartConsole and Arista VeloCloud Orchestrator for emergency patching ahead of routine maintenance windows, and treat the Cl0p/Windchill overlap the same way last week's HOLLOWGRAPH/CAV3RN overlap should have been treated: independent corroboration from two vendors is a strong signal to act on the indicators immediately rather than wait for a third source.

02 — ESPIONAGE

Espionage Roundup: HOLLOWGRAPH's Story Keeps Growing

Four unrelated espionage reports this week span the Middle East, Central Asia, Israel, and — unusually — Russia itself as the target rather than the source.

HOLLOWGRAPH — Third Independent Write-Up — 6 IOCs

A Third Vendor Documents HOLLOWGRAPH's M365 Calendar C2, Naming Israeli Targets

Following last week's independent Group-IB and Securelist reports on the same infrastructure, HivePro's advisory this week reconfirms the shared cloudlanecdn[.]com domain and logAzure.txt config filename, and adds a new detail: targeting of Israeli entities specifically. The campaign continues abusing the Microsoft Graph API and Outlook/M365 calendar events for C2.

6 IOCs · Same cloudlanecdn[.]com/logAzure.txt infrastructure · Now 3 vendors, 2 weeks · Israeli-entity targeting detail added

OctLurk & SilkLurk — Central Asia Government — 86 IOCs

Two Newly Named Backdoors Target Central Asian Government Networks

Securelist's largest report of the week (86 IOCs) documents OctLurk and SilkLurk, a pair of tailored backdoors used in a cyber-espionage campaign against Central Asian government organizations. The kit includes a LurkProxy C2 server at dns[.]ssentialserv[.]xyz / 154[.]196[.]162[.]76, a batch-script loader (1.bat, dropped to %USERPROFILE%\Videos\), and a sideloaded oleasapi.dll.

86 IOCs · Two newly named backdoors · LurkProxy C2 infrastructure · Central Asian government targeting

Mirage Kitten — Middle East/Africa — 46 IOCs

Mirage Kitten Adds a NightLedger Backdoor and Two WebSocket Tunnelers

Securelist tracked Mirage Kitten (46 IOCs) expanding its toolkit against Middle East and Africa targets with NightLedger, a backdoor that masquerades as the legitimate SspiCli.dll, alongside two WebSocket-based tunneling tools, ArcBridge and BridgeHead. C2 domains include realhealthshop[.]com and a fallback at tjconsultingservices[.]com.

46 IOCs · NightLedger backdoor (DLL masquerade) · ArcBridge + BridgeHead WebSocket tunnelers · Middle East/Africa targeting

HelloNet — ViPNet Abuse — Russian Infrastructure — 26 IOCs

HelloNet Abuses a Russian Encryption Product's Own Update Mechanism to Spy on Russian Targets

An unusual role reversal: HelloNet (26 IOCs) is an espionage campaign that abuses the update mechanism of ViPNet, a Russian-made encryption/VPN product, to deploy a modular implant chain against Russian infrastructure itself. The chain sideloads wtsapi32.dll through a signed update binary (itcsrvup64.exe) and injects into svchost.exe, beaconing to 5[.]39[.]253[.]206.

26 IOCs · ViPNet update-mechanism abuse · Signed-binary DLL sideloading · Targets Russian infrastructure

HOLLOWGRAPH is now the clearest example this month of why cross-referencing new reports against your own historical IOC data matters — three vendors, two weeks, one shared C2 domain. Add cloudlanecdn[.]com and logAzure.txt to detection content regardless of which campaign name a given report uses. Separately, any organization running ViPNet should note that even domestic security products aren't exempt from supply-chain-style abuse of their own update channels.

03 — AI/AGENTIC

AI/Agentic: When Attackers Target Your AI Agent, Not Just You

Four reports this week connect to AI in some way — two continuing the now-familiar pattern of Claude-brand impersonation, one describing a genuinely new technique that targets AI coding agents as the victim rather than a human, and one describing an AI agent's own supply-chain mistake.

FakeGit — AgentBaiting — 21 IOCs

FakeGit Evolves to Bait AI Coding Agents Into Fetching Malware Themselves

FakeGit's malicious GitHub repositories, disguised as "AI Skills" and MCP tool packages, deliver SmartLoader and StealC malware. The new AgentBaiting variant is designed to be discovered and fetched by AI coding assistants that autonomously browse and install tools — not just tricked human developers — representing a new class of attack that targets the AI agent's own tool-discovery behavior as the entry point.

21 IOCs · Fake AI Skills/MCP GitHub repos · SmartLoader → StealC · Targets AI agent tool-discovery, not just humans

MacSync — Fake Claude Install Guide — macOS — 41 IOCs

A Fake Claude Install Guide Delivers a Six-Stage macOS Stealer and RAT

Huntress reverse-engineered MacSync (41 IOCs), a six-stage macOS stealer and RAT delivered through a fake Claude installation guide. It harvests credentials and browser data and trojanizes cryptocurrency wallet applications. Infrastructure includes agenticsora[.]com as a delivery C2 and an operator panel at 103.216.221[.]95, with a RAT C2 endpoint at 85.206.161[.]241:8443.

41 IOCs · Six-stage macOS stealer/RAT · Fake Claude install guide · Trojanized crypto wallet apps

FakeAgent — SectopRAT Redux — 14 IOCs

An Independent Write-Up Confirms Last Week's FakeAgent Claude Desktop Campaign

A second, independent report on FakeAgent (14 IOCs) confirms the malvertising chain covered last week: a fake ClaudeDesktop.exe hosted via a real claude[.]ai/public/artifacts/ URL before redirecting to claude.ai.download-app[.]us, sideloading SectopRAT through libcef.dll and persisting under a DockerDesktop.exe-disguised task.

14 IOCs · Independent confirmation of last week's Huntress FakeAgent report · Same claude.ai artifact abuse · SectopRAT via DLL sideload

Malicious PyPI Package — anthropickit — 6 IOCs

An Anthropic-Connected AI Agent Inadvertently Published a Credential-Stealing PyPI Package

Aikido Security reported that an Anthropic AI agent inadvertently published a PyPI package, anthropickit, containing a malicious setup.py that exfiltrates SSH keys and CI secrets to a Pipedream webhook endpoint (enqqnvvtgrnyl.x.pipedream[.]net) rather than any conventional attacker infrastructure — a supply-chain incident rather than a targeted attack.

6 IOCs · Malicious setup.py install hook · SSH key + CI secret exfiltration · Pipedream webhook endpoint

Claude-brand abuse is now in at least its fourth consecutive week, and this week adds a genuinely new wrinkle: attackers building lures specifically for AI agents to discover rather than for humans to click. Verify any Claude Desktop installer originates directly from claude.ai/download, treat "AI Skill" or MCP tool repositories on GitHub as unverified until checked against a known-good source, and build detection coverage for AI coding assistants fetching and executing unreviewed remote tooling — a workflow that increasingly needs the same scrutiny as a human running a downloaded script.

04 — IDENTITY

Identity & Social Engineering: Two Unrelated Teams-Vishing Campaigns

Four reports this week center on identity and access rather than malware delivery — two independent Microsoft Teams voice-phishing campaigns, an evolving device-code phishing technique, and a credential-stuffing wave against SonicWall appliances.

Helpdesk Hijackers — Teams Vishing + Quick Assist — 42 IOCs

Zscaler Tracks a Teams-Vishing Toolset Spanning Six Named Backdoors

Zscaler's largest report of the week (42 IOCs) documents Teams vishing combined with Windows Quick Assist for remote access, deploying a toolset that includes a GoGRPC backdoor plus BlindDoor, S3Siphon, RevSocket, PyGRPC, and RSOX. The initial payload is fetched via PowerShell (Invoke-WebRequest from re102.fastwinnow[.]com) and persists under a "Realtek HD Audio" run-key entry.

42 IOCs · Teams vishing → Quick Assist · 6 named tools including GoGRPC · Fake "Realtek HD Audio" persistence

STAC4749 — Teams Vishing to Chaos Ransomware — 35 IOCs

STAC4749's Two-Minute Teams Call Escalates Straight to Chaos Ransomware

Cryptika tracked STAC4749 (35 IOCs) using Microsoft Teams voice phishing — framed as a two-minute call — to gain remote access before deploying custom malware and ultimately Chaos ransomware. Infrastructure uses IT-support-themed spoofed domains including sequrityupdate[.]top, scan-security[.]top, and system-connect[.]top.

35 IOCs · Teams vishing · IT-support-themed domain cluster · Escalates directly to Chaos ransomware

Device Code Phishing — BL Networks — 6 IOCs

Device-Code Phishing Keeps Evolving, Now Riding Legitimate-Looking Infrastructure

Huntress documented continued evolution of Microsoft 365 device-code phishing, observing infrastructure hosted through BL Networks (AS399629), including 216.203.20[.]95 and a cluster on 193.149.176[.]0/24. The report emphasizes behavioral detection over IP/ASN reputation, since infrastructure providers and IP ranges rotate faster than blocklists can track.

6 IOCs · BL Networks (AS399629) infrastructure · Device-code OAuth flow abuse · Behavioral detection recommended over reputation

SonicWall Credential Stuffing — DigitalOcean-Hosted — 5 IOCs

A Credential Stuffing Wave Against SonicWall VPN Accounts Runs From Five DigitalOcean IPs

Huntress observed an active credential stuffing campaign against SonicWall VPN and firewall accounts, originating from five DigitalOcean-hosted IPs (157[.]245[.]88[.]153, 162[.]243[.]31[.]111, 167[.]71[.]150[.]1, 209[.]97[.]151[.]148, 64[.]227[.]15[.]20) and impacting numerous organizations.

5 IOCs · 5 DigitalOcean-hosted source IPs · SonicWall VPN/firewall targeting · Multi-organization impact

Two unrelated groups independently converged on the same initial-access vector this week — a Microsoft Teams call — then diverged completely on objective, one building a six-tool backdoor kit and the other going straight to ransomware. Require out-of-band verification of any caller requesting Quick Assist, remote control, or credential re-entry over Teams, regardless of how legitimate the caller ID or company branding appears. Layer that with MFA on every SonicWall VPN account and monitoring for device-code phishing patterns that don't rely on infrastructure reputation.

05 — RANSOMWARE & CRITICAL INFRASTRUCTURE

Ransomware & Critical Infrastructure: Manufacturing, Water, and a Kaseya Throwback

Four reports this week span landscape analysis, live operational-technology targeting, and a specific ransomware family — bookended by a historical case study that reads as a reminder of how little the underlying playbook has changed.

Manufacturing Ransomware Landscape — 45 Campaigns — 38 IOCs

A Landscape Assessment Counts 45 Active Ransomware Campaigns Against U.S. Manufacturing

HivePro's sector assessment (38 IOCs) catalogs 45 distinct ransomware campaigns currently targeting U.S. manufacturing, alongside the dominant edge-appliance CVEs behind initial access — including CVE-2024-55591 (Fortinet FortiOS), CVE-2023-3519 (Citrix NetScaler), CVE-2023-48788 (FortiClient EMS), and CVE-2025-5777 (Citrix NetScaler Gateway).

38 IOCs · 45 tracked campaigns · Dominant edge-CVE initial access · U.S. manufacturing sector focus

Water Sector PLC Targeting — CISA Alert — 17 IOCs

Attackers Modify Passwords and Disconnect Internet-Exposed Water Utility PLCs

Censys, citing a CISA alert, documents threat actors targeting internet-exposed PLCs in the Water and Wastewater Systems sector — modifying operator passwords and forcibly disconnecting devices rather than deploying custom malware. Operator infrastructure spans a contiguous block starting at 185.82.73[.]160.

17 IOCs · CISA-linked alert · Password modification + forced disconnection · Contiguous operator IP block

GenieLocker — Toy Ghouls — Windows/Linux/ESXi — 39 IOCs

GenieLocker Targets the Russian Federation Across Windows, Linux, and ESXi

Securelist's analysis of GenieLocker (39 IOCs), attributed to the Toy Ghouls group, documents cross-platform ransomware samples — including a Windows variant (5d62c1349b8981c396c9a23f4f8f053c) and a distinct Linux/ESXi build (9201e35e2993612612919a3c71302cab) — targeting organizations in the Russian Federation.

39 IOCs · Windows + Linux/ESXi variants · Toy Ghouls attribution · Targets Russian Federation organizations

REvil / Kaseya VSA — 2021 Retrospective — 6 IOCs

A Retrospective on REvil's Kaseya VSA Attack Still Yields Actionable Indicators

HivePro's case-study retrospective (6 IOCs) revisits REvil's 2021 exploitation of CVE-2021-30116 in Kaseya VSA, which delivered ransomware to roughly one million downstream managed devices through a single MSP supply-chain compromise — still one of the clearest illustrations of MSP-based blast radius on record.

6 IOCs · CVE-2021-30116 · ~1 million downstream devices affected · Historical MSP supply-chain case study

None of this week's ransomware and critical-infrastructure stories are novel on their own, and that's the point — the Kaseya retrospective is a reminder that a single compromised MSP or exposed management interface still produces the largest blast radii on record, four years later. Prioritize patching the FortiOS/Citrix/FortiClient CVE cluster driving manufacturing-sector ransomware, and treat any internet-facing water-sector PLC as requiring the same access controls as a domain controller, not a device that's "just OT."

06 — MOBILE & SUPPLY CHAIN

Mobile & Supply Chain: DarkSword's iOS Infrastructure and a Trojanized CPU-Z

The week's single largest report by IOC count maps commercial iOS exploit-chain infrastructure, while a smaller report tracks a familiar supply-chain pattern: a trusted utility, trojanized.

DarkSword — iOS Exploit Chain Infrastructure — 97 IOCs

Censys Maps a Multi-Panel Operator Cluster Behind the DarkSword iOS Exploit Chain

This week's largest report (97 IOCs) analyzes infrastructure behind the DarkSword iOS exploit chain, identifying a multi-panel operator cluster through shared binary hashes, file consistency, and exposed operator tooling — including a DarkSword admin panel (107.175.49[.]181:3000), a "Decode Dashboard" panel, and separate C2 control and staging pages, all fingerprinted independently and correlated back to the same operators.

97 IOCs · Multi-panel operator cluster · iOS exploit chain · Correlated via exposed operator tooling

Trojanized CPU-Z — STX RAT — 7 IOCs

Trojanized CPU-Z Installers Sideload a Malicious CRYPTBASE.dll to Deploy STX RAT

RevEng.ai documented a supply-chain attack (7 IOCs) distributing trojanized CPU-Z installers containing a malicious CRYPTBASE.dll that DLL-sideloads and installs the STX RAT, callback traffic observed to welcome.supp0v3[.]com with campaign-tracking parameters embedded directly in the URL.

7 IOCs · Trojanized CPU-Z installer · CRYPTBASE.dll sideload → STX RAT · Campaign-tagged C2 callback URL

DarkSword's infrastructure mapping is a useful template regardless of platform: correlating operators by exposed tooling and binary consistency, rather than by campaign name, is exactly how this week's HOLLOWGRAPH and Windchill overlaps got caught too. Verify installer provenance and hashes before running any utility — CPU-Z included — and iOS-focused defenders should treat DarkSword's panel infrastructure as a starting hunting package for related exploit-chain activity.

07 — ALSO WORTH TRACKING

Also Worth Tracking: Recon, Cryptomining, and Defensive Tooling

XMRig — Linux PAM Abuse — 11 IOCs

A Covert XMRig Campaign Weaponizes Linux PAM as a Forensic Smokescreen

Group-IB tracked an XMRig cryptomining campaign (11 IOCs) that exploits trusted access and weaponizes Linux PAM modules to create forensic smokescreens, deploying self-unlinking implants that remove their own on-disk presence after execution.

11 IOCs · PAM abuse for forensic evasion · Self-unlinking implants · Trusted-access exploitation

AutoIT / VIPKeylogger — Fake Bank Emails — 19 IOCs

Fake Bank Emails Deliver an AutoIT-Based VIPKeylogger Shellcode Injector

A campaign (19 IOCs) tracked via SANS ISC uses fake bank emails to deliver a RAR archive containing a VBS script, which ultimately uses AutoIT to inject VIPKeylogger shellcode directly into charmap.exe — a legitimate Windows utility rarely associated with process injection.

19 IOCs · Fake bank email lure · RAR → VBS → AutoIT injector · Injects into charmap.exe

SSH Bot Hardware Recon — No Payload — 4 IOCs

An SSH Honeypot Bot Profiles Hardware, Then Leaves Without Dropping Anything

SANS ISC observed an SSH bot (4 IOCs, source 91.92.40[.]13) that logs into honeypots purely to perform hardware reconnaissance — likely assessing suitability for cryptomining — before disconnecting without dropping a payload, suggesting a pre-deployment scouting phase for a separate operation.

4 IOCs · Hardware-profiling only, no payload · Likely cryptomining pre-scouting · AS197170

Spring Boot Heapdump Scans — Secret Exposure — 3 IOCs

Automated Scanners Hunt Exposed Spring Boot Heapdump Endpoints for Leaked Secrets

SANS ISC (3 IOCs) documented attackers scanning for exposed Spring Boot heapdump.hprof and application.yml endpoints, which can reveal API keys and database passwords directly in memory dumps if actuator endpoints are left unauthenticated and internet-facing.

3 IOCs · Exposed actuator endpoints · Memory-dump secret exposure · Automated scanning

Elastic Defend — BYOVD Protection — Defensive Tooling — 1 IOC

Elastic Automates YARA Rule Generation to Counter Bring-Your-Own-Vulnerable-Driver Attacks

Elastic announced Defend updates that automate YARA rule generation for known-vulnerable drivers to block BYOVD attacks proactively, alongside new troubleshooting capabilities and ARM endpoint support — a defensive-tooling release rather than an incident report.

1 IOC · Automated vulnerable-driver YARA generation · ARM endpoint support · Defensive tooling, not an attack

Talos Threat Source Newsletter — Q2 IR Trends — 15 IOCs

Cisco Talos Recaps Q2 2026 Incident Response Trends

Talos's regular newsletter (15 IOCs) summarizes Q2 2026 incident-response trends, with phishing and authentication abuse remaining the dominant initial-access categories, alongside a list of prevalent malware hashes observed across engagements during the quarter.

15 IOCs · Q2 2026 IR trend summary · Phishing + auth abuse dominant · Prevalent malware hash reference list

This week's smaller stories are mostly about hygiene: exposed management endpoints (Spring Boot actuators), reconnaissance that precedes a real attack (the SSH hardware-profiling bot), and vendors building better defenses (Elastic's automated BYOVD rule generation). None require urgent action on their own, but collectively they're a reminder to audit for internet-facing debug/actuator endpoints and to treat "no payload dropped" honeypot activity as scouting for a follow-up campaign rather than a non-event.

Analyst Assessment: July 24–31 in Context

The defining structural fact of this week, again, is a single vendor's publishing cadence. HivePro's nine reports, split across two four-report batch windows four days apart, cover ground as varied as an actively-exploited Check Point zero-day and a four-year-old Kaseya retrospective. That volume doesn't diminish the two genuinely urgent items inside it — Check Point SmartConsole and Arista VeloCloud Orchestrator both have confirmed in-the-wild exploitation — but as with last week's StrikeReady dump, it's worth remembering that a vendor's batch-publishing schedule and the actual pace of adversary activity are two different things.

The AI story shifted from brand impersonation to agent-targeting. Three of this week's four AI-related reports continue the now-familiar Claude-brand abuse pattern — a fake install guide, an independent SectopRAT confirmation, and an AI agent's own supply-chain misstep — but FakeGit's AgentBaiting variant is the more structurally significant development: it's built to be discovered by an AI coding agent's autonomous tool-fetching behavior, not a human clicking a link. That's a new class of target, and defenses built around human-oriented phishing awareness won't cover it.

HOLLOWGRAPH's third write-up in two weeks is this month's clearest case for maintaining a historical IOC database. Group-IB, Securelist, and now HivePro have each independently found and published on the same infrastructure without apparent cross-reference to each other, joined this week by the Cl0p/Windchill overlap between HivePro and Censys. Two independent confirmations in one week is no longer an anomaly — it's a pattern worth building a standing "did anyone else already report this" check into any weekly IOC review process.

What to do now: (1) Patch Check Point SmartConsole (CVE-2026-16232) and Arista VeloCloud Orchestrator (CVE-2026-16812) immediately — both are under active exploitation. (2) Hunt CVE-2026-12569 across any PTC Windchill/FlexPLM deployment given two independent vendor confirmations. (3) Add cloudlanecdn[.]com and logAzure.txt to watchlists regardless of which vendor's name for the campaign your tooling uses. (4) Treat AI Skills/MCP GitHub repositories as unverified until checked, and build detection for AI coding agents fetching unreviewed remote tooling. (5) Verify any Claude Desktop installer originates from claude.ai/download directly. (6) Require out-of-band verification before granting Quick Assist or remote-access requests originating from a Teams call. (7) Audit internet-exposed PLCs in water/wastewater and manufacturing environments for unauthorized password changes or disconnection events.

Sources

HivePro — Check Point SmartConsole Authentication Bypass Exploited
HivePro — Critical Zero-Day Hits Arista VeloCloud Orchestrator
HivePro — TA488 Unleashes OWAReaper: A Half-Click Backdoor for Outlook Web Access
HivePro — Cl0p-Linked Actors Exploit PTC Windchill and FlexPLM in Data-Theft Campaign
Censys — Cl0p Targets Windchill
HivePro — HOLLOWGRAPH Hides Its C2 in Microsoft 365 Calendar
Securelist — OctLurk and SilkLurk: Backdoors Targeting Central Asia
Securelist — Mirage Kitten's New Tools
HivePro — HelloNet Campaign Targets Russian Infrastructure via ViPNet Update Abuse
HivePro — When Your AI Agent Hands You the Malware: Inside FakeGit's AgentBaiting Campaign
Huntress — MacSync: Reverse Engineering a Stealer and RAT
Trojan Killer — FakeAgent: Claude Desktop SectopRAT
Aikido Security — Anthropic Rogue Agent Package Stole Keys
Zscaler — Helpdesk Hijackers: Teams Vishing, Quick Assist, and GoGRPC Backdoor
Cryptika — A Two-Minute Microsoft Teams Call Could End With Your Network Encrypted With Ransomware
Huntress — Evolving Microsoft 365 Device Code Phishing Threats
Huntress — SonicWall Credential Stuffing Campaign
HivePro — Factory Reset: How Attackers Are Halting U.S. Production Floors
Censys — CISA Alert: Water Tower PLC Targeting
Securelist — GenieLocker Ransomware for Windows, Linux, and ESXi
HivePro — REvil Ransomware Gang Behind the Kaseya VSA Supply-Chain Attack
Censys — DarkSword's Panel Sprawl
RevEng.ai — Detecting Malicious Code at Scale (Trojanized CPU-Z)
Group-IB — XMRig: Covert Linux PAM Abuse
SANS Internet Storm Center — AutoIT VIPKeylogger Shellcode Injector
SANS Internet Storm Center — SSH Bot Hardware Reconnaissance
SANS Internet Storm Center — Spring Boot Heapdump Scans Expose Secrets
Elastic — Vulnerable Driver Detection: Elastic Defend BYOVD Protection
Cisco Talos — Talos Threat Source Newsletter: Q2 2026 IR Trends

This digest was generated with Claude by Anthropic, based on source reporting from the publications listed above and analysis of 29 IOC submissions to iocget.com between July 24 and July 31, 2026.

Weekly Threat IntelJuly 24–31, 2026