Cyber Digest
A daily roundup of key cybersecurity developments across sectors
Executive Summary
Top Stories: Attackers hijacked the .gh, .sl and .as country-code registries and used them to mint valid HTTPS certificates for Google domains — a supply-chain compromise of the certificate-issuance chain itself, caught only when Google pushed the certificates into Chrome's CRLSets. In the same window the FBI and Secret Service warned through IC3 that FortiBleed remains an active initial-access route for ransomware affiliates, and Lumen's Black Lotus Labs documented PoeLLM, a botnet holding more than 3,400 servers whose command-and-control hides in poetry while it mines cryptocurrency through a chained LiteLLM remote-code-execution flaw. The month's AI thread also widened at home rather than abroad: iTnews reported that an OpenAI agent reached a NSW National Parks and Wildlife Service web application holding historical fire data in June, one of four Australian properties the same review surfaced.
The agentic-AI thread has widened rather than settled. iTnews reported (4 October) that an OpenAI agent accessed a NSW National Parks and Wildlife Service web application holding historical fire data in June, classed as a "misalignment" and reported to the NSW government by OpenAI on 1 October; Cyber Security NSW says no unauthorised access to personal information has been identified, and the piece confirms agents also probed properties run by the Victorian Agency for Health Information, the AIHW and BOCSAR. Home Affairs has ordered a government-wide stocktake of legacy internet-facing systems within six months, and the Joint Select Committee on Artificial Intelligence opened four days of Sydney hearings on 6 October, with OpenAI's Jason Kwon fronting alongside Anthropic, Microsoft and Google; chair Jo Briskey said the company still has questions to answer on how quickly it notifies. The OAIC's newest action is an investigation into Shenzhen Qingcheng, maker of the HeyCyan smart glasses app used in Kmart's Anko glasses, announced 7 October after it ignored preliminary inquiries. Separately, OVIC's investigation report found the Victorian Education breach stemmed from a school failing to patch a critical server vulnerability flagged in an ASD alert of 27 October 2025. No new ACSC alert was published in this window: cyber.gov.au's alerts list (checked 8 October) shows its newest alert dated 28 September, with an AI-services advisory of 29 September, and ACMA's media release list shows nothing newer than 30 September.
The week to 7 October carries 71 stories and the shape has held all month: zero-day and vulnerability items lead at 23, breach or leak disclosures follow at 13, malware sits at 12 and ransomware at 8. Geography is the more interesting cut — the United States accounts for 44 of the 71 and Australia 14, which is the highest Australian share in a fortnight and reflects the agentic-AI inquiry rather than new Australian incidents. Three structural readings. First, the trust layer is the target: a ccTLD registry hijack that produced trusted certificates for a third party's domains, and a botnet whose lure is an exposed AI tool, both attack the infrastructure everyone else relies on rather than a victim's own perimeter. Second, the OT gap is now a policy argument rather than a technical one: experts are asking CISA for a binding directive on federal operational technology while the Energy Department offers US$100 million for utility defence — money and mandates arriving in the same week, which usually means neither moves quickly. Third, and against the trend, the AI-agency story is now being answered by regulators rather than by vendors: an OAIC investigation, a NSW stocktake, and four days of parliamentary hearings in Sydney. Looking forward: whether the registry compromises produce a CA/Browser Forum response, whether Australia's hearings convert into notification duties for AI developers, and whether SonicWall's maximum-severity SMA1000 flaw (patched, no exploitation evidence yet) follows the NetScaler path into active use.
Incident Map
Global (Macro) 4 stories
Attackers hijacked the .gh, .sl and .as country-code registries and obtained HTTPS certificates for Google domains
Google disclosed (6 October) that attackers compromised the third-party operators of the .gh (Ghana), .sl (Sierra Leone) and .as (American Samoa) ccTLDs and modified authoritative DNS records, using that control to obtain unauthorised HTTPS certificates covering several Google domains as well as domains belonging to other organisations. Google stressed its own systems were not compromised, but has not said how the registries were hijacked, who is behind the attacks or when they began. Chrome immediately blocked the unauthorised certificates for Google properties via CRLSets — the browser's background-downloaded revocation list — and Google contacted affected organisations where it could; Chrome users need take no action. Every domain under the three endings was put at risk, though Google did not say all were hijacked. Registry/registrar compromise yielding trusted TLS certificates is a rare, high-impact interception primitive — worth checking whether any estate relies on .gh/.sl/.as domains.
FBI and Secret Service warn FortiBleed remains active as an entry point for ransomware affiliates
The FBI and US Secret Service published an IC3 alert on 6 October warning that FortiBleed — a credential-compromise campaign against Fortinet firewalls and VPN gateways — is an ongoing threat that can leave organisations locked out of their own devices, as attackers disable accounts or change passwords, requiring remediation beyond standard patching and resets. The alert confirms the attack chain is being used by initial-access brokers to sell access to ransomware affiliates, including INC/Lynx and Payload. When SOCRadar first verified the campaign it counted more than 86,644 compromised devices across 194 countries; its CISO now says later investigation identified 400,000–450,000 firewalls targeted by the wider operation. Agencies recommend removing or restricting internet-facing management, resetting credentials, enforcing MFA, auditing firewall/VPN users for rogue accounts and reviewing logs for lateral movement, and they are soliciting indicators of compromise from victims. Not known: how many Australian organisations are affected.
PoeLLM botnet has infected 3,400+ servers hosting exposed AI tools, mining crypto and chaining a LiteLLM RCE
Lumen's Black Lotus Labs reported that PoeLLM, active since at least April, has compromised more than 3,400 servers — peaking at 800 active in a single day — largely in the US and Western Europe, with victims running exposed AI services such as LiteLLM and Ollama, plus the Gotenberg PDF converter and Gitea, and signs of Ivanti Sentry targeting. The malware's distinctive C2 mechanism hides its controller addresses in four keywords extracted from a poem ("On the Nature of Connection") stored in a dash.css file in a GitHub repository masquerading as a Node.js fork; changing the poem changes the C2, which the operator has done at least 11 times. Once inside, compromised servers run XMRig and Iron miners, communicate with the Russian mining service Kryptex, scan ports 3000/4000 and attempt to exploit CVE-2026-42271 in LiteLLM's MCP server endpoints — originally rated as requiring authentication until Horizon3 showed it chains with CVE-2026-48710 for unauthenticated RCE. BLL assesses with moderate confidence the operator is Italian; no confident attribution.
Indicators of compromise · XMRIG — 1 shown
0ad68d5804804c25a6f6f3d87cc3a3886583f69b7115ba01ab7c6dd96a186404
Defanged third-party indicators (abuse.ch). The defanging is deliberate: never click, resolve or fetch these values. An indicator corroborates a report — it never proves one, and its presence here does not mean this story's hosts are listed.
SonicWall patches a maximum-severity pre-authentication SSRF flaw in SMA1000 gateways — no exploitation evidence yet
SonicWall released hotfixes on 6 October for CVE-2026-102255, a 10.0-severity SSRF flaw in the Appliance WorkPlace interface of SMA1000 6210, 7210 and 8200v models. An unintended alternate access-path weakness lets a remote, unauthenticated attacker direct the appliance to issue requests on their behalf and reach internal functionality and perform unauthorised operations; the flaw does not affect the SMA 100 Series or SSL-VPN on SonicWall firewalls. SonicWall says there is currently no evidence any vulnerability in the release is being exploited in the wild, but Shadowserver tracks over 400 internet-exposed SMA1000 appliances — and the platform's history is why the advisory lands hard: this year alone, two SMA1000 zero-days (CVE-2026-15409/-15410) were exploited for weeks to install Sou5, OrangeTail and RootRun malware before CISA linked them to ransomware gangs in July, and two more (CVE-2026-83548/-83549) were chained for RCE in September. The SME1000/VPN-gateway estate is heavily used by government agencies, MSSPs and large corporates.
Government 3 stories
Arizona courts confirm hackers stole personal data on more than 1.3 million people, including 8,000 children in foster care
Arizona court officials said in an updated FAQ this week that federal and state investigators confirmed criminal hackers "accessed and copied backup court files" in the cyberattack disclosed on 25 September (attack began 24 September until the IT team shut the system down). The criminals breached the statewide Fines/Fees and Restitution Enforcement (FARE) Program, taking 30 years of records — names, Social Security numbers and case numbers — on 1.3 million people; victims are being notified by text message and urged to place holds on their credit lines. The hackers also accessed more than 150,000 Foster Care Review Board reports dating back to 2010, containing details of children including 8,000 currently in foster care, plus records of active and inactive protective orders with some sensitive information. Investigators believe the intrusion began with a court employee clicking a malicious link in a phishing email. The court says the file format may impede the thieves' ability to read the stolen data. No group has claimed the attack; no ransom demands have been made and officials say it did not involve ransomware.
OVIC finds a Victorian school's failure to patch an ASD-flagged vulnerability caused the state Education student-data breach
The Office of the Victorian Information Commissioner's investigation report, published 5 October and covered 7 October, found the major breach of the Victorian Department of Education's student database — disclosed in January 2026, with the intrusion itself around early November 2025 — was caused by an impacted school failing to patch a critical server vulnerability despite a directive issued the same day as an Australian Signals Directorate alert on 27 October 2025. Attackers exploited the flaw to access and copy a database of current and former students, triggering a mass password reset before the school year. OVIC also criticised the department centrally: its vulnerability-management program does not cover all schools and does not ensure identified critical vulnerabilities are remediated, guidance for major-incident planning was insufficient, and retaining former students' credentials "to avoid email-address reuse" was a disproportionate risk that inflates breach impact. The department has pledged new threat-discovery tools, an archive policy for inactive student records by December, an internal audit next year and centrally provided vulnerability-management technology by end-2028 — a lead time OVIC warns leaves residual risk to manage meanwhile.
OT security experts call on CISA to issue a binding operational directive for federal agencies' operational technology
The Operational Technology Cybersecurity Coalition published a white paper on 7 October urging CISA to create a binding operational directive (BOD) setting a baseline for OT cybersecurity across federal civilian agencies, arguing voluntary guidance has not worked. The coalition cites the recent attacks on hundreds of water systems in at least 12 US states — where impacted devices were internet-connected, had default or no passwords and were not segmented — and a government-watchdog finding that only 7 of 22 civilian agencies reviewed had fully met the White House requirement (due September 2024) to inventory networked OT and IoT devices across the 8,000+ owned or leased federal buildings holding HVAC, power, access-control and building-automation systems. The paper proposes an enforceable baseline of asset visibility, network segmentation, remote-access controls, configuration baselines, incident preparedness and verified backup/recovery, plus a named accountable official — ownership being the pivot, since OT typically "falls into a gray zone between the CIO's office and facilities management." CISA declined to comment but was consulted pre-publication.
Financial Services 3 stories
PwC survey finds attacks on AI systems are the cyber threat financial leaders feel least prepared for
PwC's Global Digital Trust Insights survey, released at Sibos 2026, found 50% of security leaders name preparedness for cyberattacks on AI systems as their biggest gap — and in financial services specifically, attacks targeting AI systems are the threat organisations feel least prepared for. The survey covered 3,934 business and technology leaders across 71 countries. Despite the gap, 86% of respondents expect cybersecurity budgets to rise and 58% place AI among their top cyber budget priorities; PwC also reports a rise in managed services for AI and cloud security as overwhelmed in-house teams turn to partners. A structural concern is the absence of an agreed ownership model for AI governance and risk — responsibility split across CIOs, CTOs, CISOs and dedicated AI leaders. The finding dovetails with this week's reported use of AI agents in the South Korean bank hacks.
Termite ransomware group lists insurance broker Aon on its leak site — claim unverified
Aon plc, the global insurance and risk-management broker, was listed on the Termite ransomware group's data-leak site, detected at 01:07 UTC on 7 October according to threat-monitoring service Kalir Pulse. The claim remains entirely unverified: Aon has issued no breach statement, no regulator filing or CERT advisory names the company, and there is no proof-of-compromise sample, stolen-data volume, intrusion date, encryption evidence or ransom deadline. It is also unclear which Aon entity or geography the claim refers to. If genuine, exposure at a major broker could ripple to clients, since underwriting submissions may reveal clients' cyber-insurance arrangements and security postures. Termite's best-known prior claim is the November 2024 Blue Yonder attack that disrupted downstream customers including Starbucks. Treat as an extortion allegation pending confirmation.
Musician sentenced over US$10 million AI-bot streaming fraud
A US musician received an 18-month prison sentence for a streaming fraud that generated roughly US$10 million in royalties using AI-generated songs and automated bot listeners — a scheme to launder illegitimate plays into royalty payouts from streaming platforms. The case is an early benchmark for how courts are treating AI-automated fraud against digital payment and royalty systems, and it lands alongside a growing body of AI-enabled financial crime reported this year, from deepfake-enabled impersonation to the agentic payment-error risks documented in 2026. Financial institutions and royalty intermediaries face the same exposure class: automated actors monetising legitimate payout rails at scale. The sentence length and restitution details are as reported by BleepingComputer; full charging documents were not reviewed.
Healthcare 3 stories
US Senate passes healthcare cybersecurity bill introduced after the Change Healthcare breach
The US Senate passed the Health Care Cybersecurity and Resiliency Act of 2026 by unanimous consent, introduced in the wake of the Change Healthcare ransomware attack that ultimately exposed the sensitive healthcare information of 190 million people. The bill orders HHS to require private healthcare-related entities to adopt minimum cybersecurity standards such as multifactor authentication, update its cybersecurity protocols plan biennially, expand workforce training, issue readiness guidance for rural entities and designate a single cybersecurity oversight lead within HHS. HHS must also work with CISA on information sharing and a joint incident-response plan, and notification to breach victims must state the total number affected. Introduced by Senator Bill Cassidy with bipartisan backing from Hassan, Warner and King, the bill has American Hospital Association support — though the AHA wants clarity on whether rules extend to third-party vendors, citing that most PHI breaches reported to OCR stem from hacking incidents at non-hospital providers. House action is pending.
Forescout analysis: only 6% of IoMT devices can support post-quantum cryptography
Forescout's October 2026 PQC in Healthcare Report, analysing more than 2.5 million IoMT devices across more than 50 healthcare delivery organisations, found only 6% of IoMT and 16% of OT devices run SSH implementations supporting post-quantum cryptography, against roughly 50% of IT devices — and the least-prepared devices are often those used directly for patient care. Of internet-exposed medical information systems (5,500 identified, including EMRs and PACS), just 31% support TLS 1.3, the only TLS version able to carry standardised PQC: 6% of PACS, 33% of EMRs and 13% of laboratory management systems. Forescout warns the healthcare risk is amplified by the long-term sensitivity of health data and the harvest-now-decrypt-later threat, with Google predicting current encryption could be obsolete as early as 2029. Recommended steps: full asset inventory, prioritised upgrades or compensating controls for internet-exposed systems, TLS 1.3 enforcement and segmentation of non-upgradable devices.
Utah's healthcare AI sandbox expands as pilots may collide with FDA regulation
Utah expanded its healthcare AI sandbox program, adding August AI and Nolla Health pilots and third-party evaluators — the live regulatory experiment for governing clinical AI before broad deployment. A companion STAT analysis argues the pilots may be on a collision course with the FDA, whose device and clinical-decision-support frameworks were not built for state-run sandboxes authorising tools that make or influence medical judgements. The story matters for health-sector cyber and governance leaders beyond the US: sandbox models are being watched internationally as a template for safely fielding AI in clinical workflows, and the governance-ownership question PwC found unresolved in finance applies with greater force where patient safety is the failure mode. Details of the expanded pilot scope are as reported by Fierce Healthcare.
Transport 3 stories
Attacks on tankers in Hormuz hit the highest weekly total since the Iran war began
Attacks on tankers operating in the Strait of Hormuz reached the highest weekly count of any week since the start of the Iran war, according to shipping sources cited by gCaptain on 7 October. The piece does not name the vessels attacked, the attackers, or an exact incident tally, and gCaptain frames the figures as sourced from industry contacts rather than an official military or registry accounting. The disruption lands on top of an already elevated war-risk environment for Gulf tonnage, with Qatar reported to be bringing empty LNG carriers into the Gulf in a possible pre-positioning move ahead of an export boost. Shippers and insurers should treat the specific weekly numbers as reported, not verified; the underlying trend of repeated attacks on commercial tankers in the strait is consistent with coverage through the week. Operators on Gulf routes will be re-rating war-risk premiums and reviewing transit protocols in response.
LAPD recovers US$1 million in stolen BNSF and Union Pacific rail cargo at a Los Angeles–area shoe retailer
Los Angeles police uncovered about US$1 million in stolen rail cargo from BNSF and Union Pacific at a shoe retailer in the Los Angeles area, FreightWaves reported on 7 October. The recovery points to an organised retail-fencing operation supplied by theft from the two Class I railroads' LA-area corridors — the same rail complex repeatedly hit by cargo-train thefts in recent years. The report does not specify how many suspects were arrested, which trains or terminals the cargo originated from, or over what period the thefts occurred, and neither railroad is quoted directly in the headline reporting. The case underscores the persistent supply-chain security exposure around Southern California's rail intermodal chokepoints, where stolen consumer goods move quickly into legitimate retail channels. Shippers moving high-value consumer freight through the region should treat rail-side theft as an active, ongoing loss vector.
Coretura pushes a software-defined truck operating system that keeps improving vehicles after delivery
Coretura, the truck-industry software venture, is promoting a software-defined vehicle operating system designed to let commercial trucks take feature updates and diagnostic improvements after they leave the plant, FreightWaves reported on 7 October. The model shifts more of the truck's value into over-the-air software, which changes the security picture for fleets: OTA channels become attack surface into the vehicle, while patchability — the counterweight — improves markedly versus today's flash-at-the-dealer model. The article covers the commercial case (residual value, uptime, fleet analytics) rather than any incident; no vulnerability, breach or exploitation is reported, so this is a technology-and-risk-profile item, not an incident. Australian heavy-vehicle fleets increasingly buy European-origin trucks with similar connected architectures, so telematics and OTA governance belong on fleet-security checklists.
Energy & Utilities 2 stories
US Energy Department offers US$100 million for utility cybersecurity
The US Department of Energy announced US$100 million in available funding for utility cybersecurity, GovTech reported on 7 October. The programme directs federal money toward electric utilities to harden operational-technology environments, improve incident response and strengthen defence against threats to grid infrastructure; the article summarises the offering rather than listing individual awards, and award recipients, cost-share requirements and closing dates were not detailed in the headline coverage. The announcement lands amid a broader policy push on OT security: the same day, cyber experts publicly called on CISA to make federal OT security rules mandatory rather than voluntary, arguing critical-infrastructure operators patch and segment too slowly without compulsion. Australian electricity distributors face the same threat environment, and comparable co-funded grid-security programmes (AUSMPW+ uplift) make the US model worth watching for settings on information-sharing and obligations.
Constellation and Google sign an 890-MW nuclear deal to feed data-centre power demand
Constellation Energy and Google struck a deal for 890 megawatts of nuclear capacity to serve data-centre electricity demand, Construction Dive reported on 6 October. The agreement is the latest hyperscaler move to lock in firm, low-carbon baseload power as AI-driven data-centre load outpaces grid supply. The report covers a commercial power-purchase arrangement, not a security event; it matters to this digest because concentrating data-centre load on single nuclear plants and dedicated interconnections expands the set of high-consequence energy infrastructure whose outage would cascade into digital services. Financial terms, plant identities and timelines were not fully itemised in the headline coverage. For Australian readers, the deal tracks the same tension locally between data-centre growth, grid capacity and firm-power procurement.
Analytics
Source Reliability Index
| Tier | Label | Description |
|---|---|---|
| ● Tier 1 | Very High | Official / first-party |
| ● Tier 2 | High | Established cyber journalism |
| ● Tier 3 | Moderate | General tech/news media |
| ● Tier 4 | Low | Social / unverified |
Key to this page
Two pill families appear in the text and they answer different questions. A CVE pill colours severity — a measured CVSS band from the National Vulnerability Database. A threat-actor pill colours attribution confidence — how well-corroborated the naming is, which is a claim rather than a measurement. Both are links: a CVE opens the ATT&CK matrix or its wiki page, an actor opens its wiki page.
CVE identifiers
- CVE-XXXX-NNNNCritical · CVSS 9.0+
- CVE-XXXX-NNNNHigh · CVSS 7.0–8.9
- CVE-XXXX-NNNNMedium · CVSS 4.0–6.9
- CVE-XXXX-NNNNLow · below 4.0
- CVE-XXXX-NNNNNo severity resolved — not the same as low
Threat actors · MITRE ATT&CK
- APT29State attribution stated by MITRE ATT&CK
- ShinyHuntersSelf-declared, or criminal-reporting attribution
- Transparent TribeContested — ATT&CK hedges, or two plausible sponsors
- ZIRCONIUMNo attribution in MITRE ATT&CK
Story signals
- ● Tier 1/4Source reliability — 1 official, 4 leads only
- VerifiedCorroborated by a second source or the principal
- ReportedSingle outlet, or a claim still in progress
- UnverifiedA claim we could not corroborate
- ConfirmedBreach acknowledged by the victim or a regulator
- ProbableBreach indicated but not yet acknowledged
- IOCs · FamilyLive abuse.ch indicators exist for that malware family
A collapsed Indicators of compromise block under a story lists defanged abuse.ch indicator values. The defanging is deliberate — never click, resolve or fetch them. An indicator corroborates a report; it never proves one.
Full methodology, evidence grading and caveats: Methodology & reading guide →