An actively exploited Cisco SD-WAN authentication bypass joined CISA's known exploited list, NetScaler zero-day attacks were tied to custom web shells and weeks of dwell time, and a Zimbra command injection flaw was weaponised against internet-facing mail servers.
🔴 CRITICAL (2)
Cisco Catalyst SD-WAN Manager zero-day CVE-2026-76504 exploited, added to CISA KEV Confirmed
Cisco patched a critical authentication bypass tracked as CVE-2026-76504 in Catalyst SD-WAN Manager, the system organisations use to manage Cisco SD-WAN networks. A remote attacker with no login can reach the management API as an administrator because of improper handling of URI encoding, and the flaw carries a CVSS score of 9.8. Cisco says the flaw is being actively exploited, CISA added it to the Known Exploited Vulnerabilities catalog on 30 September with a federal remediation deadline of 3 October, and there is no workaround, so self-managed deployments must move to a fixed release.
Sources CISA|Cisco|BleepingComputer|The Hacker News
NetScaler zero-day attacks used WHIPSHOT web shell and SLAPSHOT tunneler for weeks of access Confirmed
Mandiant and Google Threat Intelligence Group published details of attacks exploiting the Citrix NetScaler flaws CVE-2026-88771 and CVE-2026-88772, which CISA added to its known exploited list on 27 September, against government, financial services, technology, education and legal organisations in North America and Europe. Attackers gained unauthenticated remote code execution and root access on ADC and Gateway appliances, installed a previously unreported PHP web shell called WHIPSHOT and a Python tunneler called SLAPSHOT, changed configuration to escalate privileges and opened covert channels into internal networks, with activity tracked over weeks. The flaws are rated up to CVSS 9.5 and fixes are available, so internet-facing appliances should be updated or taken offline, then examined for web shells, unexpected files in web directories and outbound tunnelling.
Sources Unit 42|The Hacker News|SecurityWeek|The Hacker News
🟠 HIGH (6)
Zimbra CVE-2026-73570 exploited to plant web shells on internet-facing mail servers Confirmed
Microsoft Threat Intelligence documented exploitation of CVE-2026-73570, an unauthenticated OS command injection flaw with a CVSS score of 8.9 in Zimbra Collaboration Suite, reachable through a crafted email that targets the SNMP notification path. Exploitation requires the optional zimbra-snmp package to be installed with SNMP notifications enabled, which limits exposure but still covers many internet-facing mail servers, and CERT Polska has also reported attacks. Intruders used the flaw to deploy web shells and reach mailbox data and authentication secrets, and Zimbra fixed it in version 10.1.20, so mail administrators should patch, check whether zimbra-snmp is installed and hunt for unexpected files in web directories.
Sources Microsoft|The Hacker News|Ars Technica
Bitget attributes USD 387.5 million theft to a third-party zero-day as stolen funds move into Zcash Confirmed
Bitget said the attackers who took about USD 387.5 million on 24 September got in through a zero-day in third-party security products, confirming that the exchange was not the initial target. On-chain reporting shows the attacker has started moving stolen funds into Zcash’s shielded Ironwood pool, roughly USD 3.8 million to USD 4 million in ZEC, after the cross-chain service Near Intents rejected more than USD 50 million in swap attempts. The incident is a reminder to treat third-party security tooling as part of the attack surface and to watch privacy-coin shielding for stolen assets.
Sources BleepingComputer|Decrypt|CoinDesk
South Africa air traffic agency seeks help after ransomware reached an operational network Under investigation
Air Traffic and Navigation Services, which manages South African airspace, confirmed a ransomware toolkit on at least one operational network, specifically the operational technology supporting weather-related air traffic services, with network monitoring indicating possible data exfiltration to external addresses in China. The agency has brought in forensic investigators, has asked for international assistance, says containment and malware removal are under way and has not ruled out insider involvement. Operators of safety-critical operational technology should prioritise network segmentation and monitoring, and check that weather and supplier data links are not internet facing.
Sources DarkReading|TimesLIVE
TeamViewer fixes five high-severity flaws, worst allows remote session permission bypass Confirmed
TeamViewer published bulletin TV-2026-1010 covering five high-severity vulnerabilities in its Full Client and Host software for Windows, Linux and macOS, tracked as CVE-2026-19743, CVE-2026-92368, CVE-2026-92369, CVE-2026-92370 and CVE-2026-92371. The worst, CVE-2026-92370 at CVSS 8.8, can let an attacker ignore the permissions a user set for a remote session, which matters for anyone who relies on those controls to limit what a support technician or intruder can do. Version 15.82 carries the fixes and the vendor is urging customers to patch as soon as possible, so unpatched remote access software should be treated as an open door.
Sources TeamViewer|BleepingComputer
DIVD breach traced to two Zammad zero-days run through an autonomous AI agent Confirmed
The Dutch Institute for Vulnerability Disclosure said its network was breached by exploiting a chain of two zero-day flaws in Zammad, the open source ticketing and helpdesk platform, in an attack it described as automated and driven by an AI agent. DIVD scans the internet for exposed vulnerable systems and warns their owners, and says this is the first intrusion into its own infrastructure in almost seven years, adding that the activity was noisy rather than stealthy. Organisations running Zammad should watch for vendor updates, apply them quickly and review ticket and API logs for unexplained activity.
Sources BleepingComputer
Talos: China-nexus UAT-11587 targets Asian government and policy bodies with the Antino backdoor Confirmed
Cisco Talos attributed a cluster of intrusions to UAT-11587, a China-nexus group targeting government and policy organisations across Asia, including Taiwan, India, the Philippines and Cambodia, to deliver a previously undocumented backdoor called Antino that appears in developer artefacts. The campaign adds to a steady pattern of state-linked espionage against governments, think tanks and policy bodies in the region. Defenders should review the indicators Talos published, harden internet-facing services and look for unexpected outbound connections from government and policy networks.
Sources Cisco Talos
🟡 MEDIUM (8)
OpenSSL patches high-severity DTLS heap disclosure, WolfSSL fixes auth bypasses Confirmed
OpenSSL released fixes for CVE-2026-84782, a high-severity flaw in its DTLS implementation, the UDP variant of TLS, where a handshake retransmission while a larger message is stalled can leak unencrypted heap memory to the remote peer or crash the process. DTLS is widely used in VPNs, VoIP and IoT products, so the affected surface is broad even though no exploitation has been reported. WolfSSL separately patched three high-severity authentication bypass flaws in version 5.9.4, and teams should apply both sets of updates to the appliances and libraries they ship.
Sources The Hacker News|SecurityWeek
Chrome and Firefox updates patch more than 100 vulnerabilities Confirmed
Google and Mozilla shipped browser updates that together fix over 100 vulnerabilities in Chrome and Firefox, some of which could let a remote attacker execute arbitrary code or escape the browser sandbox. No exploitation was reported alongside the releases. Browser updates should be pushed to managed fleets promptly, and users should restart browsers rather than leaving the updates pending for days.
Sources SecurityWeek
WatchGuard patches critical Fireware OS code injection flaw among 15 fixes Confirmed
WatchGuard released patches for 15 issues in Fireware OS, led by a critical code injection vulnerability, with the rest covering code execution, denial of service, authorisation and path traversal bugs. Firewalls sit directly on the network edge, so flaws of this class deserve urgent attention wherever management interfaces are reachable. Administrators should update Fireware OS and keep device management off the public internet.
Sources SecurityWeek
Microsoft: phishing abuses MSP360 RMM installer to deploy ScreenConnect Confirmed
Microsoft reported phishing campaigns that distribute a legitimate MSP360 remote monitoring and management installer under deceptive file names, using meeting invitations, PDF-themed lures and software update prompts, then use it to install ScreenConnect and create a second remote access channel. Running two RMM tools at once gives attackers persistence even if one is blocked, and the traffic is hard to separate from legitimate administration. Restrict which remote management agents are allowed, alert on unexpected installer execution and review remote access logs.
Sources Microsoft|The Hacker News
CSuite phishing steals Microsoft 365 sessions and deploys remote access tools Confirmed
ANY.RUN traced a US-focused phishing campaign across 351 sandbox analyses, finding 51 per cent of samples submitted from the United States and the highest exposure in technology, manufacturing, government and consulting organisations. The campaign combines Microsoft 365 session theft with deployment of remote access tools, turning a single phish into account compromise and fraud. Organisations should enforce phishing-resistant multi-factor authentication, monitor for stolen session tokens and block unsanctioned remote access software.
Sources The Hacker News
AI coding agents leak over 13,000 internal images into public GitHub repositories Confirmed
Security company Glow found more than 13,000 internal images from developers at over 300 organisations exposed in public GitHub repositories, mostly under personal accounts, after AI coding agents were asked to share screenshots of code changes for review. The images included customer billing records and screens of unreleased features, so the leak mixes commercial secrets with personal data. Organisations should review the settings of AI coding tools, scan public repositories for internal artefacts and treat screenshots as sensitive material.
Sources The Hacker News
Study finds 543,699 live credentials exposed in public GitHub repositories Confirmed
Truffle Security scanned 224 million public repositories and found 543,699 credentials that were still valid, with half of the working secrets exposed for more than two years and the oldest dating back to 2009. Google API keys with Gemini access were the most common live secret at 69,041, and the researchers note that GitHub’s push protection blocks many token types but misses database connection strings, Google keys and private keys by default. Teams should enable secret scanning, rotate anything exposed and audit cloud keys that may still be live.
FBI memo tells staff to assume ShinyHunters holds their personal data Confirmed
An internal FBI memo told employees to assume that the ShinyHunters extortion group may already hold their personal details, after the group claimed to have attacked the bureau’s jobs site. ShinyHunters says it never intended to publish the stolen data and remains defiant after the FBI publicly called on its members to surrender and a suspected leader was arrested in the Netherlands. Staff and contractors whose data may be involved should watch for targeted phishing and credential-stuffing attempts that reuse personal details.
Sources Decrypt|SecurityWeek
⚪ WATCH (7)
White House voluntary AI safety accord signed by OpenAI, Google, Nvidia and Meta Confirmed
The White House announced a one page accord on advanced AI, which officials described as morally binding, under which six companies commit to internal controls, independent audits and board oversight, with no penalties or deadlines attached. The agreement drew criticism for its lack of detail and arrives alongside rising scrutiny of AI risk from regulators and courts. Security leaders should read it as a signal of direction rather than a compliance framework for now.
Sources DarkReading|Decrypt|CoinDesk|SecurityWeek
FTC confirms investigation into OpenAI and Anthropic over consumer risk Under investigation
An FTC spokesperson confirmed that the agency is investigating OpenAI and Anthropic over possible risks to consumers, and declined to comment further. The probe adds to a growing set of regulatory and legal questions about AI services and autonomous agents, including who is liable when an agent causes harm. Enterprises embedding these platforms should expect changes to terms, safety documentation and audit obligations as the inquiries progress.
Sources SecurityWeek
Microsoft to block Entra ID script injection attacks from October Confirmed
Microsoft reminded customers that Entra ID, its cloud identity service, will gain stronger protection against external script injection attacks starting next month. The change may affect custom sign-in and application pages that rely on injected scripts to function. Identity administrators should test their customisations now and check sign-in flows for script injection that they did not intend.
Sources BleepingComputer
OpenAI says it disrupted a coordinated model-distillation campaign Confirmed
OpenAI published details of a campaign it disrupted that aimed to extract protected model reasoning through adversarial distillation, and described new defences against the technique. Model extraction is an intellectual property and abuse problem rather than a classic intrusion, but it sits in the same family of AI platform abuse. Teams running public model endpoints should monitor for the systematic querying patterns that extraction attempts produce.
Sources OpenAI
Google: AI-discovered vulnerabilities skew towards remote code execution Confirmed
Google analysed how AI-assisted vulnerability discovery changes the profile of reported bugs, and found that AI-discovered issues are more likely to enable remote code execution than those found through traditional research. The observation matters for triage, because both the volume and the severity mix of findings shift when AI tooling is used on either side of the fight. Security teams should plan for higher patch volumes and revisit how they prioritise and resource remediation.
Sources SecurityWeek
OFAC sanctions Tren de Aragua figures over a USD 40.7 million ATM jackpotting scheme Confirmed
The US Treasury’s Office of Foreign Assets Control designated ten targets connected to a Tren de Aragua ATM jackpotting operation that stole at least USD 40.73 million, including a fugitive accused of laundering the proceeds with cryptocurrency. The case highlights how organised crime blends physical attacks on cash machines with crypto rails to move and hide money. Banks, ATM operators and payment firms should keep jackpotting indicators and crypto laundering typologies in their monitoring.
Sources Chainalysis
US Senate report calls Tether's USDT a channel for Iranian sanctions evasion Confirmed
A US Senate subcommittee led by Senator Richard Blumenthal released a report describing Tether’s USDT as a superhighway for Iranian sanctions evasion and referred its findings to the Treasury and the Justice Department. The report increases pressure on stablecoin issuers to tighten sanctions controls and to know who holds their tokens. Compliance teams at exchanges and payment providers should expect closer scrutiny of USDT flows connected to sanctioned jurisdictions.
Sources Decrypt
Methodology: compiled from vendor advisories, government feeds (CISA KEV, MSRC), security news sources and on-chain/security-firm alerts. Grouping: one incident, one entry, with every source cited. Unconfirmed reports are labeled. Crypto items are incident reporting, not investment advice. Corrections from prior digests are noted at the top when applicable.