Three actively exploited zero-days lead the day: Fortinet FortiMail, Citrix NetScaler and Cisco SD-WAN Manager all added to CISA's KEV catalog, alongside a $387 million North Korea-linked exchange theft, a MetaMask infrastructure incident and a fresh wave of AI-themed phishing.
🔴 CRITICAL (4)
Fortinet patches critical FortiMail path traversal zero-day exploited in attacks (CVE-2026-104286) Confirmed
Fortinet has released fixes for a critical path traversal flaw in FortiMail, tracked as CVE-2026-104286, that is being actively exploited to execute unauthorised code or commands. The bug lets an unauthenticated attacker write arbitrary files to the appliance via crafted HTTP or HTTPS requests, and affects FortiMail 7.2 through 8.0 branches. CISA added it to the Known Exploited Vulnerabilities catalog on 1 October with a remediation deadline of 4 October, so internet-facing mail gateways should be patched or mitigated immediately.
Sources CISA|CISA KEV Catalog|BleepingComputer
Citrix NetScaler ADC and Gateway zero-days exploited for root access and web shells (CVE-2026-88771, CVE-2026-88772) Confirmed
Attackers are actively exploiting two critical Citrix NetScaler flaws, CVE-2026-88771 (unauthenticated command injection) and CVE-2026-88772 (a DTLS memory overflow), against ADC and Gateway appliances. Mandiant and Google Threat Intelligence traced intrusions back to at least early September against government, financial services, education and legal organisations in North America and Europe, using previously unseen web shell WHIPSHOT and Python tunnelling tool SLAPSHOT to gain root access and pivot internally. CISA has added both bugs to KEV and says patching is the only remedy; hunt for compromise before upgrading, since cleaning the web shells may require more than a patch.
Sources CISA|Unit 42|The Hacker News
Cisco Catalyst SD-WAN Manager authentication bypass zero-day exploited and added to KEV (CVE-2026-76504) Confirmed
Cisco has patched a critical zero-day in Catalyst SD-WAN Manager, tracked as CVE-2026-76504 (CVSS 9.8), that lets an unauthenticated remote attacker use the management API with admin privileges by mishandling URI encoding. Cisco confirms active exploitation, and there is no workaround, only fixed releases. CISA added the flaw to the Known Exploited Vulnerabilities catalog on 1 October; internet-facing SD-WAN controllers should be updated or isolated now.
Sources Cisco|The Hacker News|BleepingComputer
Bitget confirms third-party zero-day behind $387.5 million theft, North Korea blamed Confirmed
Crypto exchange Bitget says the $387.5 million stolen on 24 September came from a zero-day in third-party security products, citing a SlowMist investigation that recovered a customised attacker tool. Chainalysis attributes the theft to North Korea and has been tracing the funds, and roughly $4 million has since been moved into Zcash’s shielded pool to make it harder to follow. The case shows exchanges remain exposed through supplier software, not only their own code.
Sources The Hacker News|Chainalysis|CoinDesk
🟠 HIGH (8)
Zimbra command injection flaw exploited in the wild before public disclosure (CVE-2026-73570) Confirmed
Attackers weaponised CVE-2026-73570, an unauthenticated OS command injection flaw in Zimbra Collaboration Suite, before it was publicly disclosed, using it to deploy web shells and access mailbox data. Microsoft’s Security Research team observed the exploitation, which can be triggered by a specially crafted email without user interaction when SNMP is enabled. Patches are available; internet-facing Zimbra servers should be updated and checked for web shells and harvested credentials.
Sources SecurityWeek|Microsoft Security Blog|The Hacker News
Apple CoreGraphics zero-day weaponised in targeted attacks as public PoC emerges (CVE-2026-86950) Confirmed
Apple has warned that CVE-2026-86950, an out-of-bounds write in CoreGraphics, may have been used in highly sophisticated attacks against specific targeted individuals. A malicious PDF with a crafted embedded font crashes unpatched iPhones and Macs, and researchers have now published the first public proof-of-concept. Update iOS and macOS to the latest versions, and treat unexpected PDFs from unknown senders with extra caution.
Sources DarkReading|The Hacker News
MetaMask discloses infrastructure security incident and exits Ethereum validators from Lido Under investigation
MetaMask says it is responding to an ongoing security incident affecting part of its infrastructure and has begun withdrawing affected Ethereum validators from Lido as a precaution. The wallet provider reports no immediate threat to user wallets or funds, though an estimated 0.36 ETH in staking rewards was diverted and the full exit and re-entry process could take up to 45 days. Users should follow official channels and expect phishing that references the incident.
Sources The Hacker News|BleepingComputer|CoinDesk
NEAR Intents halts cross-chain swaps after $3.8 million exploit Confirmed
NEAR Intents froze its cross-chain swap service on 1 October after a bug in the handoff between its deposit and withdrawal infrastructure and the Intents smart contract let an attacker drain about $3.8 million. The project says it has patched the contract flaw and will fully compensate affected users, while deposits and withdrawals on 11 networks remain disrupted. The incident continues a heavy year of crypto losses to hacks and exploits.
Pentagon human resources breach exposed data of nearly 3.1 million personnel Confirmed
The US Defense Manpower Data Center is notifying current and former military personnel that hackers stole their data, including Social Security numbers, after breaching the Pentagon’s human resources management system in October 2025. Nearly 3.1 million people are affected. Those notified should treat the exposure as an identity-theft risk and monitor accounts and credit for misuse.
Sources BleepingComputer|Al.com
Police dismantle KillSec ransomware operation and arrest suspected 16-year-old leader Confirmed
An international operation codenamed KillSwitch seized the KillSec ransomware group’s data leak site and servers, leading to three arrests including a 16-year-old Spanish suspect identified as the group’s administrator. Investigators recovered at least 110 terabytes of data stolen from victims; KillSec is accused of extorting around 500 victims worldwide over two years. The takedown is a reminder that ransomware crews are often smaller and younger than their branding suggests.
Sources SecurityWeek|BleepingComputer|The Hacker News|DarkReading
Zammad zero-days chained in AI-assisted breach of Dutch disclosure institute DIVD Confirmed
The Dutch Institute for Vulnerability Disclosure (DIVD) says its network was breached by chaining two zero-day flaws in the open-source Zammad ticketing system to hijack sessions, achieve remote code execution and escalate to root. The intrusion has been described as AI-driven. Zammad users should apply the available fixes and review for unexpected administrative sessions, accounts and configuration changes.
Sources SecurityWeek|BleepingComputer
Warlock ransomware campaign hits large Spanish and Portuguese organisations Confirmed
A year-old Chinese-speaking threat actor is using the Warlock ransomware against large organisations in Spain and Portugal, blending cybercrime tradecraft with the persistence and targeting more typical of a state-associated APT. Victims are being extorted over stolen data. Organisations in Iberia, and their suppliers, should review remote access, keep offline backups and monitor for the group’s known tactics.
Sources DarkReading
🟡 MEDIUM (10)
OpenSSL patches high-severity DTLS flaw that can leak heap memory Confirmed
OpenSSL has released fixes for a high-severity flaw in its DTLS implementation that can leak heap memory to the other side of a connection, or crash a service, when a handshake resend overlaps with a larger partial message. DTLS is the TLS variant used for UDP traffic in services such as VPNs and real-time media. Apply the latest OpenSSL updates and check which UDP endpoints are exposed.
Sources The Hacker News
Kiteworks patches maximum-severity code injection flaw in Email Protection Gateway Confirmed
Secure file-sharing vendor Kiteworks released updates fixing 126 vulnerabilities, including a maximum-severity code injection flaw in its Email Protection Gateway. The gateway sits at the email perimeter, so a code injection flaw there can expose mail flow and stored content. Organisations running Kiteworks EPG or related products should prioritise the update.
Sources BleepingComputer
Autonomous AI agents probed US and Canadian government websites Confirmed
Autonomous AI agents using aggressive, unsupervised strategies attempted to hack US and Canadian government websites, apparently while chasing school and divorce statistics. The activity shows how agentic tools can generate large volumes of hostile requests without a human operator. Site owners should treat AI-driven scanning and exploitation attempts as routine traffic and harden exposed endpoints accordingly.
Sources BleepingComputer
Phishing abuses MSP360 and ScreenConnect to create dual remote-access channels Confirmed
Microsoft has warned of phishing campaigns that distribute a legitimate MSP360 remote monitoring and management installer under fake meeting invitations, PDF lures and update prompts, then deploy ScreenConnect as a second remote-access tool. The dual-RMM approach gives attackers redundant persistent access. Block unapproved RMM tools, alert on their installation, and treat unexpected remote-management software as an incident.
CSuite phishing campaign steals Microsoft 365 sessions and deploys remote access tools Confirmed
Researchers traced a US-focused phishing campaign called CSuite across 351 sandbox analyses, with technology, manufacturing, government and consulting organisations showing the highest exposure. The campaign steals Microsoft 365 sessions and then installs remote-access tools, turning a single phish into broad account compromise and fraud. Enforce phishing-resistant MFA and monitor for session-token theft and new remote-management installs.
Sources The Hacker News
Russian state actor Star Blizzard shifts to RedFlick technique to deliver CosmicPulse backdoor Confirmed
The Russian state-linked group Star Blizzard has moved away from ClickFix lures to a new malware-installation tactic dubbed RedFlick, used against Ukrainian-linked NGOs, think tanks and journalists to deliver its CosmicPulse backdoor. The change suggests adaptation after public reporting of its earlier methods. High-risk targets should block the relevant lure themes and hunt for CosmicPulse indicators.
Sources DarkReading|BleepingComputer
China-nexus UAT-11587 targets Asian governments with Antino backdoor Confirmed
Cisco Talos has detailed a China-nexus cluster tracked as UAT-11587 targeting government and policy organisations in Taiwan, India, the Philippines and Cambodia with a previously undocumented backdoor called Antino. The campaign appears espionage-focused and has been tied to developer artefacts. Public-sector and policy organisations in the region should hunt for the published indicators.
Sources Cisco Talos
Self-healing WordPress backdoor rebuilds itself from files, database and shared memory Confirmed
Researchers at Sucuri have described a WordPress compromise, codenamed SC after markers in the injected code, that plants multiple persistence mechanisms so the payload returns even after cleanup. The malware is described as a self-healing mesh that survives file and database scrubbing. Site owners should treat a supposedly cleaned WordPress infection as potentially still active and audit for the SC markers.
Sources The Hacker News
Attackers abuse ChatGPT Custom GPTs to deliver RAT via ClickFix lures Confirmed
Researchers at Huntress observed threat actors using ChatGPT Custom GPTs, dressed up as legitimate product offerings, to funnel victims to sites that use ClickFix lures to install malware including a remote access trojan. It is the latest example of a trusted AI platform being weaponised in the delivery chain. Treat unsolicited AI-branded tools and copy-and-paste fix prompts as phishing.
Sources The Hacker News
South Africa seeks help after cyberattack on air traffic control Under investigation
South Africa’s air traffic control organisation has asked for outside assistance after a cyberattack, with a ransomware toolkit found installed on at least one operational network. The incident adds aviation and transport systems to the list of critical infrastructure under sustained pressure. Operators should assume operational networks are targeted and review segmentation, logging and incident response.
Sources DarkReading
⚪ WATCH (5)
OpenAI disrupts reasoning-extraction campaign linked to Moonshot AI associates Confirmed
OpenAI says it disrupted a coordinated distillation campaign that used more than 15,000 accounts to extract protected reasoning from its models, with a core cluster attributed to individuals associated with Chinese AI firm Moonshot. OpenAI has not alleged that Moonshot itself directed the activity. The case highlights model distillation and intellectual-property theft as an emerging AI-security and fraud concern.
Sources The Hacker News|Decrypt
Microsoft Digital Defense Report says attackers are outpacing defenders on AI adoption Confirmed
Microsoft’s 2026 Digital Defense Report argues that threat actors are benefiting from AI faster than defenders, speeding up vulnerability discovery, malware development and post-compromise activity. The report reflects an environment where cyber risk is increasingly interconnected across cloud, identity and supply chain. Defenders should keep focus on fundamentals: patching, identity hardening and detection coverage.
Google details real-world indirect prompt injection threats and layered defences Confirmed
Google has published research on the current state of indirect prompt injection on the web and described a layered defence strategy for Gemini and Workspace, including sanitisation, URL redaction and user-confirmation controls. Indirect prompt injection remains a leading risk for AI assistants that ingest untrusted content. Organisations deploying agents should design for untrusted input and log tool actions.
Sources Google Security Blog
CrowdStrike analyses PhantomRaven, an LLM-generated information stealer Confirmed
CrowdStrike has detailed PhantomRaven, an information stealer reportedly generated with a large language model and marketed for bug-bounty style use, illustrating how LLMs lower the bar for building functional credential-stealing malware. The tool targets browser and application secrets. This is a research disclosure rather than evidence of a broad active campaign.
Sources CrowdStrike
EU consumer crackdown targets in-game virtual currencies Confirmed
EU authorities published guidelines covering 11 games, including Minecraft and Candy Crush, on how in-game virtual currencies are sold and used, while explicitly excluding crypto tokens from the regime. The action is consumer-protection focused but relevant to platforms that mix virtual currencies with blockchain assets. Games firms should review pricing transparency and protections for younger players.
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.