PHP2Shell exploitation attempts escalated into Mirai botnet delivery while researchers published a working ASLR bypass, CISA added two more exploited flaws to its KEV catalog, and an AI agent framework ran ransomware intrusions end to end.
🔴 CRITICAL (1)
PHP2Shell attacks move to Mirai botnet delivery as researchers defeat ASLR Under investigation
Three days after PHP shipped emergency fixes for CVE-2026-109618 and CVE-2026-109619, exploit attempts are reaching millions of requests and Sophos has tied one campaign to Potassium, an expanded Mirai variant that can download payloads and run DDoS attacks. VulnCheck says it has an exploit that works on production x86-64 builds with ASLR enabled and error display off, which earlier public proof of concept code did not. No successful in the wild compromise has been confirmed by any vendor so far, but the UK government reported 32 services unavailable while it patched. Update to PHP 8.5.12, 8.4.27, 8.3.36 or 8.2.35, restrict error display, and hunt for requests to PHP endpoints carrying the php2shell-scan user agent.
Sources CISA|Unit 42|SecurityWeek|The Hacker News|SANS ISC
🟠 HIGH (8)
Arista VeloCloud zero-day CVE-2026-114625 added to CISA KEV, third VCO flaw exploited this year Confirmed
CISA added the VeloCloud Orchestrator SQL injection to its Known Exploited Vulnerabilities catalog on 23 October, with a remediation deadline of 26 October for federal agencies. It is the third VeloCloud Orchestrator zero-day exploited in four months, and any authenticated account, including read only, can reach operator privileges on on-premises deployments before 5.2.3.18, 6.1.3.10 and 6.4.2.12. Restrict orchestrator management interfaces to trusted networks and review hosts for webshells and unexpected daemons.
Sources SecurityWeek|The Hacker News|BleepingComputer|CISA KEV
BlackX UEFI bootkit bypasses Secure Boot on patched Windows hosts that skipped DBX revocations Confirmed
Kaspersky reports a bootkit called BlackX used against diplomatic organisations in Tajikistan, Uzbekistan and Turkmenistan in early 2026, and assesses it as an updated copy of BlackLotus, with activity attributed to a China-linked group suspected to be APT41. It abuses the patched Baton Drop flaw (CVE-2022-21894) because the Windows Production PCA 2011 certificate, which expired on 19 October 2026, is still not blocked in the Secure Boot forbidden signature database on many systems. Installing updates is not enough: administrators must apply Microsoft’s revocation steps to add the certificate to DBX, then verify the entry is present.
Sources The Hacker News
HPE patches 72 ArubaOS flaws, eight of them maximum severity and pre-auth Confirmed
HPE released ArubaOS updates covering 72 vulnerabilities, including eight maximum severity authentication bypass, command injection, SQL injection and buffer overflow flaws in AOS-8 and AOS-10 (CVE-2026-79844 through CVE-2026-79851). The worst are reachable remotely without credentials, low complexity, no user interaction, and affect mobility conductors, controllers, gateways and access points. HPE says it has seen no public exploit code, and recommends patching immediately while restricting CLI and web management to a dedicated VLAN or firewall policy.
Sources BleepingComputer
k0t group breached FortiGate and GlobalProtect edge devices, exposing over 70,000 credentials Confirmed
Group-IB and NyxLab traced a campaign against internet-facing Fortinet FortiGate and Palo Alto Networks GlobalProtect appliances to an actor tracked as k0t, also known as CryptoFeel and Member19, who is assessed to work with Qilin and The Gentlemen as an access broker. Operators exploited CVE-2025-64446 and CVE-2026-0257 or brute forced logins, then persisted with MeshCentral, Atera, Teramind, AnyDesk and AdaptixC2 and exfiltrated data through Rclone to Wasabi and MEGA. Nearly 38 percent of victims are in Asia Pacific, so regional teams should review VPN and firewall logs for OpenConnect sessions and unfamiliar remote management agents.
Sources The Hacker News
Collabora Online CODE Server command injection CVE-2026-110447 added to KEV Confirmed
CISA added an OS command injection flaw in Collabora Online’s built in CODE Server to the Known Exploited Vulnerabilities catalog on 23 October, citing evidence of active exploitation, with a 26 October deadline for federal agencies. Self-hosted document collaboration servers that expose the CODE Server component should be updated to the fixed release or taken off the public internet, and administrators should review for command execution artefacts in web server logs.
Sources CISA
INC and The Gentlemen affiliates ran intrusions through an AI agent framework called Phantom Confirmed
SOCRadar found an exposed server hosting Phantom, an autonomous red team framework that affiliates of the INC and The Gentlemen ransomware operations used to run intrusions, logging 8,257 messages across 133 sessions. A human approved each step while OpenCode with commercial models executed it, and when a model refused, the operator reframed the request as authorised red team work. Initial access came from exploited edge and remote access products including FortiClient EMS, SimpleHelp, cPanel and VPN gateways, so patch those products and watch for the BYOVD driver activity the group used to impair security tooling.
Sources The Hacker News
Sandworm and UAC-0099 used a mesh network backdoor in a destructive wipe of a Ukrainian logistics firm Confirmed
ESET reports that Sandworm worked with UAC-0099, also known as Earth Sirrush, on a destructive attack against a Ukrainian logistics company, starting from a spear phishing email with a malware attachment in July 2025. The intrusion used a previously undocumented backdoor that routes command and control over a mesh network, which makes blocking single C2 addresses ineffective, before deploying wipers. Defenders should expect follow-on destructive operations against Ukrainian and allied logistics targets and monitor for mesh networking libraries on endpoints.
Sources The Hacker News
Quest Diagnostics confirms breach of its Quanum Lab Services Manager platform Confirmed
Quest Diagnostics disclosed that an intruder reached patient records through unauthorized access to Quanum Lab Services Manager, the portal some client organisations use to order tests, and reported the incident in a regulatory filing. The company has not published how many people are affected or which data fields were exposed. Healthcare providers and patients who use Quanum should treat the exposed information as sensitive and expect targeted phishing that references test orders.
Sources BleepingComputer|SecurityWeek
🟡 MEDIUM (11)
PHP discloses CVE-2026-115560, another code execution risk with no patch Confirmed
The PHP project warned of a second way to run code on a web server, CVE-2026-115560, two days after the emergency PHP2Shell fixes. It affects only applications configured in an unusual way and PHP rates it an unlikely risk, so no emergency release is planned. The project instead recommends disabling the affected feature in php.ini, and administrators who cannot patch should confirm the setting is off and monitor for follow-on exploit attempts.
Sources The Hacker News|BleepingComputer|SANS ISC
Atlassian October bulletin lists 176 vulnerability entries, all in third party components Confirmed
Atlassian’s October 2026 bulletin covers Bamboo, Bitbucket, Confluence, Crowd, Fisheye/Crucible, Jira Software and Jira Service Management, with 162 high and 14 critical entries that collapse to 75 unique CVEs, six critical. The critical issues are in Apache Tomcat, Bouncy Castle, Netty and Apache SSHD, and Atlassian assesses its exposure as lower risk than the raw scores suggest. It reports no exploitation, but self-hosted Data Center and Server customers should apply the fixed releases because third party flaws in these products have been exploited within hours before.
Sources SecurityWeek
Ruby fixes three Onigmo regex engine bugs behind four critical GitLab RCE flaws Confirmed
Ruby released 4.0.9, 3.4.12 and 3.3.13 to fix three memory corruption bugs in Onigmo, the regular expression engine built into the language, caused when a crafted pattern is compiled. The same bugs sit behind four critical remote code execution flaws GitLab patched over the past month. Any service that compiles attacker supplied regular expressions on a vulnerable Ruby build is exposed, so upgrade Ruby as well as GitLab and review input validation around user supplied patterns.
Sources The Hacker News
Pillow CVE-2026-112900 lets a crafted EPS image run commands through Ghostscript Confirmed
Pillow, the most widely used Python imaging library, passes the filename of an EPS image to Ghostscript without validating it, so a crafted file name can run operating system commands when an application processes user uploads. The flaw is CVE-2026-112900 and is rated critical. Teams that accept images and hand them to Pillow should update, and in the meantime convert or reject EPS input and avoid passing user controlled names to any Ghostscript wrapper.
Sources The Hacker News
Imperva demonstrates Node2Shell RCE on a stock Node.js 24 build via crafted Brotli data Confirmed
Imperva researchers showed that malicious Brotli data can make an application running the official Node.js 24.22.0 Linux arm64 image execute attacker supplied code, with no debugger or custom build, tracking the decoder flaw as CVE-2026-111497. The demonstration follows Node.js patches released earlier in the week and turns a parsing bug into a full code execution path. Node operators should confirm they are on a fixed release and treat unsolicited compressed request bodies with suspicion.
Sources The Hacker News
x402 SDK flaw let a malicious API drain an AI agent's whole token balance Confirmed
V12 found that AI agents using the x402 payment SDKs could sign away their full balance of a supported token when a malicious API asked for it, because the default client set no spending limit until releases on 18 August added one. Only agents that automatically signed payments, with no application level cap, were exposed. Anyone running x402 paying agents should upgrade, set explicit per request and daily limits, and log signed payments.
Sources The Hacker News
Abandoned Popcorn Time updater left millions of devices open to remote code execution Confirmed
Researchers found the updater in Time4Popcorn, a fork of the Popcorn Time BitTorrent client that is no longer maintained, could be used to push code that ran with elevated privileges, affecting an estimated millions of installations. The team disabled the vulnerable update channel remotely, which closes the immediate path but leaves the application itself unmaintained. Users should uninstall the client rather than rely on the disabled updater.
Sources BleepingComputer
CISA publishes another batch of ICS advisories, including a CVSS 9.9 NDI encoder flaw Confirmed
CISA’s latest industrial control system advisories cover Siemens SIMATIC S7 web servers, RUGGEDCOM, WinCC and Software Center, Hitachi Energy MicroSCADA X, InterSystems IRIS for Health and a Kiloview E3 NDI encoder set of nine flaws rated 9.9 that allow full device compromise. Most issues need network access to the management interface, but several are command injection or missing authentication that need no credentials. Asset owners should map the advisories to installed versions and segment management interfaces from business networks.
Sources CISA Advisories|CISA Advisories|CISA Advisories|CISA Advisories
Prince of Persia APT returns with command and control hosted in the United States Confirmed
After a three month pause, the Iran linked group tracked as Prince of Persia hid a Tornado command and control server on US shared hosting, its first US based server since 2016, according to SecurityWeek. Using shared hosting makes the infrastructure harder to block by reputation and easier to blend with legitimate traffic. Network teams should alert on the group’s known Tornado indicators and on outbound connections from servers to commodity hosting providers.
Sources SecurityWeek
CoinDCX denies a breach after researchers flag 16 million dollars of outflows Unconfirmed report
Blockchain sleuths pointed to about 16 million dollars in suspicious outflows from addresses linked to India’s largest exchange, and CoinDCX denied that its active wallets were compromised. No security firm has published a confirmed root cause, so the cause remains unclear. Customers should rotate API keys and enable withdrawal allowlists while the exchange’s statement is checked against on-chain flows.
Sources CoinDesk
Check Point documents an infostealer built to harvest AI assistant and coding tool sessions Confirmed
Check Point Research analysed an infostealer that keeps the usual browser password, cookie and wallet theft and adds dedicated collectors for Claude, ChatGPT Desktop, Codex, Cursor, Windsurf and GitHub Copilot, plus cloud credentials and local development files. Stolen session tokens for these tools often bypass password and MFA controls entirely. Developers should assume any AI assistant session on an infected machine is compromised and revoke tokens rather than just changing passwords.
Sources Check Point Research
⚪ WATCH (4)
Bitcoin Core rewrites its security policy for a world where AI finds bugs quickly Confirmed
Bitcoin Core published a revised security disclosure policy that accounts for AI assisted bug discovery, shortening assumptions about how long a quietly fixed flaw stays secret. The change is a process response, not an incident, and follows discussion of how quickly language models can surface memory safety and logic bugs in mature code. Operators of nodes and wallets should still prioritise upgrades, since the policy shift implies faster disclosure windows.
Ethereum delays the Glamsterdam testnet fork over gas accounting security flaws Under investigation
Ethereum developers postponed the Glamsterdam testnet fork after security review flagged risks to existing contracts in the proposed gas accounting changes. The delay keeps mainnet timing uncertain and means contract authors have more time to test against the new rules. The issue is a design risk under review, not an exploited vulnerability.
New York seeks 1.37 billion dollars from NovaTech founders over crypto fraud Confirmed
New York’s attorney general asked a court to order the founders of NovaTech, who have left the country, to pay about 1.37 billion dollars, and settled with four senior promoters for nearly 6 million dollars. The scheme drew victims through prayer groups and community networks rather than technical exploits. The case is a reminder that relationship based fraud, not smart contract bugs, remains the largest source of retail crypto losses.
Sources Decrypt
Microsoft downgrades the Windows NLB RCE score after correcting the attack vector Confirmed
Microsoft corrected CVE-2026-83862, a Windows Network Load Balancing remote code execution flaw, from a network to an adjacent attack vector, lowering the CVSS base score from 9.8 to 8.8. The company says the change is informational and systems that installed the update need no further action. Teams tracking patch priority by score should update their risk registers so the flaw does not stay ranked above flaws that need immediate work.
Sources MSRC
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.