PAN-OS flaw exploited by Qilin and Settra ransomware
A PAN-OS vulnerability is being actively exploited by the Qilin and Settra ransomware groups, according to ReliaQuest. CVE-2026-0257 is an authentication bypass, meaning an attacker may get around the normal identity check, in PAN-OS GlobalProtect and Prisma Access. The flaw was disclosed in May 2026. Because access comes through the organisation's own virtual private network (VPN), malicious sessions can resemble ordinary remote work while attackers steal credentials or data, move through internal systems, or deploy ransomware. ReliaQuest recommends upgrading affected systems, disabling authentication override cookies where they are unnecessary, ending every active GlobalProtect session after patching, and investigating unexpected connections, including sessions using hostnames such as “kali.” Installing the update does not remove access that attackers may already have established.
Several web and cloud products also have high-impact flaws. Four patched vulnerabilities in the cloud version of Adobe Experience Manager can be chained for unauthenticated remote code execution (RCE), which means running code without first signing in. The chain uses an unauthenticated DAM bulk import, stored cross-site scripting, unsafe XML parsing and Groovy execution through /bin/asynccommand. Samsung patched CVE-2026-110392, a critical command-injection flaw in QA-Board caused by improper handling of user-controlled data. It could let an unauthenticated attacker run shell commands. The vulnerability was flagged by Claude Opus 5.5.
Active exploitation has also been reported against CVE-2026-106931 in EG Secure Solutions SiteGuard Server Edition. It affects version 2.20 or later when statistics are enabled, which is the default, and can allow operating-system commands to run with root or SYSTEM privileges. The method, operators and targets remain unknown. Rapid7 separately disclosed CVE-2026-108065, a critical SQL injection flaw in Channel Software WebAlliance eCommerce for which no patch was announced. SQL injection manipulates a site's database requests. Attackers could read personal information and passwords, or run system commands if the database account has administrator privileges. Harbor has patched GHSA-2phj-cp9f-qq6w, a server-side request forgery flaw that OX Security says lets a registered user who can create a project direct a webhook to a cloud metadata endpoint and obtain the server's IAM credentials.
Infrastructure operators face flaws at the host, container, virtual-machine and DNS layers. CVE-2026-47595, also called Bad UVM, affects the NVIDIA GPU Display Driver for Linux because memory permissions are not preserved. A researcher demonstrated that an unprivileged user could write to read-only memory and potentially gain root access, tamper with data, disclose information, cause denial of service or escape a container to reach the host. Xen addressed CVE-2026-79610, a grant-table race condition affecting every Xen version from 4.0 onward, with xsa517.patch. An unprivileged guest could potentially obtain host privileges. The listed mitigations are gnttab=max-ver:1 for the hypervisor or max_grant_version=1 for guests. NSD 4.15.3 or newer fixes four command-execution vulnerabilities: CVE-2026-101341, CVE-2026-101339, CVE-2026-101338 and CVE-2026-101337. Quarkslab also found that MyReport Updater 2.1.1 can let a standard Windows user choose an arbitrary reference server. If the service reaches it over gRPC on TCP port 9095, files can be written with SYSTEM privileges and a malicious DLL can be loaded, potentially taking control of a Citrix server or another affected machine.
Website owners should also note a resilient WordPress backdoor documented by Wordfence. Installed as a must-use plugin, the malware uses several mechanisms to restore itself after removal. It also uses EtherHiding, which stores the location of attacker-controlled servers behind an Ethereum smart contract to make the command channel harder to disable. Its capabilities include creating and concealing an administrator account, collecting administrator passwords, changing the site remotely, removing data and spreading to other WordPress installations on the same file system. That combination makes checking administrator accounts, must-use plugins and neighbouring installations important after cleanup.
DNS is also being used to support cryptocurrency theft. Infoblox analysed the Noir wallet-drainer kit, whose loader queries DNS TXT records through public DNS over HTTPS (DoH) services to find an active backend hosted on Cloudflare Pages. The drainer does not require the victim's seed phrase, password or private key. It first requests read access, adapts its content to the wallet, then seeks spending permissions that may look less alarming than a direct transfer. Theft is subsequently performed from attacker-controlled infrastructure. Infoblox suspects, but has not confirmed, that artificial intelligence assisted the malware's development and maintenance.
Other malware campaigns broaden the risks to endpoints and developer accounts. Darktrace observed the Rust-based CikaBika Bot arriving through a document lure containing a malicious LNK file and a PDF with an embedded payload. It uses a Nim loader, a PowerShell User Account Control bypass and DLL search-order hijacking, then relies on Discord for command and control, or C2, as well as reconnaissance and remote execution. It can capture screenshots, transfer files and exfiltrate data through ngrok. Point Wild described BotHelper RAT, a remote access trojan that fingerprints a Windows host, disables TLS certificate validation and downloads payload.bin from easyllms.xyz. Its second stage hides itself, runs every 30 minutes through a scheduled task, interferes with the Antimalware Scan Interface and polls PHP endpoints for instructions.
Group-IB reports that KGB RAT has been marketed since late 2025, although a public reference to server version 0.1 dates to January 2022. Its advertised capabilities combine remote desktop control, credential theft, keylogging, camera and microphone access, persistence, ransomware and harassment features such as jump scares, loud visual interruptions and forced system crashes. Plans range from $150 for one month to $800 for one year. XWorm has meanwhile expanded from 35 to 46 plugins, adding GPS collection without a prompt, live microphone audio, webcam capture and control of authenticated browser sessions that can bypass multi-factor authentication. Versions detected since July 2025 include V6.4, V6.5 “XWorm Reborn,” V7.2 and V7.4.
Credential-stealing malware is increasingly interested in development and AI tools. Gen Digital says the Rust-based Warden Stealer targets local data and configuration files belonging to AI assistants and agentic coding tools, while also providing a loader and cryptocurrency clipper. Symantec says Aoki targets credentials and session files for Claude, Codex, Cursor, Codeium and Gemini.