
SAP CVSS 10 Flaw, Chrome and Windows Zero-Days, MikroTik Exploits
Coverage: 09 Sept 2026 (UTC)
< view all daily briefs >Patch activity and confirmed exploitation dominate the security news for 9 September 2026. SAP fixed a CVSS 10.0 pre-authentication kernel flaw, Google and Microsoft addressed actively exploited zero-days, and MikroTik shipped RouterOS fixes for vulnerabilities already used to take over routers. Research on memory-resident web shells, a zero-click WeChat worm and identity-focused intrusion campaigns adds to the picture, alongside several AWS platform updates.
Critical Patches and Exploited Zero-Days
SAP has issued emergency security updates for several critical vulnerabilities, led by a SAP kernel flaw tracked as CVE-2026-44756 (CVSS 10.0). Discovered by Onapsis and dubbed OVERPASS, the bug is a memory corruption issue in the kernel's Extended Passport (EPP) processing. Missing boundary validation during EPP deserialization allows unauthenticated remote attackers to send crafted requests that trigger memory safety violations and ultimately execute arbitrary OS commands as the SAP administrative user. Because the vulnerable code is shared across multiple protocols and reachable from internet-facing components, including web, SAP GUI and RFC, a single network control cannot fully mitigate it, and the flaw bypasses authentication and authorization protections. Successful exploitation can expose the SAP secure store, recover database credentials and password hashes, read live user sessions, enable lateral movement to other SAP systems, and allow modification of application data and binaries. SAP also patched CVE-2026-58240 (CVSS 9.8), named S4GET by Onapsis, a logic flaw in NetWeaver Message Server that allows unauthenticated network attackers to achieve remote code execution as the OS-level <sid>adm account. Two further critical issues were fixed: CVE-2026-76969, a credential disclosure in SAP CAP, and CVE-2026-66768, an improper access control flaw in SAP GUI for Java. No exploitation has been observed, but Onapsis and SAP urge immediate inventorying and patching, prioritizing internet-facing instances, reducing exposure and increasing monitoring, since standard SAP authorization and segregation-of-duties controls do not prevent exploitation of these pre-authentication flaws.
Google pushed a Chrome security update fixing 230 vulnerabilities, including CVE-2026-87491, an out-of-bounds write in the V8 JavaScript and WebAssembly engine that is being actively exploited. Researcher Jihyeon Jeong of Compsec Lab reported the issue on August 6, 2026; Google acknowledged working exploits but withheld technical details pending broad deployment. The release also addresses five critical flaws in WebGL and Cast and a high-impact WebPackaging use-after-free credited to OpenAI Codex Security. Google said it reported 195 of the 230 flaws itself, citing sanitizers and fuzzing tools such as AddressSanitizer, MemorySanitizer, libFuzzer and AFL. Fixed versions are 153.0.8010.36/.37 on Windows and macOS and 153.0.8010.36 on Linux, and Chromium-based browser vendors are expected to roll out corresponding patches. CISA added CVE-2026-87491 to its Known Exploited Vulnerabilities catalog on September 9, with a remediation deadline of September 23, 2026 for Federal Civilian Executive Branch agencies.
Microsoft set a Patch Tuesday record with a 974-flaw release: 723 in Windows, 111 in Office, 62 in SQL Server and 22 in Developer Tools, reaching 999 when some non-Microsoft CVEs are included. Two of the flaws are actively exploited zero-days that allow local privilege escalation to SYSTEM: CVE-2026-85880, a heap-based buffer overflow in ALPC, and CVE-2026-81963, an improper link resolution issue in the Windows Update Stack. CISA added both to KEV with a September 22, 2026 deadline for federal agencies. High-severity issues also affect Exchange, Authenticator, SharePoint, SQL Server, Remote Desktop Services, DNS and DHCP, several rated above 9.0 CVSS. Separately, researcher Chaotic Eclipse published a Defender bypass PoC named ShieldCrash, which the researcher says re-exploits CVE-2026-69414 (ShieldBreak) because the original fix was incomplete. The PoC demonstrates arbitrary file read with SYSTEM privileges on fully patched Windows desktop systems, with all supported desktop versions reportedly affected. Microsoft shipped Malware Protection Engine version 1.1.26080.3 to address CVE-2026-69414, noting that the update requires no customer action and that systems with Defender disabled are not affected. The researcher has recently released PoCs against products from CrowdStrike, Kaspersky, Avast and NVIDIA.
MikroTik released fixes for six RouterOS vulnerabilities after CERT Polska disclosed active router exploitation, a campaign termed MikroTrick. The most severe issue, CVE-2026-67276, is an SSH public-key validation bug that lets an attacker who knows a username and the public modulus craft a private key and authenticate as that user; CVE-2026-86060, in SSH username handling, enables escalation to root. Affected components include the SSH server and client, the bandwidth-test service, X.509 handling and the WebFig interface. Patches are available for RouterOS 7.25 beta 3, 7.24.2, 7.23.4 and 6.49.21, and scans show more than 122,500 devices with SSH reachable from the internet. MikroTik added a Flagged marker indicating signs of compromise, though CERT Polska cautions that its absence does not guarantee safety. Recommended actions include immediate updates, isolating and factory-resetting flagged devices, preserving logs and configurations, rotating all keys and passwords, and blocking SSH, WWW/WWW-SSL and bandwidth-test services from untrusted networks.
cPanel disclosed a cPanel root flaw, CVE-2026-67401, an SQL injection in the EmailTrack component that allows an authenticated hosting account with mail-related privileges to create arbitrary files and execute code as root. Every supported cPanel & WHM release line is affected, with fixed builds listed for the 11.110, 11.134, 11.136, 11.138 and WP Squared lines. The advisory provides no interim mitigations, does not explain the exploitation chain and does not state whether patching remediates already compromised systems. No public exploit or confirmed exploitation was found at the time of reporting. The flaw was reported by researchers Ali Mustafa (rz1027) and abed1526.
Stealthy Implants, Zero-Click Worms and Identity Lures
Sophos published an analysis of F5 BIG-IP malware that hides a PHP web shell entirely in memory on Access Policy Manager appliances. The code infects the Apache httpd binary, hooks an APR loader function and rewrites file-open and mapping calls so that three APM webtop scripts appear to include a web shell only in memory, allowing it to evade file scanning and known-good comparisons. Sophos links the activity to the c05d5254 cluster and to appliances vulnerable to CVE-2025-53521, which F5 patched in October 2025. An installer component disables SELinux and can infect installation images, potentially persisting across upgrades, while a local socket at /run/bigtlog.pipe can provide an interactive bash shell. F5, the UK NCSC and Ireland's NCSC recommend investigating for compromise even on patched systems, collecting qkview reports, running integrity checks and considering rebuilds where full investigation is not feasible. The exploitation timeline and the attacker's identity remain unknown.
Researchers from Calif disclosed WeWorm, a WeChat zero-click worm that abuses a memory corruption flaw in the app's VoIP stack to achieve remote code execution on Android and iOS. An incoming WeChat call triggers the exploit without user interaction, giving the attacker control of the account to read and send messages and place calls. The attacker must be on the victim's friend list, but compromising one contact allows the worm to propagate. Calif found the flaw using large language models combined with custom analysis. Tencent released patched versions, Android 8.0.77 and iOS 8.0.76, and the researchers warn that chaining with other platform vulnerabilities could lead to full device takeover.
Microsoft Security Research is tracking passkey-themed social engineering observed since May 2026. Attackers impersonate IT support via phone calls, SMS or forwarded Teams messages, then use phishing sites or device-code flows to capture session tokens or induce OAuth approvals. After access, they add attacker-controlled authentication methods for persistence and use Microsoft Graph, SharePoint, OneDrive and Outlook Web to enumerate data, often with Node.js tooling and rotating proxy infrastructure. Because initial contact often occurs on unmanaged devices, Microsoft recommends correlating reported calls and texts with sign-in, device-code, token and authentication-method changes, revoking sessions and removing unauthorized MFA methods.
Group-IB detailed how the Gigabud banking trojan now works with Vwork, a weaponized fork of the open-source Shelter app, to clone banking apps into Android work profiles, separating fraud from earlier malware alerts. Samples target 11 countries, including Brazil, Mexico, Thailand and Turkiye. Between February and July 2026, Group-IB observed about 1,469 compromised devices and 1,281 potentially compromised logins in Indonesia, with estimated losses near $960,939. Banks are advised to monitor signals such as unexpected work profiles and unexplained accessibility access, treating two or more as high risk, and to apply device binding.
Identity Data and Machine Credentials Under Pressure
The ShinyHunters extortion group claims a Florida DMV breach affecting the Driver and Vehicle Information Database (DAVID), alleging theft of more than 200,000 driver records via a password-reset vulnerability used to take over DMV accounts. The group set a September 11 negotiation deadline. The Florida DMV has not publicly confirmed the breach, and the incident is distinct from the Nexus exposure involving scanned licenses collected through IDScan's service. In related commentary, Bruce Schneier discusses a driver's license listing on the dark web offering 153 million records, arguing that AI models accelerate vulnerability discovery and that retention of primary ID documents should be limited to tasks that genuinely require them, with such data treated as crown-jewel assets.
SpyCloud's Identity Threat Report, based on a survey of 750 security leaders, finds that compromised non-human identities, including AI agents, service accounts, API keys and tokens, accounted for 31% of intrusions, compared with 17% for social engineering. While 95% of organizations believe they have adequate visibility into NHIs, only 36% monitor them, and only 56% have formal processes to govern AI tool privileges. Check Point's August threat landscape reports an average of 2,422 weekly attacks per organization, up 22% year over year, 1,042 ransomware attacks, nearly double the prior year, and phishing in 1 of every 112 emails. Average GenAI prompts per user rose to 106, and 86% of organizations using GenAI regularly experienced risky prompt activity, with healthcare showing the highest exposure rate at 1 in 25 prompts.
AI tooling also featured in a critical DeepSeek Harness flaw, CVE-2026-82533, scored 9.4 by VulnCheck. A sandboxed coding agent could call the tool's unauthenticated local web interface and switch its session to danger-full-access mode, disabling file-write confinement and approval prompts with a single shell command. Versions 0.1.1-rc.2 and earlier are affected; the fix, which adds a one-time-token exchange and signed cookie, shipped in 0.1.2-alpha.2 and is included in 0.1.2-rc.1. Until upgrading, users should stop the web interface when not in use, remove tunnels or port forwards to it, and verify the harness version bundled in third-party desktop wrappers.
AWS Expands Certificate, RAG and Lambda Capabilities
AWS made Private CA connectors for Kubernetes, as a managed Amazon EKS add-on, and for Active Directory available in AWS GovCloud (US-East) and (US-West). The EKS add-on integrates with cert-manager to automate certificate issuance and renewal for ingress TLS and service meshes such as Istio and Linkerd, while the AD connector supports smart card authentication, LDAPS and 802.1x, with private keys protected by FIPS 140-3 Level 3 HSMs. AWS also added a Bedrock Confluence connector that lets Amazon Bedrock Managed Knowledge Base crawl and incrementally sync self-hosted Confluence Data Center content, with filters by space or content type.
For serverless workloads, Lambda Managed Instances now support Graviton5 instances (C9g, C9gd, M9g and M9gd), which AWS says offer up to 25% better compute performance than Graviton4; functions and dependencies must be Arm-compatible to benefit. Lambda Managed Instances also gained a 90-minute timeout for asynchronous and Event Source Mapping invocations, a sixfold increase over the previous 15-minute limit, while synchronous invocations remain capped at 15 minutes.