Microsoft confirmed exploitation of a perfect-score flaw in Entra ID, the identity service behind most Microsoft 365 and Azure estates, then fixed it on its own infrastructure with nothing for customers to install. SAP's commerce platform came under attack three days after its patch, with no public exploit code to work from. CISA added a VMware vCenter flaw to the Known Exploited catalogue on 18 August, where a campaign has already reached victims in 47 countries and left persistence behind, and an MLflow flaw on 19 August now being used to steal cloud credentials out of AI engineering stacks. Citrix published a critical gateway authentication bypass on 19 August with no exploitation yet reported, which makes this the rare week where the patch window on a NetScaler flaw is still open.
1CriticalCVSS 10.0
Microsoft Entra ID Deserialisation Flaw Exploited in the Wild — CVE-2026-69836
CVSS 10.0 | Microsoft advisory, 20 August 2026* Deserialisation of untrusted data in Entra ID allows an unauthenticated attacker to execute code over a network. It scores 10.0 because it is network reachable, needs no credentials, requires no user interaction and is low in complexity. Microsoft credited its own principal security engineer Robert Fitzpatrick with the discovery, confirmed exploitation had occurred, and declined to name the actors, the timeframe, the number of tenants affected or what the attackers were after. Because Entra ID is a managed service, Microsoft says it has already mitigated the flaw across its infrastructure: no patch, no KB article, no configuration change for customers.
India exposureIndian BFSI, IT services, GCCs and government departments run authentication, conditional access and single sign-on through Entra ID without operating any of it. The exposure here is not a device to patch but a dependency to account for: an identity provider compromise reaches every application behind it, and the customer has no way to inspect the platform independently.
ActionTreat this as a detection and assurance task rather than a patching one. Review sign-in logs, privileged role assignments and application consent grants for the period before 20 August. Confirm that Microsoft service health and security messages route to the SOC rather than to an unread mailbox. Test that break-glass accounts work and are alerted on. Record the dependency in the risk register, because "no customer action" is not the same as "no customer impact".
SourceMicrosoft advisory, 20 August 2026; Help Net Security, 21 August 2026; The Hacker News and BleepingComputer, August 2026.
2CriticalCVSS 9.8
VMware vCenter Path Traversal Exploited for Persistent Access — CVE-2026-59310
CVSS 9.8 | Broadcom patch 29 July; CISA KEV, 18 August 2026* A path traversal weakness in the vCenter Syslog Server component allows an attacker with network access to write outside the intended directory and execute code without authenticating. Broadcom released fixes on 29 July 2026. The German security firm QUIRSO reports that exploitation began on 3 August, five days later, and counted 361 unique victim IP addresses across 47 countries, with Germany, the United States, Turkey, Iran and France most represented. The intrusions install a cron job that runs reverse_ssh, an open-source tool that dials out to attacker infrastructure. Attribution remains unclear, though the tradecraft has been described as consistent with an APT actor.
India exposurevCenter is the control plane, not just another server. Indian data centres, BFSI private clouds, hospital virtualisation estates, manufacturing plant IT and managed service providers frequently expose vCenter to a wider management network than intended, and one compromised vCenter is administrative reach over every workload beneath it.
ActionMove to Broadcom's fixed builds now if you have not already, given the patch is over three weeks old and exploitation began within a week of it. Patching does not evict an attacker who arrived earlier, so hunt separately: unexplained cron entries, outbound SSH or long-lived encrypted sessions from vCenter, new vSphere SSO administrator accounts, and unusual REST API activity. Restrict vCenter management access to a dedicated administrative network.
SourceThe Hacker News, 12 August 2026; Infosecurity Magazine, August 2026; CISA KEV addition, 18 August 2026.
3CriticalCVSS 9.3
MLflow SSRF Bypass Used to Steal Cloud Credentials — CVE-2026-64849
CVSS 9.3 | CISA KEV, 19 August 2026* MLflow is a Linux Foundation-supported open-source platform for machine learning and LLM engineering, with reported downloads in the tens of millions per month. Its default tracking server exposes the model-registry webhooks API without authentication, including a test endpoint that returns the upstream response body to the caller. That turns the server into a server-side request forgery primitive, bypassing protections added in version 3.10.0, and lets an unauthenticated attacker make it fetch internal, loopback and cloud metadata endpoints and read the answers. watchTowr observed scanning within hours of the CVE being assigned on 17 August, and confirmed incidents where AWS IAM credentials were exfiltrated and then used to enumerate cloud resources, run cryptominers, or create new IAM users for persistence.
India exposureIndian IT services firms, GCCs, fintechs and product companies have stood up MLflow quickly during the current wave of AI work, often on cloud instances managed by data science teams rather than by infrastructure security. Those servers sit inside the cloud account with an instance role attached, which is precisely what makes the metadata endpoint worth reaching.
ActionUpgrade to MLflow 3.15.0 or later. Remove any tracking server from internet exposure and put authentication in front of it. Enforce IMDSv2 so metadata access requires a session token, and scope instance roles to the minimum the workload needs. Review CloudTrail or the equivalent for credential use from unexpected addresses, new IAM principals, and unexplained compute.
SourceBleepingComputer and SecurityWeek, 20 August 2026; CISA KEV addition, 19 August 2026.
4CriticalCVSS 10.0
SAP Commerce Cloud Flaw Attacked Three Days After Patch — CVE-2026-58231
CVSS 10.0 | SAP Security Patch Day, 11 August 2026* Insufficient authorisation checks and input validation let an unauthenticated attacker abuse a default authentication client, submit crafted input to functions that do not validate it, and execute arbitrary code, compromising internal components. SAP shipped the fix on 11 August. Defused recorded exploitation attempts against its honeypots on 14 August, at a point when there was no public proof-of-concept and no reported in-the-wild activity; KEVIntel independently confirmed attempts the same week, and SecurityWeek reported exploit material available by 15 August. Three days from patch to attack, with no PoC to work from, is the number to note.
India exposureSAP Commerce Cloud underpins storefronts and B2B ordering for Indian retail, automotive, FMCG and industrial distribution. These platforms hold order, pricing and customer data and connect back into ERP, so code execution on the commerce tier is rarely contained to the commerce tier.
ActionApply the 11 August SAP patch and confirm it reached every environment, including staging systems that share credentials with production. Restrict the platform's externally reachable endpoints to what the business actually needs. Review access logs from 11 August onward, and rotate credentials used for downstream ERP and payment integrations if anything unexplained appears.
SourceThe Hacker News, 15 August 2026; SecurityWeek, 17 August 2026.
5CriticalCVSS 9.3
Citrix NetScaler Gateway Authentication Bypass Published Before Exploitation — CVE-2026-19490
CVSS 9.3 | Citrix bulletin CTX696939, 19 August 2026* An authentication bypass through an alternative path affects NetScaler ADC and Gateway appliances configured as a gateway, covering SSL VPN, ICA Proxy, CVPN and RDP Proxy, or as an AAA virtual server. On newer firmware the appliance must also have a SAML action configured; on older builds the gateway or AAA configuration alone is enough. The same bulletin covers CVE-2026-19489, a memory overflow scored 8.8 that applies only where SIP ALG is enabled on a Large Scale NAT group and can cause unpredictable behaviour or denial of service. Rapid7 reported no observed exploitation as of 19 August, while noting that authentication bypasses in this product line have historically not stayed unexploited for long.
India exposureNetScaler carries remote access, application delivery and partner portals for Indian banks, insurers, IT services firms and hospital chains, and those appliances are internet-facing by design. An authentication bypass on a gateway is a bypass of the control that everything behind it assumes is working.
ActionUpgrade to 14.1-73.32 or later, or 13.1-63.21 or later; FIPS builds move to 14.1-73.32 FIPS or 13.1-37.277 for FIPS and NDcPP. Check whether a SAML action is configured on gateway or AAA virtual servers, and whether SIP ALG is enabled on any LSN group. Prioritise internet-facing appliances, and verify that cloud marketplace images you deploy from carry a patched build.
SourceCitrix bulletin CTX696939, 19 August 2026; Rapid7, 19 August 2026; Help Net Security, 21 August 2026.
6
Evooo1Bot Converts Routers and Firewalls into Outbound Proxy Relays
No CVE | FortiGuard Labs, 13 August 2026* FortiGuard Labs documented a previously unreported Linux botnet, named for a hardcoded string present in every sample. It reuses the leaked Mirai DDoS engine and adds encrypted command and control, SSH credential scanning, a credential sniffer, interactive shell access, sixteen flooding methods, and checks for debuggers, sandboxes and virtual machines. Its exploit dispatcher works through known vulnerabilities going back as far as 2007 in products from Alcatel, NETGEAR, Tenda, D-Link, TP-Link, Zyxel, Telesquare, Hikvision and Mitsubishi Electric, alongside Atlassian, Kubernetes, PHP-CGI and WSO2. The relay mode deserves attention: rather than opening a listening port, the infected device dials out to an operator-controlled server, making it a durable anonymisation proxy that outbound-permissive firewall rules will not stop. FortiGuard reports activity since July 2026.
India exposureThe exposure sits in branch offices, retail outlets, clinics, campuses and plant networks, where SOHO routers, CCTV recorders and small firewalls run firmware nobody owns and default credentials nobody rotated. A relay generates no ransom note and no outage, so it can sit for months while traffic leaves through an Indian commercial IP address, which is the attacker's reason for wanting it.
ActionInventory internet-facing edge devices, including cameras and recorders procured outside IT. Update firmware, remove WAN-side management interfaces, disable Telnet, and rotate default credentials. Restrict outbound traffic from device VLANs to what those devices genuinely need, and alert on sustained outbound encrypted sessions or proxy-like behaviour from an appliance that should only be talking to a handful of destinations.
SourceFortiGuard Labs, 13 August 2026; The Hacker News and BleepingComputer, August 2026.
Takeaway
The highest-severity items this week were the ones furthest from the patch cycle. Entra ID was fixed without customer involvement, MLflow was installed by data science teams rather than infrastructure teams, and the Evooo1Bot devices sit in no one's asset register. Where a patch did exist, the interval kept shrinking: five days for vCenter, three for SAP Commerce Cloud. Against that, NetScaler is the one genuine opportunity, with fixed builds available and no exploitation reported as of 19 August. Two of the six items end at stolen cloud credentials rather than at the compromised host, which is the direction worth planning for.
This week's vendor sweep named Fortinet, Cisco, Palo Alto Networks, Check Point, Juniper, SonicWall, Sophos, Barracuda, WatchGuard, Zscaler, Citrix NetScaler, Ivanti, F5 BIG-IP, Versa, VeloCloud, Aruba EdgeConnect and Seqrite. Fortinet's 18 August advisories across FortiWeb, FortiClient Windows and FortiManager carry no reported exploitation. Three further KEV additions did not displace the items above but merit patching where they apply: Windows IKE Service Extensions CVE-2026-33824, a double free in IKEv2 fragment reassembly patched in April and escalated to known-exploited on 18 August, which matters wherever Windows terminates IPsec VPNs; SharePoint CVE-2026-55040, a JWT forgery bypass added the same day as exploitation rose after a public proof-of-concept; and TrueConf Server CVE-2026-72529 and CVE-2026-72530, added on 20 August.
