Top 5 Cybersecurity News
This week’s Top 5 Cybersecurity News for November 20, 2026 is about what companies rely on without checking. Microsoft’s own load tests missed a bug that later took Microsoft 365 mail down for almost a day. Google’s Project Zero is removing the grace period between disclosure and exploitation that patch plans were built around. A ransomware crew wiped a virtualization estate with administration features, backups first. North Korean operators used recordings of familiar faces to get people to run a “fix”. And two automakers’ login systems fell to an admin console that still had its documented default password. Five everyday dependencies, one question: what is your company relying on that has not been checked this year?
1) One Latent Bug Took Microsoft 365 Mail Down for Nearly a Day

Microsoft says a bug from a July update to its Microsoft 365 gateway servers, a Teams failover and reduced capacity combined to cause the November 12 outage, which lasted until 11:10 AM ET on November 13.
Microsoft says a software bug introduced in July, in an update for its Microsoft 365 networking gateway servers, caused the worldwide outage of November 12. According to its preliminary post-incident report, as relayed by BleepingComputer, the incident ran from 11:20 AM ET on November 12 to 11:10 AM ET the next day and affected Outlook, Exchange Online, Teams and Copilot. Microsoft names three factors: the bug itself, which its load tests missed because they ran on weaker hardware with fewer cores; extra traffic after Teams switched to the infrastructure Exchange Online uses about 15 hours earlier; and roughly 14% of gateway capacity that was already offline for an unrelated reason. In under two hours the defect hit the whole fleet, and at the lowest point more than 80% of the servers were unresponsive. As of this writing, Microsoft’s final report is not yet available; we will update this article once it is published.
Teams, OneDrive and SharePoint could switch to other infrastructure and recovered earlier. Exchange Online could not, so mailboxes stayed out of reach. Restarting servers did not help at first, because the backed-up traffic overwhelmed them again. Microsoft also warned of phishing that impersonates Microsoft 365 service health notices and cites the real incident ID EX1550543; it called the activity consistent with credential phishing and said credential collection and malware delivery were not confirmed. Agree in advance how your company communicates when email is down, and check service status only in the admin center, never through a link in a message. Assumption: many companies send password resets and alerts to the same mailbox they would be locked out of.
Nobody attacked anything here, and the outage still lasted close to 24 hours. A defect that sat quietly for months, an ordinary failover and an unrelated capacity loss lined up. That is what supplier concentration looks like in practice: when most of your communication, identity and files run on one provider’s shared plumbing, availability risk belongs in the same register as attacker risk. The phishing that followed within hours shows that attackers watch for these days too.
Read more on: Microsoft Post-Incident Report (365Fanatic summary) · BleepingComputer
2) Google Project Zero Removes the Grace Period Defenders Counted On

From November 13 Project Zero discloses 90 days after reporting, whether or not a fix exists; the deadline falls to 60 days on February 2, 2027 and to 30 days on January 3, 2028.
On November 13 Google’s Project Zero announced that it will publish technical details 90 days after reporting a flaw to the vendor, whether or not a fix exists. The previous “90+30” policy added 30 days after a fix for patch rollout, and that extension is gone. The deadline drops to 60 days on February 2, 2027 and to 30 days on January 3, 2028. Project Zero says AI-assisted discovery and exploitation made the old timeline “clearly insufficient”, and its blog post calls zero patch-to-exploitation time a reasonable, though imperfect, model of attacker behavior. Flaws that are already exploited, already public, or where the vendor asks for earlier disclosure are exceptions. According to Dark Reading, Anthropic will shorten its own disclosure window to 30 days for high and critical bugs from January.
Experts are split. Dustin Childs of the Zero Day Initiative called it “a bold move” and fears patches that do not fully fix the flaw. Katie Moussouris of Luta Security said “a deadline doesn’t magically manifest a fix” and, as quoted by Dark Reading, noted that Anthropic reports 629 of the 5,087 vulnerabilities it disclosed as patched (not independently verified). For your company this means details of an unpatched flaw in software you run may become public on a fixed date. Ask which of your internet-facing products are critical, who decides on a mitigation when no patch exists and how fast a workaround can be applied.
The same week supplied a live example. On November 18, F5 released NGINX 1.31.9 and 1.30.7 for two HTTP/2 flaws, CVE-2026-123007 and CVE-2026-123633, both rated 9.2 (CVSS v4.0, 8.1 under v3.1) in F5’s advisories. F5 says an attacker can crash the worker process or execute code where ASLR is off or can be bypassed. Wiz, one of five teams that found the flaws independently, demonstrated code execution on CVE-2026-123007, which dates back to NGINX 1.9.5 in 2015, and says its AI-based tooling found both. Wiz has not released its exploit and suggests turning HTTP/2 off if you cannot patch at once. The days between a vendor fix and the first exploit were a buffer that patch plans quietly relied on, and a disclosure policy now removes it. Many firms run NGINX without knowing it, inside appliances, ingress controllers or a hosting provider’s stack, so ask where it runs and who patches it. Hypothesis: the gap will show first in products whose vendors are small or slow to publish fixes.
Read more on: Google Project Zero · Dark Reading
3) BlackNevas Wiped Four ESXi Hosts With Administration Features, Not an Exploit

Kaspersky reports a ten-day intrusion that wiped four ESXi hosts, two vCenter appliances and the backups behind them without exploiting a hypervisor flaw; ransom notes named BlackNevas.
Kaspersky’s Global Emergency Response Team published a case on November 13 in which an attacker destroyed a whole virtualization estate over ten days: four ESXi hosts, their virtual machines, two vCenter appliances and the backup infrastructure. No hypervisor vulnerability was exploited. Credentials recovered from systems outside the virtualization layer reached vCenter. A root SSH login to one appliance was followed within 43 seconds by the install of MeshAgent, a legitimate remote management agent, and about four hours later by a Ligolo-ng tunnel named vmware-vdma. The attacker then used vCenter Host Profiles to switch on SSH on the hosts; the service was running 33 seconds after the push. Before the final run, antivirus on the backup environment was disabled, backup services were stopped and recovery systems were encrypted. Incident artifacts indicate that roughly the first 2 KB of the virtual disks were overwritten. Ransom notes named BlackNevas; Kaspersky did not recover the binary, so the match to the BlackNevas and Trigona family rests partly on the command-line style. The victim is not named, and the report does not give the date of the attack.
The access path crossed boundaries many companies treat as separate. Backup infrastructure held credentials that helped administer production, and the vCenter authority that sets host configuration could also reopen SSH. Kaspersky recommends treating recovery as its own security boundary, with independent admin identities, restricted management paths and tested restores, and treating vCenter as a Tier 0 asset. It also advises sending ESXi logs to remote storage, because ESXi keeps them in a ramdisk that is cleared on reboot, so collect them before anyone restarts a host.
Every step used a legitimate function, so no single event looks wrong. The order does: a root login followed by an agent install within a minute, a configuration push followed by a running shell service seconds later. Detection built on single alerts misses this, and the attacker also attacked recovery before production. “We have backups” is therefore a claim about the whole recovery path, including the accounts and systems needed to use it, and not only about the data.
Read more on: Kaspersky Securelist · Gridinsoft Blog
4) FaceStealer: Familiar Faces Became the Delivery Mechanism for a $400 Million Theft

The U.S. government says it disrupted FaceStealer, a North Korean campaign that ran for 18 months, stole more than $400 million in cryptocurrency and used more than 900 secretly recorded webcam videos.
In the week of November 9 the FBI published an alert on FaceStealer, and the U.S. government seized the cloud infrastructure behind the campaign, which BleepingComputer reports ran for 18 months and stole more than $400 million in cryptocurrency. Attackers hijacked accounts on Telegram, LinkedIn, Slack and Discord and used them to invite targets to video meetings. During the call they showed recorded footage of people the target knew, then claimed a technical problem, displayed a fake error and offered a link to a “fix” that installed malware. The malware steals crypto data and the credentials of Zoom, Google Meet and Microsoft Teams, and gives access to contacts, messaging apps and the webcam. While victims waited for a meeting to start, the attacker secretly recorded video that was reused against the next target; U.S. authorities identified more than 900 such videos. Arctic Wolf attributes the campaign to BlueNoroff and identified 677 people in 58 countries whom the group had prepared material against, 84% of them founders and C-suite executives; these are targets, not confirmed compromises.
The FBI says attackers sometimes waited months, or more than a year, after the first compromise before going after a victim, likely until holdings had grown. It also describes a maintainer of a popular JavaScript library with tens of millions of weekly downloads who was lured into a fake Teams meeting with a familiar participant; quick detection stopped that attack, and the FBI confirmed other attempts against developers in the npm and Rust ecosystems. As relayed by BleepingComputer, the FBI advises joining meetings only through official apps with an invite code, verifying unexpected invitations through another channel, keeping the webcam off unless needed and never running a script offered as a fix during a call. Executives and developers with privileged access are the people to brief first.
Recognizing a face or a sending account worked as a trust check for years, and it is the thing being harvested and replayed here. The control therefore has to sit on the action, not on the person: nothing gets installed or run because someone on a call asked for it. The exposure also reaches beyond company systems, because the entry point was personal and professional accounts on consumer platforms that no corporate policy covers.
Read more on: FBI · BleepingComputer
5) A Documented Default Password Gave Researchers Control of Two Automakers’ Login Systems

Shubham Shah and Sam Curry found an OpenIDM admin console on two carmakers’ ForgeRock login servers still using admin and admin; both systems were fixed about two hours after the report.
Researchers Shubham Shah and Sam Curry found that the login systems behind two car brands’ mobile apps, both built on the ForgeRock identity platform (now part of Ping Identity), exposed an OpenIDM administrator console that still used its documented default: the word admin as username and password. The console was blocked from the internet, but the edge and the Java server behind it read the same web address differently, and encoded characters got the request through. Inside, a built-in scheduler ran the researchers’ code, which let them change a target account’s email and password, sign in through the normal app and obtain the tokens the real app uses. They unlocked the doors and started the engine of a test car, on their own and consenting owners’ cars in the U.S. in February 2026. The carmakers fixed the issue about two hours after the report; the write-ups were published in early November. A second attack on the Australian systems used admin credentials left in the app code for a test setup, and The Hacker News reports commands could reach cars in Australia, the U.S. and Canada. The automakers are not named.
This was a deployment error, not a product bug, so there is no CVE and no vendor advisory to trigger an alert. The product’s own hardening guide already tells operators to remove the console or protect it before go-live. For teams running the same software, now sold as PingIDM, The Hacker News lists two checks: look for the Apache Felix console at /system/console and remove it or change its password, and do not rely on edge or WAF path rules to hide an admin page. Changing the owner’s email and password did not log the owner out, and alerts could be redirected, so the owner had little to notice. Neither write-up says whether anyone else used the path.
An identity platform sits underneath every application that signs people in, so one default on its admin side opens every account above it. Nothing here was new or hidden. It was documented for years and survived because the gap between the build documentation and the production checklist belongs to nobody, and a scan for new vulnerabilities will never report it. The same question applies to your SSO, VPN and remote management consoles: who confirmed, after go-live, that the defaults are gone?
Read more on: Ping Identity Documentation · The Hacker News
If this week tells us anything, it’s this:
Each of the five incidents started with something ordinary that people relied on without checking. Microsoft’s lab hardware stood in for production. Patch plans counted on days that vendors will soon stop granting. The backup hung on the same accounts and paths as the systems it protected. A familiar face counted as proof of a real person. A default stayed in place because the step between build and go-live belongs to nobody. No new attack technique was needed; each case used a dependency that worked until the day it mattered.
For boards and CISOs the useful exercise is a short list. For each critical dependency, write down what it relies on and the date it was last checked under realistic conditions: the provider’s availability, the patch timeline, the restore path, the way you verify a person, the state of every admin console. The entries without a date are the exposure. Checking is cheaper than learning the answer from an incident.
DIESEC works on the parts of this list that are hardest to check from the inside. Our experts run penetration tests and red teaming that look for this kind of exposure, from default credentials on admin consoles to recovery paths that share accounts with production. SOC as a Service adds continuous monitoring and 24/7 support, and NovaMDR™ is our managed detection and response solution. For security awareness and human-risk management, the Phished solution helps employees learn to question a familiar face on a call. Our governance, risk and compliance team helps you document which dependencies matter and when each was last checked.
If you would like help checking one of these dependencies in your own environment, for example how your team would work through a day without Microsoft 365 mail, you can reach us on our contact page.

