My PGP key for secure contact:
74E6 EEFD E178 BD87 74FB 8CA3 C0A9 6976 2554 EE58
https://t.co/vSXii5vfpK
Check the fingerprint before you trust it. Don't hate the hacker, hate the code.
CISA confirms exploitation of Puppet CVE-2026-115731.
An authenticated agent can compromise the server managing it.
A valid certificate proves identity. It doesn't make the data trustworthy.
Patch the control plane. Review certificate trust.
https://t.co/AR6AuGyMRz
The FBI warns ARTEX is accelerating real intrusions. Some reached privileged access in under 3 hours.
AI didn't invent the vulnerabilities. It shortened the time needed to exploit them.
The weaknesses are familiar. The clock isn't.
https://t.co/l7Nce8ETAV
N-central servers were compromised even after Hotfix 2, according to new research.
The 4EverAble2Shell chain requires an internet-exposed Legacy API.
Check configuration. Investigate compromise. Don't confuse a patched version with a secure system.
https://t.co/M7iF9wwqPm
288 endpoints. 48 organizations.
Attackers abused ScreenConnect management access to deploy rogue remote clients.
The entry point is still under investigation. The persistence problem isn't.
Removing an extension doesn't remove an intruder.
https://t.co/YEyp71OuJK #MSP
OpenSSH CVE-2026-123158: a password-hashing error can become successful authentication on affected musl systems without PAM.
Patch. Restart sshd. Verify the actual configuration.
An error is not an authenticator.
https://t.co/DGRwK1WYd6
Crédit Mutuel Arkéa says attackers came in through a service provider's platform, reached the bank and used its own account for transfers out.
Supplier risk isn't just who can connect. It's what that connection can make you do.
https://t.co/o730lDl7SB #SupplyChain
Every zero-day in Volt Typhoon's Silent Train hit a platform that already manages other systems.
On Quest KACE the actor never moved laterally. It used the server's own admin features to take credentials.
Count what controls your endpoints.
https://t.co/Fvb5maqSgP #ZeroDay
LiteLLM's real repo is BerriAI/litellm. The malicious 1.107.0 was built from Berri-AI/litellm.
Stolen developer tokens pointed the official pipeline at the lookalike. It signed and shipped the result, and provenance recorded the fork.
https://t.co/lYyv2lBm6Z #SupplyChain
Chrome and Firefox stopped trusting eight compromised eMudhra CA keys within hours.
A Go binary keeps trusting them until it is rebuilt. Many CA bundles do until they are updated.
Your browser is not your only TLS client.
https://t.co/fsqNXOjkH1 #PKI
Your backup plugin may carry a remote control you never switched on.
UpdraftCentral keys were enrolled by default and never expired. Stolen ones pushed backdoors to WordPress sites.
Check: wp option get updraft_central_localkeys
https://t.co/yVgb3pmk9D #WordPress
Disabling the compromised vendor app stops new sign-ins.
It does not remove a secret already added to another app in your tenant. ConnectWise says so itself.
Check every app credential added since Oct 13.
https://t.co/R3AAoQtXBn #EntraID
The malicious AOMEI update configuration sometimes existed for only 5 minutes.
Then the server looked clean again.
That's the lesson: a clean vendor server now does not prove a clean endpoint then.
Trust needs historical evidence.
https://t.co/buM0TBGaK5 #SupplyChain
Revoking access is not deleting data.
In Lørenskog, some managers took lost access to an old storage area as a sign the data was gone. It wasn't. ~1.3M files later surfaced in a leak.
Deletion needs evidence.
https://t.co/aasi0RQcin #DataSecurity
Cisco's exploited Secure Email Gateway flaw is delivered as an email attachment—the exact thing the gateway exists to inspect.
Root access. Possibly erased evidence.
Capture forensics before you patch.
https://t.co/CmRP1czgiD #EmailSecurity
A valid TLS certificate can still begin with a false premise.
Attackers compromised three ccTLD registries and obtained unauthorized certificates.
Crypto worked. The identity evidence underneath it did not.
https://t.co/HgHsQAkclp #DNS#TLS
8,547 wind and solar systems were found exposed to the internet across Europe.
Not 8,547 hacks. That's the point.
Reachability comes before exploitation. Know what the internet can see before someone else maps it for you.
https://t.co/ChKuCflyIO #OTSecurity
8.8M Danish CPR records. The access itself was legitimate.
That's the lesson: authentication proves who gets in. It doesn't prove what happens next.
Monitor behavior after login. Bound access. Detect enumeration.
https://t.co/yrGULzxgxf #IdentitySecurity
Some security works because finding the weakness costs more time than anyone bothers spending.
Agents change that equation.
If your control depends on attacker inconvenience, it isn't much of a control.
https://t.co/XxgzjCKuCK
New: a CVSS 4.0 and 3.1 calculator.
CVSS tells you how bad. This tells you how far.
Score the flaw, then check exposure, credentials, trust boundaries and control planes.
Runs in your browser. Nothing leaves the page.
https://t.co/VsRkVzjpmO #CVSS#infosec
iRhythm knew in June that data was exfiltrated from hosted business apps. Patient notifications began in October.
“Business application” is an internal label.
Sensitive health data makes the security classification.
https://t.co/7KKJgylIfJ #HealthData#Privacy