TRANSKRİPT 🔗
WEBVTT
00:00:10.320 --> 00:00:13.759
Here's your briefing for Thursday, November 12th, 2026.
00:00:15.120 --> 00:00:24.640
Someone is logging into nearly 5,000 hijack GitHub accounts and quietly planting a secret stealer in the build pipelines of every repository those accounts can touch.
00:00:27.920 --> 00:00:30.160
Your build pipeline may have been robbed this week.
00:00:30.400 --> 00:00:42.000
Aikido Security says that starting Tuesday, an attacker used stolen credentials for 4837 GitHub accounts to plant a stealer that grabs every secret and cloud credential a repository's automated workflows can reach.
00:00:42.239 --> 00:00:44.799
It ran in at least 1800 repositories.
00:00:44.960 --> 00:00:55.039
Each attack opened with a harmless-looking test pull request titled Docs Dependency Audit Notes from a branch called WFX Docs sent to more than 61,000 repositories.
00:00:56.079 --> 00:01:04.159
SafeDEP counted more than 8,000 repositories that received malicious commits, some through a package install hook that also runs on a developer's laptop.
00:01:04.320 --> 00:01:11.200
The attacker usually reverted the change within minutes and deleted the workflow run and its logs, so a clean looking history proves nothing.
00:01:11.359 --> 00:01:14.320
Fresh test pull requests were still appearing early Thursday.
00:01:14.480 --> 00:01:15.120
Do this today.
00:01:15.280 --> 00:01:19.920
Search every repository you can write to for that pull request title and that branch name since Tuesday.
00:01:20.079 --> 00:01:27.040
And if you find one, revoke that account's tokens, keys, and sessions, then rotate every secret and cloud role the repository's workflows could read.
00:01:30.240 --> 00:01:32.560
Screen Connect is being used as a worm again.
00:01:32.719 --> 00:01:47.280
Huntress says attackers got into cloud-hosted Screen Connect panels, likely with stolen credentials or hijack sessions, and use the run command feature to push rogue remote clients to at least 288 endpoints across 48 organizations starting Wednesday afternoon.
00:01:47.840 --> 00:01:49.200
It spreads both ways.
00:01:49.359 --> 00:01:58.480
When a technician connects to an infected machine, the malware can jump back to their workstation, steal browser sessions and tokens, and pivot into every other panel that technician can reach.
00:01:58.640 --> 00:02:07.760
ConnectWise says its own software isn't vulnerable, but it disabled five third-party extensions, including LabTech Plus and the RMM Plus family in about 400 partner environments.
00:02:07.920 --> 00:02:08.639
Do this today.
00:02:08.879 --> 00:02:13.199
Disable those RMM Plus and LabTech Plus extensions on any on-premises Screen Connect server.
00:02:13.359 --> 00:02:20.240
Block the indicators in ConnectWise's advisory, and audit your panels for new admin accounts and run commands you didn't send.
00:02:23.757 --> 00:02:33.680
Apache web servers have a critical bug in their HTTP2 support, A flaw in the module that handles that protocol lets an unauthenticated attacker run code and take over the server process.
00:02:33.840 --> 00:02:40.080
Red Hat rates it 9.8 and says public research shows a reliable exploit against a default setup with HTTP2 turned on.
00:02:40.240 --> 00:02:43.840
Wiz says it hasn't seen working exploit code or attacks in the wild yet.
00:02:44.800 --> 00:02:50.159
It was found by Anthropics AI-powered open source scanner and separately by Google researchers.
00:02:50.319 --> 00:02:52.960
The Apache project shipped the fix on Wednesday.
00:02:53.120 --> 00:02:58.719
Red Hat has packages for its last two major releases, while Debian stable and Amazon Linux were still waiting.
00:02:58.879 --> 00:03:02.400
The standalone version of the module is archived and won't get a fix at all.
00:03:02.560 --> 00:03:03.200
Do this today.
00:03:03.360 --> 00:03:15.920
Update Apache to version 2.4.71 or your distro's patched package, and on any server you can't patch yet, remove the H2 and H2C entries from the protocol's setting so it only speaks HTTP 1.1.
00:03:19.575 --> 00:03:23.495
An AI agent just ran a two-week telecom intrusion with almost no human help.
00:03:23.735 --> 00:03:31.974
CISA added a Cisco Broadworks flaw to its exploited list, an XML parsing bug that lets an unauthenticated attacker read configuration data.
00:03:32.159 --> 00:03:43.120
Malbear Labs says a Chinese-speaking group used it in Operation Snowball, an autonomous framework built on a jailbroken DeepSeek model that hit a European telecom provider and 13 companies hosted on it.
00:03:43.280 --> 00:03:49.439
The agent pulled private keys and credentials and walked off with thousands of subscriber billing and device records.
00:03:50.400 --> 00:03:59.680
Malbear says the operator delegated the intrusion and checked in now and then, while the agent chose exploits, worked around block controls, and wrote new tools when old ones failed.
00:03:59.840 --> 00:04:00.640
It's not alone.
00:04:00.800 --> 00:04:09.840
Arctic Wolf ties an open source AI pen testing platform called Antifox to intrusions that went from a web shell to reading production secrets in Amazon's cloud.
00:04:10.080 --> 00:04:10.800
Do this today.
00:04:10.879 --> 00:04:20.639
Confirm Broadworks is on Cisco's August Fix, block its client interface port from the internet, and rotate any keys stored on that host.
00:04:23.840 --> 00:04:27.439
The world's busiest ransomware gang may have just lost its anonymity.
00:04:27.680 --> 00:04:42.000
Brian Krebs reports that Kylin, which claimed at least 365 victims in the past three months, is run by an administrator called Heyze and a decade of old forum handles, reused email addresses and leaked Russian government records point to a 36-year-old man in St.
00:04:42.160 --> 00:04:43.120
Petersburg.
00:04:43.840 --> 00:04:52.240
Kylin pays affiliates at least 80% of each ransom, steals data before encrypting it, and works with English-speaking crews tied to Scattered Spider.
00:04:52.480 --> 00:04:59.439
A name won't stop the attack since he's out of reach in Russia, but it shows how one sloppy old account can unravel careful OPSEC years later.
00:04:59.680 --> 00:05:10.639
Do this today, test the restore from your offline backups, and make sure every remote access and admin login requires phishing resistant multi-factor authentication because that's how Kylan's affiliates get in.
00:05:13.920 --> 00:05:24.079
Hijack developer accounts, hijack support panels, a web server bug found by an AI, an AI that ran the whole break in, and a ransomware boss undone by his own old accounts.
00:05:24.160 --> 00:05:28.079
The tools keep getting smarter, but the way in is still a credential somebody trusted.
00:05:31.920 --> 00:05:32.879
That's the briefing.
00:05:33.040 --> 00:05:36.160
Stay sharp, keep your systems patched, and we'll see you tomorrow.