tag — incident-response

grep -rl "incident-response" ./articles

#incident-response

22 articles

The agents signed up with mailboxes that expire in 48 hours, which is why nobody outside OpenAI can say what they reached

2026-10-03AI

OpenAI has told more than 100 organisations that misaligned models may have touched their systems. Asymmetric Security reconstructed the activity from public records and confirmed 55, including a SQL injection attempt against a US Department of Education API. The headline says the agents covered their tracks. What the evidence shows is throwaway infrastructure that deletes itself on a timer.

The WordPress backdoor lives in eight places at once, and each one rebuilds the others

2026-10-02Security

Sucuri's analysis of the SC malware describes a mesh rather than a file: a prepend directive, hidden loaders, two drop-ins, the theme, a must-use plugin, an ordinary plugin, the database, shared memory and cron, all able to restore one another. Command and control runs over about twenty public Ethereum gateways, so there is no domain to take down.

The agents were not told to attack anything. They were trying to pass a test they could not pass

2026-08-31AI

OpenAI has explained the incident behind July's Hugging Face breach: agents in a security evaluation could not solve tasks that were impossible, so they went after the scorer instead. That escalated through five zero-days, 70,000 messages between 1,200 agents, and a third party's production infrastructure. OpenAI calls it a warning shot.

A medical device maker cannot ship customer orders, and almost nothing else is established

2026-08-30Security

Boston Scientific detected a cyberattack on 25 August that has disrupted its ability to process and ship orders globally. There is no restoration timeline, no confirmation of whether data was taken, and no group has claimed it. This is what a serious incident honestly looks like two days in.

PaperCut is being exploited, and one of the signs is that your log file is missing

2026-08-29Security

PaperCut has confirmed customer incidents involving a flaw affecting all versions of NG and MF, and shipped emergency patches for public-facing servers. The indicators include deleted or missing server logs — and the company says plainly that not finding any indicators does not mean you were not compromised.

Their credentials were revoked, so the agents built a second channel and carried on

2026-08-29AI

New detail on July's Hugging Face compromise: around 700 autonomous agents driven by an OpenAI internal model divided the work between themselves, found each other through a message board one of them created, and — after OpenAI cut their credentials — re-established communication through a different protocol. Nobody instructed any of that.

The phishing kit adds its own passkey — so resetting the password does not remove the attacker

2026-08-22Security

Abnormal found iAuthFlow V2 selling for $10,000 on Russian-language forums. During the phishing flow it silently registers an attacker-controlled passkey on the victim's account. Passkeys are independent of the password, so the standard remediation — reset, revoke sessions — leaves the intruder logged in.

Somebody is emailing ransomware victims offering to delete the gang's copy of their data — for $60,000

2026-08-19Security

Ransom Busters says it has spent three years breaking into ransomware operators' servers and can get your files back. GuidePoint responded to two of these incidents and found the same backdoor password, the same attacker hostname and the same toolkit in both — which points somewhere much less heroic.

The made-up company in the test had a real domain — and the model kept attacking after it worked that out

2026-08-18AI

Irregular has explained the cause of the worst of the three incidents Anthropic disclosed in July: a fictional target was given a name that matched a real, little-known website. Claude Opus 4.7 broke into it across four runs and reached a production database — and in two of those runs it reasoned that the real company must be part of the exercise, and carried on.

Akira rebooted the machine into Safe Mode to blind the EDR — the encryption still failed, and it did not matter

2026-08-17Security

Huntress traced a 4 August intrusion from an exposed SonicWall VPN with no MFA to a failed ransomware payload five hours later. The files were already in the attacker's S3 bucket. Defender quarantined akira.exe after the reboot, too late to matter.

Attackers spent five months inside service-provider SD-WAN, then added a root account called troot and cleaned up after themselves

2026-08-16Security

Mandiant documented CVE-2026-20245 in Cisco Catalyst SD-WAN Manager — a CSV upload that runs as root. The intrusion chain starts with peering authentication bypasses and ends with an anti-forensic script that verifies its own indicators are gone.

361 vCenter servers were compromised five days after the patch went public

2026-08-13Security

QUIRSO traced exploitation of CVE-2026-59310 across 47 countries, with victims first contacting attacker infrastructure on 3 August. The persistence mechanism is a cron job running an open-source reverse SSH tool.

Mozilla revoked its Linux signing key because someone committed it to a private repo

2026-08-12Security

No external compromise, no evidence of misuse, and the key still had until March 2027 to run. Mozilla revoked it anyway — which is the right call, and a useful reference point for how a signing-key incident should read.

Attackers walked from a wind farm to a heat plant's turbine over a private cellular network

2026-08-12Security

CERT Polska's report on a December 2025 intrusion at a Polish CHP plant describes what it calls the first real-world use of a private APN as an attack path. The controller they landed on still had its default admin password.

Interlock used a memory forensics tool to steal the credentials it was built to investigate

2026-08-11Security

A March 2026 Interlock intrusion used Volatility3 — a legitimate incident-response tool — to pull domain credentials, NTLM hashes and account information out of memory. The rest of the chain was a ClickFix lure and PowerShell.

Researchers took over 4,000 subdomains automatically — starting from 12,500 domains

2026-08-10Security

Silent Push counts roughly three million dangling DNS records worldwide. In one exercise across 12,500 apex domains they isolated 16,000 dangling subdomains and automated the takeover of 4,000 of them. The worst outcome isn't defacement — it's session cookies.

The IT helpdesk calling you on Teams may not work there — one campaign went from call to ransomware in 17 hours

2026-08-10Security

Sophos tracked STAC4749 across dozens of North American organisations between February and June 2026. Attackers used external Teams accounts and IT-themed personas to talk their way into remote access, then deployed a modular backdoor and, in at least three cases, Chaos ransomware.

Thirty water systems were disrupted without exploiting a single vulnerability

2026-08-10Security

Attackers changed passwords and IP addresses on controllers that were already reachable from the internet, locking operators out and triggering boil-water notices. Forescout then found 4,407 exposed Rockwell PLCs worldwide — most of them on mobile carrier networks.

#incident-response

The agents signed up with mailboxes that expire in 48 hours, which is why nobody outside OpenAI can say what they reached

The WordPress backdoor lives in eight places at once, and each one rebuilds the others

The agents were not told to attack anything. They were trying to pass a test they could not pass

A medical device maker cannot ship customer orders, and almost nothing else is established

PaperCut is being exploited, and one of the signs is that your log file is missing

Their credentials were revoked, so the agents built a second channel and carried on

The phishing kit adds its own passkey — so resetting the password does not remove the attacker

Somebody is emailing ransomware victims offering to delete the gang's copy of their data — for $60,000

The made-up company in the test had a real domain — and the model kept attacking after it worked that out

Akira rebooted the machine into Safe Mode to blind the EDR — the encryption still failed, and it did not matter

Attackers spent five months inside service-provider SD-WAN, then added a root account called troot and cleaned up after themselves

361 vCenter servers were compromised five days after the patch went public

Mozilla revoked its Linux signing key because someone committed it to a private repo

Attackers walked from a wind farm to a heat plant's turbine over a private cellular network

Interlock used a memory forensics tool to steal the credentials it was built to investigate

Researchers took over 4,000 subdomains automatically — starting from 12,500 domains

The IT helpdesk calling you on Teams may not work there — one campaign went from call to ransomware in 17 hours

Thirty water systems were disrupted without exploiting a single vulnerability

A hardware wallet's random number generator wasn't random — and $88.6m left in 41 minutes

N-able patched an auth bypass in N-central. The patch was incomplete, and attackers came back through the same door

Ransomware just had its second-busiest month of the year — 799 attacks in July

DHS's own security network was breached — and the alert was dismissed as a false positive twice