Atarus Research https://atarussecurity.com/research/ Research and analysis from Atarus Offensive Security: how current attacks work, what they exploit, and what stops them. en-us Wed, 29 Jul 2026 00:00:00 GMT An AI agent cheated its way out of the lab and into someone else's production. One step was genuinely new. Four of them were things you already know how to fix. https://atarussecurity.com/research/agent-lab/ Two OpenAI models broke out of a test sandbox and compromised Hugging Face production. One step was a genuine zero-day. The other four were familiar weaknesses. Wed, 29 Jul 2026 00:00:00 GMT https://atarussecurity.com/research/agent-lab/ AI is now on both sides of the attack: the thing we break into, and the thing that breaks in. Here is what actually changed, and what didn't. https://atarussecurity.com/research/ai-on-both-sides/ Six AI coding assistants fell to one booby-trapped repository, and the first AI-run ransomware leaned on an already-fixed bug. Neither needed a novel exploit. Mon, 13 Jul 2026 00:00:00 GMT https://atarussecurity.com/research/ai-on-both-sides/ We get past MFA on most engagements, and we almost never break it. Here is how, and what actually stops us. https://atarussecurity.com/research/mfa/ MFA stops password attacks, so we go around it: stolen sessions, proxied logins, help desk resets and weak fallbacks. Here is what stops us. Fri, 26 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/mfa/ Researchers caught a low-skilled attacker using AI agents to breach 14 companies. The uncomfortable part is that the AI mostly weaponized bugs you already have a patch for. https://atarussecurity.com/research/ai-agents-14-companies/ An attacker with little skill drove AI agents into 14 companies using bugs that already had patches. The recovered sessions show how little skill it took. Mon, 22 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/ai-agents-14-companies/ Five days, ten exploited bugs, almost none of them touched a core app. A week attackers spent walking in through everything else. https://atarussecurity.com/research/edge-week/ In one week, about ten exploited bugs hit VPNs, file transfer, plugins, browsers, switches and npm. Almost none touched a core application. Fri, 19 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/edge-week/ Microsoft patched three more Nightmare Eclipse zero-days on Tuesday, six in six weeks now. The story isn't the bugs, it's how fast a public PoC turns into a real attack. https://atarussecurity.com/research/nightmare-eclipse/ We wrote about Nightmare Eclipse on June 1, when the story was attackers turning Windows Defender against the people it was meant to protect. Wed, 10 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/nightmare-eclipse/ Your network assumes the attacker comes through the firewall. A red team with a thirty-dollar badge cloner starts inside, as a trusted employee. https://atarussecurity.com/research/physical-access/ Legacy 125 kHz badges can be cloned in seconds with about $30 of hardware. A red team that clones one starts inside your network. Thu, 04 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/physical-access/ Attackers are forging a cookie to log into Palo Alto VPNs as any user, admin included. It was rated medium severity, and they were inside four days after the patch. https://atarussecurity.com/research/palo-alto-vpn-cookie/ CVE-2026-0257 is an authentication bypass in Palo Alto Networks GlobalProtect, the VPN built into PAN-OS firewalls and Prisma Access. Wed, 03 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/palo-alto-vpn-cookie/ Sophos found a ransomware crew using AI coding agents to build EDR-evasion malware and automate Active Directory attacks. The tooling works. The AI's claims about beating EDR were hallucinated. https://atarussecurity.com/research/sophos-ai-ransomware/ The reaction to AI in attacker hands tends to skip straight to the ceiling: autonomous malware, machine-speed intrusions, defenders hopelessly outpaced. Wed, 03 Jun 2026 00:00:00 GMT https://atarussecurity.com/research/sophos-ai-ransomware/ CISA added three supply chain CVEs to KEV including the Nx Console that compromised GitHub. The methodology is federally confirmed. https://atarussecurity.com/research/cisa-kev-supply-chain/ CISA added three actively exploited supply chain vulnerabilities to its Known Exploited Vulnerabilities catalog on May 27. Fri, 29 May 2026 00:00:00 GMT https://atarussecurity.com/research/cisa-kev-supply-chain/ A single syscall escapes the browser sandbox and lands SYSTEM. CVE-2026-40369 is patched. The public exploit code is everywhere. https://atarussecurity.com/research/browser-sandbox-escape/ CVE-2026-40369: one Windows kernel syscall escapes the Chrome, Edge and Firefox sandboxes to SYSTEM. Patched 12 May 2026, with public exploit code. Wed, 27 May 2026 00:00:00 GMT https://atarussecurity.com/research/browser-sandbox-escape/ A new open-source framework turns 7 AI coding tools into 50 autonomous bug-hunting agents. Your adversaries now have it too. https://atarussecurity.com/research/ai-bug-hunting-agents/ A researcher going by H-mmer open-sourced Pentest Agent Suite on May 25. Tue, 26 May 2026 00:00:00 GMT https://atarussecurity.com/research/ai-bug-hunting-agents/ The long weekend in review: Glasswing found 10,000 vulnerabilities. The supply chain attacks didn't break for the holiday. https://atarussecurity.com/research/memorial-day-review/ We hold posts over the weekend, so this is the catch-up. The long Memorial Day weekend was not quiet. Mon, 25 May 2026 00:00:00 GMT https://atarussecurity.com/research/memorial-day-review/ Trend Micro patched an actively exploited Apex One zero-day. Compromise the endpoint security server, own every endpoint it manages. https://atarussecurity.com/research/trend-micro-apex-one/ CVE-2026-34926 is a directory traversal flaw in the on-premise version of Trend Micro Apex One, the company's enterprise endpoint security platform. Fri, 22 May 2026 00:00:00 GMT https://atarussecurity.com/research/trend-micro-apex-one/ GitHub confirmed overnight: TeamPCP exfiltrated 3,800 internal repositories via a poisoned VS Code extension. https://atarussecurity.com/research/github-teampcp-repos/ TeamPCP poisoned a VS Code extension a GitHub employee installed. The stolen credentials reached GitHub internal source code repositories. Wed, 20 May 2026 00:00:00 GMT https://atarussecurity.com/research/github-teampcp-repos/ Grafana lost its codebase to Coinbase Cartel, the new offshoot of ShinyHunters, Scattered Spider, and LAPSUS$. https://atarussecurity.com/research/grafana-coinbase-cartel/ Grafana Labs confirmed Sunday that an unauthorized party obtained a token granting access to its GitHub environment and downloaded the company's codebase. Wed, 20 May 2026 00:00:00 GMT https://atarussecurity.com/research/grafana-coinbase-cartel/ Pwn2Own Berlin 2026 paid out $1.3 million across 47 zero-days. OpenAI Codex got hit three times. https://atarussecurity.com/research/pwn2own-berlin-2026/ The three-day contest at OffensiveCon ran May 14 through 16 with TrendAI's Zero Day Initiative awarding $1,298,250 across 47 unique zero-day vulnerabilities. Mon, 18 May 2026 00:00:00 GMT https://atarussecurity.com/research/pwn2own-berlin-2026/ CISA gave federal agencies 48 hours to patch a CVSS 10.0 Cisco SD-WAN auth bypass. UAT-8616 is already in. https://atarussecurity.com/research/cisco-sd-wan-uat-8616/ CVE-2026-20182 is an authentication bypass in Cisco Catalyst SD-WAN Controller (formerly vSmart) and SD-WAN Manager (formerly vManage). Fri, 15 May 2026 00:00:00 GMT https://atarussecurity.com/research/cisco-sd-wan-uat-8616/ 32 lawmakers told the White House this week: AI is now finding bugs faster than companies can patch them. https://atarussecurity.com/research/lawmakers-ai-patch-gap/ A bipartisan group of 32 House lawmakers, led by Rep. Bob Latta (R-Ohio), wrote to National Cyber Director Sean Cairncross on May 13. Thu, 14 May 2026 00:00:00 GMT https://atarussecurity.com/research/lawmakers-ai-patch-gap/ Checkmarx Confirmed on May 11 That a Modified Version of Its Jenkins AST Plugin Was Published to the Jenkins Marketplace Between May 9 and May 10. TeamPCP Took Credit by Renaming the GitHub Repository to "Checkmarx-Fully-Hacked-by-TeamPCP-and-Their-Customers-Should-Cancel-Now" With the Description "Checkmarx Fails to Rotate Secrets Again." This Is the Third Time TeamPCP Has Compromised Checkmarx in Three Months. The Trivy Scanner Started It in March. The KICS Docker Image, GitHub Actions, and VS Code Extensions Followed in April. The Jenkins Plugin Is May. CVE-2026-33634, CVSS 9.4. The Pattern Is the Story. https://atarussecurity.com/research/checkmarx-jenkins-plugin/ A malicious version of the Checkmarx Jenkins AST plugin (2026.5.09) was published to the Jenkins Marketplace on Saturday May 9, 2026 at 01:25 UTC. Wed, 13 May 2026 00:00:00 GMT https://atarussecurity.com/research/checkmarx-jenkins-plugin/