☕️ McDonald's staff records sold by hacker

Hi there, this is your daily ☕️ Cyberpresso.

In today's Cyberpresso:

🔓 McDonald's staff records sold by hacker

🍎 macOS flaw lets hackers hijack Macs

🇫🇷 France tax hack exposes 678,000 taxpayers

🤖 New botnet hijacks routers as proxies

🛒 SAP Commerce Cloud flaw actively exploited

Plus: 💡 5 strategies & tactics, 🎁 7 other news you might like, 🧰 6 tools, and 📚 5 papers.

🔓 McDonald's staff records sold by hacker LINK

A seller on a data-trading forum is offering what they claim are 1.7 million McDonald's employee records, allegedly pulled from the company's Azure tenant using stolen login credentials.

The seller, TheHatman, posted an 8,000-row sample whose column names like FacsimileTelephoneNumber match Microsoft's Entra ID export output, with genuine McDonald's email domains, the internal mcdonaldscorp.onmicrosoft.com address, and encoding damage consistent with a real Export-Csv run without UTF-8.

The sample contains no passwords or hashes, so the risk is social engineering from full names, job titles, phone numbers, and internal email formats; the report advises treating unsolicited contact that knows your role with suspicion and confirming instructions out-of-band.

🍎 macOS flaw lets hackers hijack Macs LINK

Hackers are breaking into internet-exposed Macs through an authentication bypass in macOS Screen Sharing, tracked as CVE-2026-65400, that CISA rescored on August 14 from 7.1 to critical (CVSS 9.8) and now calls automatable.

The flaw lets an attacker on the network authenticate to Screen Sharing on TCP port 5900 without valid credentials thanks to faulty state management, and in every case reported to NCSC-NL attackers gained root and installed a Monero cryptocurrency miner.

Apple patched the bug on August 6 for macOS Tahoe 26.6.1, Sequoia 15.7.9, and Sonoma 14.8.9, but with public proof-of-concept code available, users who can't update can disable Screen Sharing under System Settings > General > Sharing.

🇫🇷 France tax hack exposes 678,000 taxpayers LINK

France's tax agency confirmed that hackers stole data on 678,000 taxpayers, including private individuals and businesses, in a cyberattack that officials described as more complex than anything they had faced before.

The Directorate-General for Public Finances (DGFiP) said stolen data includes income figures, tax rates, and family circumstances, but does not grant access to secure accounts on impots.gouv.fr; for businesses, exposed data covered SIREN registration numbers and business addresses.

Authorities disclosed no technical details on how attackers broke in or their motivation; the Paris Public Prosecutor's cybercrime unit and OFAC are investigating, and affected taxpayers will be notified from Monday about identity theft and fraud risks.

🤖 New botnet hijacks routers as proxies LINK

A new Mirai-based Linux botnet called Evooo1Bot is hijacking internet-facing gateway devices from Alcatel, NETGEAR, Tenda, Mitsubishi Electric, Telesquare, and D-Link, turning them into SOCKS5 relay nodes to hide the attackers' traffic.

Active since at least July, the malware breaks in by exploiting known vulnerabilities in these devices, though Fortinet found some embedded exploits are wrongly implemented and fail, then downloads one of 12 builds matching the CPU architecture and clears Bash history.

Beyond proxying, Evooo1Bot steals logins via a sniffer capturing HTTP Basic Authentication and Cookie headers, brute-forces SSH using 150 enterprise username-password combinations, and launches DDoS attacks with 16 flood methods including UDP, DNS, SYN, and HTTP floods.

🛒 SAP Commerce Cloud flaw actively exploited LINK

Attackers are actively exploiting a maximum severity (CVSS 10.0) flaw in SAP Commerce Cloud that lets them break in and run code remotely without any login, just days after SAP shipped a fix in its August 2026 Patch Day.

The bug, tracked as CVE-2026-58231, stems from an improper authorization weakness in the platform's core Data Hub Adapter extension, where an attacker abuses a default authentication client and sends crafted input to functions that lack sufficient validation.

Threat intelligence firm Defused detected exploitation attempts against its honeypots three days after the patch despite no known public proof-of-concept, and SAP, now investigating, urges affected organizations to apply the latest security updates.

💡 Strategies & Tactics

> Hazmat: Open-source containment for AI agents: Hazmat runs AI coding agents in a separate account that sees only your project folder, keeping your keys and cloud credentials out of reach.

> Ruby 4.0 Marshal.load RCE Gadget Chain Exposes Critical Deserialization Risk: A new attack turns Ruby's built-in data-loading feature into remote code execution, so developers should stop feeding it untrusted input and switch to safer formats like JSON.

> How Cloudflare detects MCP traffic and helps secure it: Cloudflare One now spots AI agent traffic on company networks and blocks connections that bypass approved servers, since agents can repeat harmful actions faster than humans catch them.

> Escalating a Blind Upload to RCE via Path Traversal into Cron and DNS-Restricted Callback Bypass: Turning a locked-down file upload into server takeover by writing a scheduled task file into a directory that the system's cron scheduler automatically runs.

> Metasploit Wrap Up: Lot of summer shells and fit http profiles: Metasploit 6.5 adds thirteen new exploit modules plus disguisable HTTP traffic and Windows-on-ARM shells, expanding what penetration testers can attack and evade.

Other news you might like

• GLM-5.3 is here with advanced cyber capabilities, and reportedly already found a 'serious vulnerability' in CursorLINK

• Microsoft Plans to End SMS and Voice Authentication for Entra IDLINK

• Plaintiff hid invisible AI instructions in court filings to secretly influence automated reviewLINK

• 12 KB Windows Backdoor Hides C2 Domain in desktop.ini Whitespace to Evade DetectionLINK

• MessiahGPT Unrestricted AI Model Lets Hackers Generate Ransomware and Phishing KitsLINK

• Malicious Google Apps Script Profiles Crypto Victims Before Delivering Signed Windows MalwareLINK

• SafePal data breach impacts 39,798 customers, stolen info for saleLINK

🧰 Trending tools

FireTail: an AI security and governance platform that gives visibility into AI usage across your environments, helping teams detect and secure risks before they cause harm.LINK

BestDefense.io: continuously pentests every deploy, validates which vulnerabilities are truly exploitable, and auto-generates fixes so SaaS teams patch real risks fast.LINK

Sequirly: browser extension that scans prompts and file uploads before they reach ChatGPT, Claude, or Gemini, flagging API keys and personal data.LINK

DeepFrame: an authorized penetration testing studio for fast-moving web apps, delivering deep security assessments with clear reporting and follow-up retests.LINK

Origin: a confidential agentic stack pairing a private LLM gateway, AI IDE, agents, attestation, and sandboxes for regulated teams in defense and finance.LINK

Skill Inspector: scans AI-generated code and skills for security vulnerabilities, delivering automated remediation guidance directly within developer workflows to reduce risk.LINK

📚 Trending papers & reports

Encrypted traffic screening spots malicious industrial network commands hidden inside scrambled TLS traffic, lifting rare-attack detection by ~43 points and hitting over 95% accuracy while running fast enough for real-time use on factory edge devices.LINK

Safety filter tune-ups let companies fix a content moderation model's mismatched judgment calls without retraining it, catching up to 0.81 of previously missed unsafe content and flagging every one of ten simulated attack campaigns in testing.LINK

Apple's Lockdown Mode falls short as protection for high-risk users like journalists or activists, offering little clarity on what threats it blocks, clunky controls, and noisy alerts that annoy more than they protect.LINK

Passkey backup design lets people export and restore locked hardware login credentials onto new devices without exposing private keys, closing a recovery gap that currently forces buying spare authenticators in advance.LINK

Threat intel extraction shows a specialized text-reading system can pull attacker names, targeted industries, and breach locations from messy real-time posts on X with ~89% accuracy, beating fine-tuned chatbots while running faster.LINK

See you tomorrow for a new dose of ☕️ Cyberpresso!