UPDATE: jscrambler npm compromise hit five versions, not one.

Socket links 8.14.0, 8.16.0, 8.17.0, 8.18.0, and 8.20.0 to the same actor. Jscrambler says a stolen publishing credential was used.

Later versions moved beyond preinstall, so --ignore-scripts would not block them. Upgrade to 8.22.0 and audit affected systems.

Read: https://thehackernews.com/2026/07/compromised-jscrambler-8140-npm-release.html

🚨 Joomla sites running iCagenda or Balbooa Forms face two exploited CVSS 10.0 flaws now in CISA’s KEV catalog.

Both allow attackers to upload PHP files and execute code on affected servers.

What the attacks looked like and what admins should check: https://thehackernews.com/2026/07/icagenda-and-balbooa-forms-joomla-flaws.html

🛑 One open directory exposed three Microsoft 365 phishing operations.

The operations used two paths into #Microsoft 365: Evilginx session theft and device code phishing. One campaign logged 218 captured accounts across 12 countries, 94% of them corporate mailboxes.

Read the full investigation: http://thehackernews.com/2026/07/misconfigured-server-reveals-three.html

⚠️ An attacker used compromised credentials to access a domain-joined Windows Server over RDP, then ran a suspected AI-generated PowerShell script to map Active Directory.

The script was noisy, aggressive, and oddly helpful.

Read how researchers traced it: https://thehackernews.com/2026/07/attacker-uses-suspected-ai-generated.html

AI SOCs do not fail because Claude or Codex are useless. They fail when teams use them for the wrong job.

Research cited in this article says 98% of alerts can be resolved autonomously, while less than 2% warrant human review.

The problem is where the human and agent judgment starts.

Learn more: https://thehackernews.com/2026/07/thinking-fast-and-slow-in-soc-case-for.html

⚡ Meta just filed a patent for an Artificial Intelligence (AI) that listens through your day and logs how you're feeling.

Not a mood check here and there, but a timestamped record of your emotions, keyed to where you are and what you're doing.

And it doesn't stop at your voice.

What else it reads, and whether Meta is actually building it: https://thehackernews.com/2026/07/meta-files-patent-for-ai-that-can.html

🚨 Forg365 targets Microsoft 365 with device code and AitM phishing.

This $400/month service can use a legitimate Microsoft sign-in to authorize an attacker-controlled session, with tools for post-compromise mailbox operations.

How the attack works: https://thehackernews.com/2026/07/forg365-phaas-targets-microsoft-365.html

🛑 [New] One email could rewrite what an inbox-reading AI assistant remembers about you.

MemGhost adds false information to an AI agent’s memory and changes what it says later. In #OpenClaw lab tests, the change could stay hidden from the user.

How it works, and what limits it: https://thehackernews.com/2026/07/new-memghost-attack-plants-persistent.html

"AI changed both sides of this fight.” Stephen Sims is keynoting AND teaching at Network Security 2026, bringing offensive security expertise straight into the classroom. Caesars Palace, Vegas | Sept 21–26.

His code SIMS777 = $777 off any course 👉 https://thn.news/cyber-training-sans

⚡ THN Recap: Before you start the week, know what's out there:

▪️ ShareFile servers pulled offline
▪️ npm package hijacked to drop infostealer
▪️ Critical Zimbra email flaw patched
▪️ GigaWiper backdoor wipes disks
▪️ 1.4M WordPress sites targeted
▪️ Citrix Bleed 2 leads to ransomware
▪️ AI coding assistants tricked into malware
▪️ Django flaw actively exploited
▪️ Fake VPN installer drops RAT
▪️ Helix crew hits SharePoint data

Full recap → https://thehackernews.com/2026/07/weekly-recap-sharefile-threat-citrix.html

🚨 Google and Microsoft pulled ModHeader after researchers found a dormant browsing-history collector inside its genuine store build.

The extension had roughly 1.6 million installs across Chrome and Edge, with most of the collection pipeline already in place.

Here's how it stayed hidden - https://thehackernews.com/2026/07/google-and-microsoft-pull-modheader.html

CrashStealer uses a signed and Apple-notarized macOS dropper to pass Gatekeeper checks.

Once launched, it can steal browser credentials, wallet data, password manager records, files, and keychain material.

How the attack chain works: https://thehackernews.com/2026/07/crashstealer-macos-malware-uses.html

🚨 Hackers spent a year stealing data from #Salesforce environments without exploiting a single platform flaw.

#Microsoft mapped the ShinyHunters-linked activity to trusted OAuth apps, stolen vendor tokens, and misconfigured guest access while sign-in monitoring stayed quiet.

How the three attack paths worked: https://thehackernews.com/2026/07/microsoft-maps-year-long-shinyhunters.html

🚨 148 npm packages disguised as student web proxies turned visitors’ browsers into a DDoS botnet.

The code never ran at install time. It waited inside working proxy sites, beyond scanners focused on install-time behavior.

Here's how the campaign worked: https://thehackernews.com/2026/07/148-npm-packages-disguised-as-student.html

💪 U.S. sanctions hit First VPN, its administrator, and a cryptor seller over support for #ransomware attacks on American businesses, hospitals, financial firms, and local governments.

The VPN was dismantled in May 2026 after allegedly helping attackers hide their origins.

How the operation worked: https://thehackernews.com/2026/07/us-sanctions-first-vpn-service-and.html

🛑 Grok Build uploaded entire Git repositories, full commit history and all, to xAI storage, not just the files the agent read.

A captured bundle contained a file the agent had been explicitly told not to open.

What was uploaded, and which secrets users should rotate: https://thehackernews.com/2026/07/grok-build-uploads-entire-git.html

⚠️ Attackers can validate stolen #Microsoft Entra credentials without generating a successful sign-in event.

Researchers tracked two campaigns using spoofed OAuth client IDs, including one that targeted over 2 million users. App-scoped detections may miss it.

Read why Entra logs may miss it: https://thehackernews.com/2026/07/oauth-client-id-spoofing-lets-attackers.html

🚨 Researchers tested 85 browser crypto wallets representing 35.16 million Chrome Web Store users.

Some linked separate addresses, kept exposing previously granted addresses after logout, or enabled tracking across unrelated sites.

How it works: https://thehackernews.com/2026/07/study-of-85-crypto-wallet-extensions.html

🛑 11 old Microsoft-signed UEFI shims could let attackers with admin access bypass Secure Boot and run code before the OS starts.

Microsoft revoked the vulnerable shims in June 2026, but systems without the revocation update may still accept them.

Read how the bypass works: https://thehackernews.com/2026/07/11-old-microsoft-signed-linux-uefi.html

⚠️ A RabbitMQ flaw can leak an OAuth client secret in one unauthenticated request, potentially giving attackers full broker control in affected setups.

A second bug lets logged-in users read queue and exchange metadata beyond their permissions.

Details: https://thehackernews.com/2026/07/rabbitmq-flaws-could-leak-oauth-secrets.html

⚡ Pentera’s MCP Server gives AI assistants validated attack paths, not just scanner scores.

A critical finding may be unreachable, while a medium-severity flaw may be the one that leads to privileged access.

How validation changes the workflow: https://thehackernews.com/2026/07/how-pentera-turns-ai-security-workflows.html