Jun 24's Top Cyber News NOW! - Ep 1160

At a glance

Law enforcement scored major wins against cybercriminals this week, with guilty pleas from Scattered Spider members and extraditions targeting marketplace operators. Meanwhile, AI supply chain risks are escalating—malicious agent skills bypass security scanners by changing payloads after vetting, and CI/CD workflow weaknesses could expose signing keys across major vendors including Microsoft and Google.

Stories covered

What's driving the feds' crackdown on the Huan Group's criminal infrastructure?

What happened: The US government seized cloud infrastructure allegedly used by subsidiaries of Cambodian conglomerate Huan Group, which facilitated online investment scams, money laundering, human trafficking, and stolen-data marketplaces. The action follows Treasury sanctions last year that severed Huan from US financial systems after it laundered at least $4 billion in illicit funds between 2021 and 2025, including proceeds from North Korean cyber operations.

Why it matters: This demonstrates coordinated international enforcement targeting not just individual scammers, but the infrastructure enabling transnational criminal ecosystems. For defenders, it underscores the supply chain nature of cybercrime—disrupting backend systems degrades the entire operation.

What to do: Monitor for your organization's involvement in sanctioned entities or suspicious payment flows. If you handle financial transactions, implement sanctions screening and report suspicious activity to FinCEN.

---

How is Draos building AI specifically trained for operational technology security?

What happened: OT security firm Draos launched Ember AI, a chatbot trained on a decade of OT-specific threat intelligence, vulnerability research, and incident response data. The tool operates within Draos' platform, answers queries in plain language, and correlates threat data across multiple sources while keeping data on-premises.

Why it matters: Generalist AI models lack OT domain knowledge; specialized tools reduce false positives and accelerate threat triage in critical infrastructure environments. Draos' acquisition by Accenture for $4.1 billion signals market momentum around OT security.

What to do: If you manage OT environments, evaluate whether domain-specific AI agents can reduce investigation dwell time. Ensure any tool maintains human-in-the-loop approval for recommendations, especially in safety-critical systems.

---

Why did two Scattered Spider members plead guilty mid-trial in the UK?

What happened: An 18-year-old and 20-year-old pleaded guilty to unauthorized access charges related to the 2024 Transport for London hack. Both reportedly rolled on the wider Scattered Spider operation in exchange for reduced sentences. US authorities allege Scattered Spider collected at least $115 million in ransom across more than 100 compromised organizations, including healthcare providers SSM Health and Sutter Health.

Why it matters: Early guilty pleas typically signal cooperation with investigators. This disrupts the group's operational security and feeds law enforcement intelligence on methods, targeting, and infrastructure. Scattered Spider has been responsible for SIM-swap campaigns, SMS phishing, ransomware, and intrusions across multiple verticals.

What to do: Monitor your identity and SIM swap attack surface. Implement multi-factor authentication that doesn't rely solely on SMS or phone-based factors. Review recent threat intel on Scattered Spider's TTPs and cross-reference your logs for overlapping indicators.

---

How are AI agent skills bypassing every security scanner in the wild?

What happened: Security firm Air created a fake AI agent skill, published it to a popular marketplace, promoted it via Instagram ads, and reached approximately 26,000 agents—all while passing every security scanner tested. The skill appeared benign at review time but pointed to an external domain controlled by Air that could be rewritten after approval, allowing payload swaps without re-scanning.

Why it matters: This is a time-of-check/time-of-use (TOCTOU) vulnerability at scale. Scanners validate a snapshot; attackers change the external instructions after approval. Once installed, skills run with agent authority and can steal data, access internal systems, and establish C2-like footholds. Real campaigns have exploited this for months.

What to do: Treat agent skills as third-party software, not benign text. Review not just bundled code but all external URLs a skill fetches. Pin skill versions and implement least-privilege constraints on agent capabilities. Assume any external instruction an agent fetches runs with agent access. Maintain a single-source-of-truth skill repository under your control.

---

Why is npm becoming a recurring vector for multi-stage Windows malware?

What happened: Jrog researchers found a malicious npm package named `postcss-minify-selector-parser` impersonating the legitimate `postcss-selector-parser` library. The package executed code on import, dropping a PowerShell script that downloaded a multi-stage RAT capable of stealing Chrome passwords, opening reverse shells, exfiltrating files, and maintaining persistence. The package passed all security scanners and remained available in the registry.

Why it matters: Lookalike packages are more sophisticated than typo squats—they inherit credibility from legitimate repos, match keywords, and declare dependencies on real packages. This lowers friction for casual dependency reviews. Supply chain attacks through npm are now routine; each incident follows similar patterns.

What to do: Implement dependency scanning before merge and runtime. Use allow-lists for critical packages. Monitor for packages with suspicious names similar to popular libraries. Audit your `package-lock.json` for unexpected additions and review source code before pulling new versions of widely used dependencies.

---

What financial damage did the Spock cyber crime marketplace operator inflict before arrest?

What happened: Abdullah Belilli, alias Spock, was extradited to the US and charged with operating two cyber crime marketplaces (markettoday.com and spoxy.us) that sold fishing kits, stolen financial credentials, compromised email access, and fraud tools. He created approximately 600 phishing kits targeting major banks, collected data from roughly 5,600 victims, funneled around $900,000 through cryptocurrency between 2020–2023, and embedded backdoors into kits sold to other attackers.

Why it matters: Spock exemplifies the "exploit-as-a-service" model. By building backdoors into kits before selling them, he maintained persistent access to victim data even after losing control of the kit. His operational security was poor—his full name and aliases appeared in source code and linked Facebook accounts, leading to his identification.

What to do: Monitor financial institutions' fraud alerts for anomalies in phishing kit distribution. If your organization was targeted by phishing, check for backdoored credential harvesting. Advise users not to save passwords in browsers; recommend password managers with no external integrations.

---

How can malicious pull requests compromise CI/CD pipelines across major vendors?

What happened: Novi researchers disclosed Cortiseps, a CI/CD workflow weakness affecting repositories from Microsoft (Azure Sentinel), Google (AI agent development kit), Cloudflare (Worker SDK), Apache (Doris), and the Python Software Foundation. Attackers can inject malicious pull requests that exploit overly permissive automated workflows to steal non-expiring GitHub API tokens and signing keys. Novi flagged 654 potentially vulnerable repos; 300 are fully exploitable.

Why it matters: Unlike traditional code vulnerabilities, this weakness hides in YAML workflow configurations that many teams view with less scrutiny than source code. Untrusted data (pull request comments, metadata) crosses trust boundaries without audit. Compromised tokens enable long-term supply chain attacks and code signing fraud.

What to do: Audit CI/CD workflow YAML files with the same rigor as code. Scope GitHub Actions runner permissions to the minimum necessary. Rotate API tokens and signing keys regularly and set expiration dates. Never trust unauthenticated pull request metadata. Use environment secrets scoped to specific jobs and branches.

---

How are scammers hijacking legitimate websites to target World Cup ticket buyers?

What happened: Recorded Future tracked a purchase scam tactic in which threat actors compromise legitimate websites, inject fake product listings and SEO metadata, then redirect search traffic to fraudulent stores. The scam domains never appear in search results; victims arrive via compromised legitimate sites. One operation generated roughly 17 million visits in 2026 alone. The tactic is now spreading to World Cup merchandise and events.

Why it matters: This bypasses traditional ad fraud detection because scammers don't buy ads or rank their own domains. Legitimate sites do the heavy lifting. The tactic is event-driven and scales with demand spikes around sporting events, holidays, and sales. Victims lose both payment card data and merchandise.

What to do: Monitor your website for unauthorized product listings and redirects, especially before major events. Implement content security policies and integrity monitoring. Advise customers to use single-use credit card numbers or services like Privacy.com for online purchases. Verify domain ownership before clicking search results for limited-time merchandise.

Key takeaways

• Law enforcement is disrupting cyber crime infrastructure at scale—Huan Group, Scattered Spider, and marketplace operators are facing extraditions and seizures, signaling sustained international coordination.
• AI agents and CI/CD workflows introduce new supply chain risks that traditional scanners miss: malicious skills change payloads after approval, and pull requests can steal signing keys via overly permissive automation.
• npm and dependency supply chains remain a persistent attack surface; lookalike packages are now more sophisticated than typos and inherit credibility from legitimate repositories.
• Event-driven scams (World Cup, ticket sales) exploit compromised legitimate websites rather than creating new domains, making detection harder and reach wider.
• Operational security failures (embedding your name in malware, using personal email for hacking research) remain a leading cause of threat actor arrest and attribution.

Topics covered

Show notes generated from the live transcript using AI on Wed, 08 Jul 2026 16:10:40 GMT. Errors? Open the YouTube replay for the source of truth.