Jun 26's Top Cyber News NOW! - Ep 1162

At a glance

A mixed news day: Madison Square Garden held firm against Shiny Hunters' ransom demand and the group leaked 26 million visitor records anyway, Cal Water found Iranian hackers only accessed non-critical systems, and a novel macOS malware called Gaslight uses prompt injection to evade AI-assisted malware analysis. Also covered: federal SASE guidance, a new sub-1nm chip from IBM, and a 29-year-old memory leak in the Squid proxy server.

Stories covered

Did Madison Square Garden do the right thing refusing Shiny Hunters' ransom demand?

What happened: Shiny Hunters claimed it breached Madison Square Garden on June 12, threatened to expose 26 million records unless ransomed, and followed through on June 16 by leaking customer data, internal emails, celebrity contacts, facial recognition profiles, and security reports tied to venue visitors.

Why it matters: MSG's refusal to negotiate sends a signal that ransom-paying won't guarantee data stays private. But the leak exposed sensitive surveillance data collected from visitors—faces, security details, addresses—raising questions about data retention practices for transient venue attendees. Multiple class action lawsuits have already been filed.

What to do: Audit what visitor and customer data you actually need to keep, set aggressive retention limits, and treat surveillance-related data as high-value assets in your risk calculus.

How bad was the Iranian hack on California Water's critical infrastructure?

What happened: Cal Water and investigators including Mandiant found the breach was limited to unauthorized access to user accounts at two third-party service provider platforms. Handala, believed to be an Iranian state-sponsored front, leaked 5 GB of data but did not gain access to operational technology (OT) or industrial control systems (ICS) running the water utility.

Why it matters: This is a lower-impact outcome than initial headlines suggested—no active compromise of water treatment systems themselves. However, the breach started with stolen credentials, likely via info stealer or phishing, which is a persistent weak link in supply chain access patterns.

What to do: Monitor for now. Stay tuned on Cal Water's fuller investigation timeline. For your own critical infrastructure: isolate OT from IT networks, enforce strict credential hygiene on third-party service accounts, and assume supply chain partners will be compromised.

What does CISA's new SASE guidance mean for federal agencies moving to zero trust?

What happened: CISA released vendor-agnostic guidance to help federal agencies migrate from legacy TIC 2.0 (centralized internet gateways) to TIC 3.0 using secure access service edge (SASE) technology, bundling SD-WAN, firewalls, CASB, and zero trust network access into distributed cloud services.

Why it matters: TIC 2.0 created traffic bottlenecks that slowed remote and branch users. SASE distributes security closer to users but creates a visibility gap: agencies lose telemetry at central inspection points. CISA now requires agencies to feed equivalent data to CLAW (comprehensive log aggregation warehouse) instead. Also notable: CISA is stepping back from universal TLS decryption, favoring encrypted traffic analysis with ML to reduce latency.

What to do: If you support federal customers or operate at scale, begin planning SASE architecture migration now. Ensure you have log aggregation and forwarding to CLAW-compatible endpoints. Test encrypted traffic pattern analysis and machine learning detection in your environment.

What's the risk from the EdgeQuation malicious Edge extension used in ransomware attacks?

What happened: A malicious Microsoft Edge browser extension dubbed EdgeQuation uses native messaging APIs to escape the browser sandbox and deploy a Python-based backdoor. Attackers pose as IT support on Teams, direct victims to fake Microsoft Outlook management pages, and serve download buttons that either deploy malware, copy PowerShell commands to clipboard, or harvest M365 credentials.

Why it matters: This weaponizes trust in native browser-extension APIs and social engineering over Teams. Initial access brokers connected to the Payout Kings ransomware operation are deploying it. Zscaler researchers link the campaign to previous IAB activity, meaning this is a repeatable infection chain.

What to do: Educate users not to install software based on unsolicited Teams contact—advise them to independently verify by calling IT back. Disable or restrict native messaging for Edge extensions in your organization. Monitor for scheduled tasks spawning Edge with unusual flags. Block known malicious extension names in your endpoint detection rules.

Is Squid's SquidBleed memory leak a practical threat after 29 years?

What happened: Researchers and Mythos AI discovered CVE-2024-50382, a 29-year-old memory leak in Squid proxy server that silently leaked plaintext HTTP requests and credentials. It exists in Squid's FTP directory listing parser (code from 1997) and was patched this month.

Why it matters: Squid runs on thousands of corporate, school, and ISP networks. The leak requires plaintext HTTP (not HTTPS) traffic and attacker-controlled FTP server access, so real-world exposure depends on your environment's encryption posture. Most modern traffic is already HTTPS-encrypted. However, if you still cache or inspect plaintext HTTP, session tokens and credentials are at risk of exfiltration.

What to do: Audit your Squid deployment for plaintext HTTP handling. If present, upgrade immediately. Enforce HTTPS-everywhere policies to minimize plaintext exposure. Monitor outbound connections to FTP servers (port 21) as a side control.

How is Gaslight malware using prompt injection to evade AI-assisted malware analysis?

What happened: North Korea-aligned threat actors deployed Gaslight, a Rust-based macOS implant that embeds fabricated system failure messages designed to make LLM-assisted malware analysis tools doubt themselves and abort analysis. It uses a Telegram bot C2 channel for interactive shell access and harvests command histories.

Why it matters: This is the first documented use of prompt injection to disrupt AI-assisted triage. It attacks the agent's perception of reality rather than the sandbox, which is a novel evasion technique. As AI tooling becomes standard in SOCs, expect more adversaries to exploit prompt injection for defense bypass.

What to do: Don't over-rely on single-vendor AI analysis. Pair LLM-assisted triage with traditional sandbox analysis and behavioral monitoring. Sanitize and validate outputs from AI tools before acting on them. Monitor macOS systems for launch agents with labels like `com.apple.system.services.activity` and embedded Python scripts.

Will IBM's 0.7-nanometer chip advance change the security landscape?

What happened: IBM announced a prototype sub-1nm chip (0.7nm) that crams ~100 billion transistors on a fingernail-sized die, delivering 50% better performance and 70% better energy efficiency than its current 2nm chip. Production is several years away.

Why it matters: Denser, more energy-efficient chips will enable more powerful AI accelerators and distributed compute at lower power draw—good for efficiency, potentially enabling edge AI at scale. However, centralized chip manufacturing remains a geopolitical single point of failure.

What to do: Monitor for now. Advocate internally for distributed manufacturing partnerships to reduce supply chain concentration risk. Plan for more capable edge AI workloads in your threat modeling.

Why is Bajaj Auto confirming a ransomware attack but not revealing details?

What happened: Indian automotive giant Bajaj Auto disclosed a ransomware incident affecting its operations and subsidiary Bajaj Auto Technology Limited on Tuesday morning. The company stated mitigation efforts were successful but did not disclose the threat actor identity, whether data was stolen, or if ransom was demanded.

Why it matters: Ransomware incidents are now routine. The vague disclosure pattern—"we found it, we contained it, trust us"—is common but unhelpful for threat intelligence. Without threat actor attribution, defenders can't hunt for similar TTPs in their own environments.

What to do: Maintain offline, encrypted backups separated by trust boundary from production systems. Assume every major vendor will be hit; treat backups as your primary defense, not ransom negotiation.

Key takeaways

• MSG refusing Shiny Hunters' ransom and the group leaking data anyway reinforces the reality that paying ransoms doesn't guarantee data deletion—avoid creating vast surveillance datasets to begin with.
• Cal Water's breach stayed limited to IT/third-party systems because OT/ICS was segmented; isolation remains your best critical infrastructure defense against both espionage and destructive attacks.
• Gaslight's use of prompt injection to confuse AI-assisted malware analysis is a preview of adversaries weaponizing LLM weaknesses; don't treat AI triage outputs as ground truth.
• CISA's shift from universal TLS decryption to encrypted pattern analysis signals practical acknowledgment that full decryption doesn't scale; apply this lesson to your own visibility architecture.
• SquidBleed required HTTPS displacement and FTP access to exploit, but if you still run plaintext HTTP caching anywhere, patch and re-architect immediately.

Topics covered

ransomware, madison square garden, data breach, shiny hunters, prompt injection, squid proxy, cve-2024-50382, gaslight malware, north korea, edge extension malware, sase, zero trust, federal networks, cisa guidance, cal water, operational technology, ot/ics, ibm chip manufacturing, bajaj auto, critical infrastructure

Show notes generated from the live transcript using AI on Wed, 08 Jul 2026 16:09:25 GMT. Errors? Open the YouTube replay for the source of truth.