Jul 7's Top Cyber News NOW! - Ep 1168
At a glance
Seven distinct cyber threats dominated today's brief: Chinese attackers targeting Indian taxpayers via fake tax PDFs, AI agents being manipulated into crypto payments through prompt injection, France mandating quantum-safe encryption by 2027, a critical Linux/Android privilege escalation flaw, an actively exploited Adobe ColdFusion RCE, North Korean supply chain compromises targeting developers, and Opera's new clickjacking defense. The standout: the prompt injection attacks show adversaries weaponizing AI agent capabilities at scale.
Stories covered
Why are Chinese hackers impersonating India's tax authority?
What happened: Researchers at Sacchrite Labs identified a multi-stage campaign called "Operation Dragon Return" using fake income tax department emails and malicious PDFs to deliver a remote access Trojan targeting Indian taxpayers, tax professionals, and corporate finance teams during filing season.
Why it matters: This is a well-established attack pattern pivoted to a new geography. Once the RAT gains execution, attackers own the machine and can deploy any payload—info stealers, crypto miners, spyware, destructive malware. The campaign demonstrates precision: bilingual content, real legal citations, and active payload rotation indicate sustained, resourced threat operations.
What to do: If you operate in India or have Indian offices: deploy email security gateways to block spoofed tax authority messages; educate users that governments don't email credentials requests; enforce EDR or application whitelisting to block unknown executables; monitor for C2 beaconing. Even without these controls, defense-in-depth across email, user education, firewalls, and endpoint detection stops this chain at multiple points.
How are AI agents being tricked into making crypto payments?
What happened: Researchers at Zscaler uncovered two campaigns using indirect prompt injection to trick AI agents into making unauthorized payments or trusting fraudulent crypto platforms. Attackers hide prompts in hidden HTML divs and schema markup on SEO-poisoned or typosquatted domains targeting developers searching for legitimate libraries and DeFi tools.
Why it matters: This shows threat actors weaponizing prompt injection for financial theft, not just data exfil. The same malicious site targets both AI agents and human developers with payment options, indicating thorough adversary thinking. As organizations grant AI agents access to financial instruments and email, attack surface expands dramatically.
What to do: If you deploy AI agents: implement human-in-the-loop approval for any financial transactions; restrict AI agent access to sensitive systems and credentials; educate developers on what permissions and access levels they grant to AI tools; monitor for unexpected payments or API calls. Treat autonomous AI agents with the same caution you would a privileged service account.
Why is France banning non-quantum-safe encryption products?
What happened: France's cybersecurity agency (ANSSI) announced it will stop certifying security products lacking quantum-resistant encryption starting 2027, and will require government and critical infrastructure operators to use only quantum-safe products by 2030.
Why it matters: Quantum computers can break current encryption. This policy forces vendors to adopt quantum-safe algorithms now, not later. Without regulatory mandate, vendors have no financial incentive to re-engineer working products. This is a leading indicator—other governments will follow, and it will eventually pressure private sector adoption.
What to do: If you evaluate or procure security products, begin asking vendors about quantum-safe encryption roadmaps. Don't force the issue tomorrow, but start asking. Expect a compressed timeline once a major breach exploits cracked encryption; retrofitting millions of devices is expensive and slow.
What's the privilege escalation flaw in Linux and Android's epoll subsystem?
What happened: A CVE (BadEPoll) in the Linux kernel's epoll system allows local attackers with no privileges to gain full root access on affected Linux and Android devices. The flaw was missed by AI but discovered by a security researcher. Patches are available.
Why it matters: This is a local attack requiring physical or prior system access—threat actors operate remotely and don't live in your house. EPSS scoring places it in the lower 2% of vulnerability severity. No known exploitation in the wild yet.
What to do: Monitor for now. If you run Linux or Android, apply patches during normal cycles. This is not an emergency patch situation. The story's framing ("AI missed it, humans found it") is academic research clickbait; don't conflate AI's current limitations with real risk.
Why is the Adobe ColdFusion RCE being exploited right now?
What happened: Adobe released patches for a maximum-severity remote code execution vulnerability in ColdFusion 2025.9 and 2023.20 and earlier. The Canadian Center for Cyber Security reports active exploitation in the wild.
Why it matters: Max severity + active exploitation = urgent. ColdFusion is used to build enterprise web applications and may be internet-facing. Unknown patch adoption rate means unpatched instances likely exist in production.
What to do: If you run ColdFusion: inventory your instances immediately; prioritize patching to this week or weekend; communicate with application owners and track remediation through completion—don't just fire an email and move on. If you don't know whether you run ColdFusion, that's a sign your asset inventory is weak; fix that first.
How are North Korean operators compromising open-source software at scale?
What happened: A campaign active since December 2025 compromises legitimate GitHub, npm, Packagist, and Chrome extension repositories by hijacking maintainer accounts, injecting JavaScript loaders, and rewriting git history to obfuscate malicious commits. Payloads include DevPopper RAT and Omni Stealer targeting developers.
Why it matters: Developers trust open-source maintainers. This exploits that trust at the source. The threat actor tricks developers in fake job interviews into installing backdoored packages, turning them into their own attack vector.
What to do: Educate developers that job interview "technical assessments" from unknown companies requesting local downloads are high-risk. Frame this carefully to avoid signaling that job searching is discouraged. Implement software composition analysis (SCA) to detect known-bad open-source packages in your supply chain. Monitor for unusual package releases (version bumps with suspicious commits).
Why should you care about Opera's clickjacking defense?
What happened: Opera browser added paste-protection features (injection protection) to block ClickFix-style attacks where victims are tricked into pasting malicious PowerShell commands into command line interfaces.
Why it matters: ClickFix is a viable social engineering vector. Opera's mitigation is good; Chrome and Edge should follow. Browser-level protections reduce user friction compared to endpoint controls.
What to do: If you use Opera, the feature is enabled by default. If you use Chrome or Edge, monitor for similar features. Educate users: legitimate verification processes do not ask you to open Run and paste commands. If someone is asking you to do that, it's malicious.
Key takeaways
- Supply chain and open-source attacks remain the most persistent threat vector; asset inventory and software composition analysis are non-negotiable.
- Prompt injection attacks on AI agents represent a new financial attack surface—restrict what your AI systems can access and require human approval for sensitive operations.
- Patch management is not a send-email-and-forget game; you must track and follow up until remediation is confirmed, especially for max-severity flaws with active exploitation.
- Quantum-safe encryption is coming by regulatory mandate in 2027+; start asking vendors about roadmaps now rather than scrambling later.
- Academic research (airgap exfil via pixel emissions, etc.) is interesting but not actionable; don't let novel attacks distract from high-probability threats like phishing and unpatched critical flaws.
Topics covered
Want the live experience? The Daily Cyber Threat Brief airs live every weekday at 5am PT / 8am ET on YouTube. 400+ practitioners join the chat in real time.