Threat Brief — 2026-08-15 — Fake Flash, Real Stealth

Executive summary: Two intelligence-relevant items stand out from today's feed. A Chinese threat group is combining government espionage with cryptocurrency theft using a revived fake-Flash-updater social engineering vector — a reminder that legacy delivery mechanisms remain effective. Separately, Chinese researchers demonstrated a low-cost AI model achieving 97% accuracy in classifying F-22 and F-35 aircraft by thermal signature, lowering the barrier to counter-stealth detection.

Top items

• Chinese APT blends state espionage with crypto theft via fake Flash updates. A newly spotlighted Chinese hacking group combines traditional government-targeted spying with cryptocurrency theft, using a long-known social engineering trick: weaponized "Flash Player update" prompts that deliver malware. The continued effectiveness of this vector underscores that legacy delivery mechanisms — once assumed obsolete after Flash's EOL — still succeed against unpatched or unsavvy targets. (src: RSS:securitylab-ru)

• Chinese AI achieves 97% F-22/F-35 thermal-signature recognition on commodity hardware. Researchers trained an AI model to identify fifth-generation stealth fighters by their thermal signatures, reaching 97% accuracy using inexpensive hardware costing only a few hundred yuan. This lowers the cost barrier for counter-stealth detection and could inform adversary air-defense targeting workflows. (src: RSS:securitylab-ru)

Themes

• Legacy vectors, modern payloads: The fake-Flash-updater trick has persisted well past Flash's 2020 EOL, demonstrating that threat actors continue to exploit user habits and outdated mental models long after underlying technologies are retired. Paired with dual-mission operators (espionage + financially motivated crypto theft), this blurs the line between nation-state and cybercrime activity.
• AI as offensive enabler: The thermal-signature classifier joins a growing pattern of AI being weaponized for defense-relevant targeting at low cost, complementing other recent AI-in-intelligence developments covered this week.