現代のトロイの木馬~ウクライナの「クモの巣」作戦とドローン運用の技術的側面:NIDSコメンタリー第385号

This article provides a detailed analysis from technical perspectives of Ukraine's "Operation Spiderweb" conducted by the Security Service of Ukraine (SBU) on June 1, 2025, suggesting that modern drone tactics combining open-source technology and AI could overturn traditional defense concepts.

Key Points

Scale and Results of the Operation

  • 41 aircraft were destroyed including Russian Air Force strategic bombers Tu-95MS, Tu-22M3, and early warning and control aircraft A-50, with total damage amounting to $7 billion
  • Inflicted damage equivalent to 34 percent of Russian Air Force strategic bombing capability
  • Simultaneously attacked 4 Russian military bases including Olenya base in Murmansk Oblast 1,800 kilometers from Ukraine and Belaya base in Irkutsk Oblast 4,500 kilometers away
  • 117 FPV drones were deployed with one operator assigned to each aircraft
  • Operation was meticulously prepared over 18 months, with over 100 small drones smuggled into a secret facility in Chelyabinsk, Russia

Technical Innovations

  • Utilized Russia's domestic 4G/LTE mobile communication networks, remotely operated from within Ukraine using pre-installed SIM cards
  • Used open-source autopilot software ArduPilot to achieve stable flight through dead reckoning while avoiding GPS jamming
  • Made detection difficult by embedding control signals in civilian data traffic, receiving control signals via UART interface
  • Set multiple waypoints in cruise control mode, maintaining autopilot following pre-set flight routes
  • Failsafe function during communication loss locked drone altitude, speed, and direction, maintaining flight path for a certain period

Features of AI-Equipped OSA Drone

  • Used OSA drones developed by Ukraine's First Contact company, with maximum payload of 3.3 kilograms, maximum flight time of 15 minutes, speed of 150 km/h
  • Robust closed structure with electronics and wiring housed inside the fuselage, adapted to severe weather and poor road conditions
  • Equipped with Raspberry Pi small computer (approximately 85mm × 56mm, weighing less than 50 grams) to enable AI processing
  • Shaped charges containing incendiary materials embedded in landing skids for precision strikes on aircraft vulnerable parts
  • Completed development of AI-embedded unmanned aircraft system by January 2024 and began final testing

AI Terminal Guidance System

  • Switches to autonomous flight mode using AI along pre-planned route when signal is lost, warhead automatically detonates upon approaching/contacting target
  • AI development through machine learning using hundreds of Soviet-era bomber images stored at Poltava military museum
  • Marked fuel tank locations of Tu-95MS and Tu-22M3 with red X marks, identifying vulnerable parts as targets
  • Possibly equipped with Skynode-S terminal guidance solution developed by American company Auterion
  • Repeated epoch (learning iterations) work, verifying target recognition under various angles, lighting, and weather conditions

Detailed Analysis of Operation Execution

  • In attack on Belaya Air Force Base in Irkutsk Oblast, flew approximately 6 kilometers from trailer to base in 6 minutes (approximately 60 km/h)
  • Strong winds suppressed speed and battery consumption limited attack time after base arrival to just 9 minutes
  • Due to mobile network bandwidth limitations, drones launched sequentially rather than simultaneously
  • With 2-3 second signal delays, weather conditions, and jamming risks, AI autonomous guidance functioned as essential backup for mission success
  • Series of remotely opening wooden container roof, launching drones, and self-destructing trailer all controlled via 4G/LTE

The article warns that the high cost-effectiveness of inflicting billions of dollars in damage with assets worth only thousands of dollars, combined with the ability to access previously out-of-range areas by utilizing globally deployed ICT infrastructure, represents a new reality that could become an attractive option not only for states but also for non-state actors.

※ This summary was automatically generated by AI. Please refer to the original article for accuracy.