Detect and Trace: An Australian Field Trial Using Machine-Learning Tools to Combat Illegal Wildlife Trade

Area of Science:

• Wildlife conservation
• Forensic science
• Artificial intelligence

Background:

• Illegal wildlife trade poses a significant global threat to biodiversity, ecosystems, and human health.
• Limited real-world testing of advanced technologies hinders efforts to combat wildlife trafficking.
• Existing methods for detecting smuggled wildlife are often insufficient against sophisticated trafficking operations.

Purpose of the Study:

• To evaluate the effectiveness of two machine-learning tools in detecting smuggled wildlife during real-world seizures.
• To assess the safety and utility of radiation-exposure data from CT X-ray scanning in wildlife interdiction.
• To demonstrate the potential of integrating novel technologies into intelligence-led enforcement strategies.

Main Methods:

• An opportunistic Australian trial involving seven months of real-world wildlife seizures.
• Deployment of two machine-learning algorithms with a CT X-ray baggage scanner (RTT®110) and a portable X-ray fluorescence (pXRF) device (Olympus Vanta).
• Analysis of 116 intercepted animals (reptiles and crustaceans) across five genera and 48 scanned parcels.

Main Results:

• Automated AI detected smuggled wildlife in 56% of scanned consignments using the AT.3 algorithm.
• High-resolution 3D X-ray images enabled the identification of concealed wildlife.
• Machine-learning provenance models differentiated between wild-caught and captive-bred lizards (Tiliqua sp.), aiding enforcement.

Conclusions:

• Integrating machine-learning technology significantly enhances the detection capabilities for smuggled wildlife.
• The trial demonstrated a reduction in the export of seized parcels via postal pathways following technology implementation.
• These findings support the adoption of advanced technological solutions for more effective wildlife trafficking interdiction.