evidoria

← Back to browse

Good practice Imported

Detect and Trace — DCCEEW's AI X-Ray Pilot for Catching Wildlife Smugglers at Sydney's International Mail Gateway

Australia · Sydney · See the Australia profile · See the Sydney profile

Evidence: Observational / pre–post Top 83% 40/100 · Ask Evidence Copilot about this practice

DCCEEW-backed researchers trialled a 3D X-ray CT scanner and machine-learning classifier on real seized mail at Sydney's international gateway; across 18 consignments (48 parcels, 116 reptiles and crustaceans) the best algorithm flagged concealed wildlife 56% of the time.

82 %
Lab validation detection rate (curated images) (2022)
1.6 %
Lab validation false-alarm rate (2022)
294 images across 13 species
Lab validation image set (2022)
56 %
Field trial detection rate (real seized mail) (Jul 2023 - Jan 2024)
18 consignments (48 parcels)
Field trial consignments scanned (Jul 2023 - Jan 2024)
116 animals across 5 genera (87% reptiles)
Animals detected in field trial (Jul 2023 - Jan 2024)
200+ parcels (780 native animals)
Wildlife parcels intercepted (wider crackdown) (Jun 2023 - early 2025)
545 % increase
Rise in wildlife interceptions (Jun 2023 - early 2025)
8 years
Sentence given to wildlife trafficker (Feb 2026)
Detect and Trace — DCCEEW's AI X-Ray Pilot for Catching Wildlife Smugglers at Sydney's International Mail Gateway

Details

Maturity
Pilot
Promoter
Department of Climate Change, Energy, the Environment and Water (DCCEEW) Wildlife Crime Special Investigators
Period
2022 lab validation; Jul 2023 – Jan 2024 field trial; published 2026
Keywords
wildlife conservation, customs & border security, law enforcement

Context

Researchers at Macquarie University and the Taronga Conservation Society, backed by Australia's DCCEEW and Department of Agriculture, Fisheries and Forestry, developed 3D X-ray CT scanning paired with machine-learning classifiers to detect wildlife concealed in international mail and luggage.

Objectives

The work aimed to give DCCEEW's Wildlife Crime Special Investigators a screening tool that could flag concealed live animals in mail parcels at Sydney's international gateway, building on a controlled 2022 validation study.

Activities

A first controlled validation, published in 2022, scanned 294 curated images across 13 species; DCCEEW investigators then trialled a Rapiscan RTT®110 scanner at the Sydney Gateway Facility between July 2023 and January 2024, scanning parcels that Australian Border Force staff had already flagged as suspicious, with no negative controls.

Results

The 2022 lab validation reached an 82% detection rate with a 1.6% false-alarm rate; in the real-world field trial across 18 seized consignments (48 parcels, 116 animals across five genera, 87% reptiles), the best-performing algorithm correctly flagged concealed wildlife in 56% of scans — a sharp drop from the lab figure. The wider DCCEEW crackdown intercepted more than 200 parcels containing 780 native animals between June 2023 and early 2025, a 545% rise in interceptions, and in February 2026 a linked investigation led to an 8-year sentence for a wildlife trafficker, the longest ever given to an Australian wildlife smuggler.

Conclusions

Researchers recommend a cloud-based mobile app to speed field use and formal investigator training on the companion pXRF provenance device, whose scan numbers were limited by staff availability, and note the approach depends on 3D CT scanners that are "not available everywhere."

Implementation

Indicative cost
Medium (€50k–€500k)
Time to results
Medium (1–3 years)
Staffing & skills
DCCEEW Wildlife Crime Special Investigators operated the scanner during the field trial, Researchers recommend formal training for investigators on the companion pXRF provenance device

Conditions for success

  • Access to a 3D X-ray CT scanner (Rapiscan RTT®110), which researchers note is 'not available everywhere'
  • Prior flagging of suspicious parcels by Australian Border Force staff to narrow the scanning workload
  • Sufficient trained-staff availability, which the study found limited pXRF scan numbers

Common failure modes

  • Real-world detection accuracy (56%) fell sharply from the curated lab validation (82%)
  • The field trial had no negative controls, since only already-flagged parcels were scanned
  • The approach has not yet been documented as replicated at another port or country

Where it fits

Governance type
national government agency (DCCEEW) with university/NGO research partners
Scale
single facility pilot (Sydney Gateway Facility)
Income level
high-income

Commonly funded by

National / regional programmes Philanthropic / foundation funding

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful