AI-Driven Detection and Analysis of Handwriting on Seized Ivory: A Tool to Uncover Criminal Networks in the Illicit Wildlife Trade

Will Fein, Ryan J. Horwitz, John E. Brown III, Amit Misra, Felipe Oviedo, Kevin White, Juan M. Lavista Ferres, Samuel K. Wasser · 2025

Read AI-Driven Detection and Analysis of Handwriting on Seized Ivory: A Tool to Uncover Criminal Networks in the Illicit Wildlife Trade by Will Fein, Ryan J. Horwitz, John E. Brown III, Amit Misra, Felipe Oviedo, Kevin White, Juan M. Lavista Ferres, Samuel K. Wasser (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-10219.txt.

Introduction

Will Fein and co-authors present a 2025 AI tool that reads handwriting on seized ivory to map trafficking rings. Elephant tusks carry marks scrawled by the traffickers who moved them. DNA tests on tusks are conclusive but costly and sometimes impossible. Handwritten marks are easy to photograph yet rarely studied.

Their pipeline finds and extracts the marks automatically. It then links marks across seizures to expose shared hands and routes. For twenty years DNA has tied shipments to poaching grounds. Handwriting could add a cheaper, faster tie layer. The work aids law officers fighting the ivory trade.

Catalog record

CREATORWill Fein, Ryan J. Horwitz, John E. Brown III, Amit Misra, Felipe Oviedo, Kevin White, Juan M. Lavista Ferres, Samuel K. Wasser
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-10219

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

AI-Driven Detection and Analysis of Handwriting on Seized Ivory: A Tool to Uncover Criminal Networks in the Illicit Wildlife Trade Will Fein, Ryan J. Horwitz, John E. Brown III, Amit Misra, Felipe Oviedo, Kevin White, Juan M. Lavista Ferres, Samuel K. Wasser Abstract The transnational ivory trade continues to drive the decline of elephant populations across Africa, and trafficking networks remain difficult to disrupt. Tusks seized by law enforcement officials carry forensic information on the traffickers responsible for their export, including DNA evidence and handwritten markings made by traffickers. For 20 years, analyses of tusk DNA have identified where elephants were poached and established connections among shipments of ivory. While the links established using genetic evidence are extremely conclusive, genetic data is expensive and sometimes impossible to obtain. But though handwritten markings are easy to photograph, they are rarely documented or analyzed. Here, we present an AI-driven pipeline for extracting and analyzing handwritten markings on seized elephant tusks, offering a novel, scalable, and low-cost source of forensic evidence. Having collected 6,085 photographs from eight large seizures of ivory over a 6-year period (2014-2019), we used an object detection model to extract over 17,000 individual markings, which were then labeled and described using state-of-the-art AI tools. We identified 184 recurring “signature markings” that connect the tusks on which they appear. 20 signature markings were observed in multiple seizures, establishing forensic links between these seizures through traffickers involved in both shipments. This work complements other investigative techniques by filling in gaps where other data sources are unavailable. The study demonstrates the transformative potential of AI in wildlife forensics and highlights practical steps for integrating handwriting analysis into efforts to disrupt organized wildlife crime. Introduction The illegal ivory trade is a multibillion-dollar industry that threatens to drive African elephants toward extinction [ 1 ] . Each year, thousands of African elephants are killed for their tusks, and transnational criminal networks traffic shipments of ivory across borders [ 2 ] . As part of the fight against the illicit ivory trade, we developed a method to extract forensic evidence from photographs of tusks in large seizures of ivory using AI and computer vision tools. Elephant ivory is poached in areas expansive enough to support large numbers of elephants and provide ample cover for poachers [ 3 ] . Middlemen acquire the ivory from poachers and move it across borders in small quantities, consolidating the ivory into large shipments before exporting it [ 4 , 5 ] . By weight, most illicit exported ivory is packed in shipments of 100kg or more [ 6 ] . DNA analyses of large ivory seizures have been very useful in determining where the ivory was poached [ 7 , 8 , 9 , 3 ] and have also shed light on the number and connectivity of the major transnational criminal organizations consolidating and exporting the ivory [ 4 , 5 ] . As DNA evidence is expensive to obtain, these analyses have been limited to very large seizures and rely on data taken from a representative sample of tusks. In addition to the DNA evidence they contain, many tusks carry handwritten markings made by traffickers. For over a decade, the United Nations Office on Drugs and Crime has recommended documenting these markings, noting that they carry information about the individuals and organizations involved in the trafficking of a tusk [ 10 ] . Between 2014 and 2019, the University of Washington Center for Environmental Forensic Science collected photographs of tusks in large ivory seizures, resulting in an extensive database of 6,085 photographs taken from 8 shipments each containing at least 1,800kg of unworked ivory. Though photography itself is inexpensive and the value of tusk markings is well-documented, no scalable approach for analyzing them has yet been formulated, and this evidence remains underutilized. With automated techniques, these data can be unlocked. In this work, we fine-tuned an AI vision-language model to detect handwritten markings in photographs of seized tusks and implemented a semi-automated pipeline that applies meaningful labels to those markings while minimizing the manual work required. By identifying ”signature markings,” we revealed new forensic connections among ivory shipments. This work introduces a scalable and cost-effective method for extracting forensic insights from tusk markings, enabling law enforcement to uncover new links between ivory seizures and trafficking networks. Results Using our semi-automated pipeline, we extracted and labeled 17,071 handwritten markings from 6,085 images to create a curated dataset for analysis (See 1 Materials and Methods for full pipeline details). Handwritten markings hold evidentiary value because they enable us to trace the involvement of criminal entities across multiple tusks. Thus, we defined the concept of “signature markings,” or markings with sufficient consistency and specificity to connect the tusks on which they appear. We established seven cross-seizure connections using handwriting data (Table 1 ). We also examined the frequency of the most common signature markings, providing indications as to the number of middlemen involved in each seizure and highlighting the types of marking that merit greater attention in the future. Table 1: Shared Signature Markings Identified Across Seizure Groups. Seizures Handwriting Evidence 2+3 Lambda X/O variation 2+8 7 shared signature markings 3+8 1 shared signature marking 5+8 8 shared signature markings 7+8 1 shared signature marking 2+5+8 1 shared signature marking 2+7+8 1 shared signature marking Figure 1: Signature Markings Identified in Multiple Seizures. Examples of 4 signature markings identified in multiple seizures. Seizures in which each marking was found: a. 8; b. 2; c. 5; d. 8; e. 8; f. 5; g. 2; h. 5. Connecting Seizures Seven cross-seizure connections were established using handwriting data (Table 1 ). No transitive connections are assumed, and three-way connections are included as separate combinations of seizures from their component pairwise combinations. Six relied on the co-occurrence of signature markings, meaning that we observed the same signature markings in multiple seizures. The seventh is based on observation of the ”Lambda X/O variation” which is discussed in more detail below. Four of the connections involve seizure 2, a seizure for which genetic data was unavailable. While DNA evidence can provide the most conclusive connections among seizures, handwriting analysis can reveal vital links when that data is unavailable. Four of the connections involve seizure 2, a seizure for which genetic data was unavailable. While DNA evidence can provide the most conclusive connections among seizures, handwriting analysis can reveal vital links when that data is unavailable. Evidence From Matched Signature Markings Our dataset included 184 signature markings that appeared more than once, including 20 that appeared in multiple seizures. Table 2 provides the number of observed instances of each signature marking and Figure 1 shows select examples. The matched signature markings included 16 pairs of initials, 1 longer text, and 3 non-text symbols. Markings not shown in figures are named generically in this table (e.g. “Initial Pair A”) to minimize the amount of evidence exposed to possible adversarial actors. Notably, when we refer to the number of instances or occurrences of a given signature marking, we refer only to the number of examples in the corpus of photographs. Since some tusks were photographed multiple times, this value is not necessarily equivalent to the true number of unique appearances

Read the complete text