Araceli: Artificial Intelligence in the Fight Against Child Sexual Abuse

• Grant number: 23/12086-9
• Support Opportunities: Regular Research Grants
• Start date: December 01, 2023
• End date: November 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Sandra Eliza Fontes de Avila
• Grantee: Sandra Eliza Fontes de Avila
• Host Institution: Instituto de Computação (IC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Associated researchers: Fabrício Benevenuto de Souza ; Jefersson A dos Santos ; Leo Sampaio Ferraz Ribeiro
• Associated scholarship(s): 24/16685-7 - Transferability Estimation Scorers for Efficient Transfer Learning, BP.DR; 24/09375-1 - Zero-Shot Age Estimation for Child Sexual Abuse Material, BP.IC; 24/09372-2 - Scene Graph Generation as Proxy Task for Child Sexual Abuse Material Detection, BP.IC; 24/01210-3 - Child Sexual Abuse Material Recognition Models Based on Proxy Tasks, BP.TT

Abstract

Child sexual abuse materials (CSAM) have reached alarming proportions in the digital age. According to the National Center for Missing and Exploited Children's CyberTipline, over 32 million reports of suspected CSAM were received in 2022, making it a record year. The availability and sharing of such harmful content online not only exacerbate the trauma inflicted upon the victims but also significantly burden law-enforcement agents who manually inspect thousands of files, leading to emotional strain. In light of that, there is a need for reliable automated tools that can handle this kind of material securely and efficiently. In the Araceli* project, we aim to design, develop, and deploy solutions based on machine learning to detect CSAM, supporting forensic analysis automatically. Specifically, we will investigate how scene recognition and representation can be explored for CSAM classification while respecting the limitations of working with CSAM: low-data regimes, restricted testing, and less powerful hardware available for law-enforcement agents. Additionally, data-centric AI techniques will be leveraged to produce a high-quality and well-documented benchmark for evaluating CSAM classification approaches, available to the research community through model submission. Due to legal and ethical barriers, such sensitive data can only be accessed by police agents. For this reason, we will cooperate with Brazil's Federal Police and Technical-Scientific Police experts to evaluate the models on real child sexual abuse material. We highlight that our research group is at the forefront of such research worldwide, being responsible for groundbreaking results associated with digital forensics, machine learning, and computer vision.*In memory of the 8-year-old girl Araceli who was kidnapped, raped, and murdered on May 18, 1973. This crime still goes unpunished to this day. The National Day to combat abuse and sexual exploitation of children and adolescents was officially established through Law 9.970/2.000. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

PACCOTACYA-YANQUE, ROSA Y. G.; BISSOTO, ALCEU; AVILA, SANDRA. Are Explanations Helpful? A Comparative Analysis of Explainability Methods in Skin Lesion Classifiers. 2024 20TH INTERNATIONAL SYMPOSIUM ON MEDICAL INFORMATION PROCESSING AND ANALYSIS, SIPAIM 2024, v. N/A, p. 5-pg., 2024-01-01. (23/12086-9, 20/09838-0, 13/08293-7)

CAETANO, CARLOS; FERRAZ RIBEIRO, LEO SAMPAIO; LARANJEIRA, CAMILA; DOS SANTOS, GABRIEL OLIVEIRA; BARROS, ARTUR; PETRUCCI, CAIO; DOS SANTOS, ANDREZA APARECIDA; MACEDO, JOAO; CARVALHO, GIL; BENEVENUTO, FABRICIO; et al. Mastering Scene Understanding: Scene Graphs to the Rescue. 2024 37TH SIBGRAPI CONFERENCE ON GRAPHICS, PATTERNS AND IMAGES, SIBGRAPI 2024, v. N/A, p. 6-pg., 2024-01-01. (24/01210-3, 24/09372-2, 20/09838-0, 22/14690-8, 24/09375-1, 13/08293-7, 23/12086-9)

BENATTI, RAYSA; SEVERI, FABIANA; AVILA, SANDRA; COLOMBINI, ESTHER LUNA. Gender Bias Detection in Court Decisions: A Brazilian Case Study. PROCEEDINGS OF THE 2024 ACM CONFERENCE ON FAIRNESS, ACCOUNTABILITY, AND TRANSPARENCY, ACM FACCT 2024, v. N/A, p. 18-pg., 2024-01-01. (13/08293-7, 23/12086-9, 20/09838-0)