At AWS re:Invent 2023, Dr. Rebecca Portnoff, Head of Data Science at Thorn, shares how their nonprofit organization utilizes Amazon Web Services (AWS) to combat child sexual abuse. Thorn, a nonprofit that builds technology to defend children from sexual abuse, leverages AWS's robust capabilities to enhance their machine learning tools.
According to the National Center for Missing and Exploited Children (NCMEC), 88 million files of suspected child abuse were reported by online platforms in 2022 alone.
With just one second spent on review of each file, this would take a single analyst nearly three years of non-stop review to complete.
Thorn’s Safer tool has led to the identification of more than 2.8 million potential child sexual abuse material (CSAM) files. It uses the scalability of AWS technology to build critical testing components for its CSAM Classifier, using Amazon Elastic Cloud Compute (Amazon EC2) and Amazon Elastic Kubernetes Service (Amazon EKS) to empower end users with fast remote access to its on premise solution for model training, debugging and fixing any issues with its training pipeline prior to training with on-premises data.
Dr. Portnoff emphasized the proactive role of content-hosting platforms in combating this issue by utilizing Thorn's Safer tool to protect their platforms from CSAM. Safer is designed to help any platform with an upload button to identify, review, and report CSAM, ensuring comprehensive and proactive detection. Dr. Portnoff also explains how the organization’s CSAM Classifier helps platforms identify new and previously unknown CSAM, often signifying active abuse situations.
Thorn also uses Amazon Simple Storage Service (S3) to store benign data that is crucial for training its CSAM Classifier, and powers user feedback within its cloud-hosted Safer services where users can submit false positives to be incorporated back into training. Thorn also uses Amazon Elastic Container Registry (Amazon ECR) to distribute its trained models to end users.
Thorn is a nonprofit that builds technology to combat child sexual abuse at scale. AWS is our preferred cloud provider to match the scale of our ambition and power our machine learning tools.”
Dr. Rebecca Portnoff
Head of Data Science, Thorn
About Thorn
Founded in 2012, Thorn builds technology to defend children from sexual abuse. The organization collaborates with law enforcement agencies, tech companies, and nonprofits. Using a variety of Amazon Web Services (AWS) solutions, Thorn, an AWS Partner, developed Safer, an application that works within customer storage environments to detect CSAM. The application then elevates suspected CSAM to the company for review and helps report confirmed CSAM—which is uniquely positioned to engage law enforcement to rescue victims.
By using a full stack of services from AWS, Thorn built an accessible tool that any company can use to identify, remove, and report CSAM from content-hosting sites without a large investment in staff headcount, unnecessary exposure of employees to disturbing material, or unforeseen legal risk.
Customer Speaker: Dr. Rebecca Portnoff
Dr. Rebecca Portnoff has dedicated her career to building tools and techniques to seek out and help child victims of sexual abuse. She is currently Head of Data Science at Thorn, where she owns strategy and vision for Data Science across the organization. She works cross functionally with business and technical functions to develop, deploy and maintain machine learning/artificial intelligence (ML/AI) and algorithms to: accelerate victim identification, stop re-victimization (via the removal of child sexual abuse material from the open web), and prevent abuse from occurring. She leads Thorn’s engagement with ML/AI as a field, including emerging threats and emerging technologies. She holds a B.S.E. in Computer Science from Princeton University, where she also minored in vocal jazz, and a Ph.D. in Computer Science from UC Berkeley.
/SMOKEBALL_logo%401x.febfb961d8faa07b6a06937fe1a232690321b7b4.png)
/IOTA_MEDTECH_logo%401x.0c664af34b06ab64c1ac214c5f12077884223ab3.jpg)
More to Explore
# Screens < wide
# screens = wide
