When a deepfake targeted him personally, Vaishnav Anand panicked. But when everything settled down, he turned that panic into purpose.
The California high-school junior was inspired by the incident to ask a different question than most: If people already doubt celebrity videos and viral images, what about the satellite maps that governments and corporations quietly trust every day? If these could be altered to create distortions, such as faking natural disasters or hiding weak infrastructure, that could have serious effects, he pondered.
"I began because I was a victim of a deepfake, and it kind of made me realize how easily something true can be absolutely manipulated with AI and look so realistic," he says.
In the months that followed, Anand turned that shock into a research project he recently presented at the IEEE Undergraduate Research Technology Conference at the Massachusetts Institute of Technology. The project focuses on how to detect altered satellite imagery before it can distort public decisions.
His work fills a gap in a field with surprisingly little research. Only a handful of studies have explored what some scientists now call "deepfake geography." One 2021 paper showed how artificial intelligence can blend features from one city into the satellite imagery of another to create convincing but false landscapes.
Related:Data Dump From APT Actor Yields Clues to Attacker Capabilities
Why Geospatial Deepfakes Matte r
Anand first explored voice and face forensics before widening his focus to geospatial data, recognizing how it underpins everything from disaster response to national security planning. Through his work on the National 4-H Geospatial Team, he learned to "take every point as data" and to interpret maps and satellite imagery with precision.
Anand says he was surprised by how little research exists on detecting manipulated satellite imagery — despite the fact that the potential consequences could be "catastrophic," he says.
"Satellite imagery is really a national security issue," he says, adding that if adversaries can alter even a "little bit of that data," the downstream impact on infrastructure and government decisions could be severe.
... continue reading