What is Fake Finder — Is It Real or Fake?

Research context and background

Fake Finder is a software prototype created by IQT Labs to help users determine if a video has been altered by artificial intelligence. The tool was built using open-source deep learning models and includes a front-end application, an API for programmatic access, and a containerized back-end that runs on cloud infrastructure. Its main purpose is to let users upload a short video clip and receive a prediction on whether the video is real or fake. However, recent audits have highlighted that the tool is still in a prototype stage and has specific limitations regarding its scope, potential bias, and security readiness for enterprise use.

Benefits

The primary benefit of Fake Finder is its ability to use advanced deep learning to analyze video content. It is designed to detect manipulations that are not easily visible to the human eye. The tool offers a user-friendly interface built with React, making it accessible for non-technical users to upload and test videos. It also provides an API, which allows developers to integrate the detection capabilities into their own applications or workflows. The open-source nature of the project means that the code is available for others to study, learn from, and potentially improve.

Use Cases

Fake Finder is intended for scenarios where verifying video authenticity is important. It can be used by individuals or organizations who need to check if a video has been tampered with. The tool is particularly focused on detecting face swap techniques, which are a common method used to create deepfakes. Users might employ it to verify the identity of people in videos or to ensure that media content has not been altered maliciously. It is important to note that the tool is currently a prototype and is best suited for testing and educational purposes rather than critical security decisions in a production environment.

Pricing

Pricing details are not available for Fake Finder. The tool is described as an open-source project, which often means it is free to use, but specific pricing models or enterprise licensing costs have not been disclosed in the available information.

Vibes

Public reception and reviews are limited because the tool is a prototype and has not been widely deployed. An internal audit conducted by IQT Labs between July and September 2021 provided significant insight into its current state. The audit found that while the tool is innovative, it has significant limitations. Experts noted that the tool is effectively a face swap detector rather than a general deepfake detector. There are also concerns about potential bias related to race and gender due to the training data. Furthermore, the audit identified critical security vulnerabilities, including a bug that could allow remote code execution and issues with infrastructure security. These findings suggest that while the concept is promising, the tool requires substantial improvements before it can be trusted for serious real-world applications.

Additional Information

Fake Finder was developed by colleagues at IQT Labs as a software prototype. It is composed of six different deepfake detector models that were developed externally and made open-source. The project includes a front-end application built with React, an API for programmatic access, and a containerized back-end that facilitates running compute resources on AWS. The audit highlighted that the tool is not yet production-ready for enterprise use due to its narrow scope, potential biases, and security flaws. The developers recommend that users clearly understand the tool's limitations, specifically that it only detects face swaps, and that security vulnerabilities must be patched before the tool is used in any environment beyond a controlled prototype.