top of page
ellinamanagement

Company Spotlight: Reality Defender

By Elliot Eizik - Ellina Management




Who is Reality Defender, and what do they do?

Reality Defender is a pioneering deepfake detection platform founded by a team of experts with strong academic backgrounds and extensive experience in data science, cybersecurity, and artificial intelligence (AI). Their mission involves assisting corporations, governments, and journalistic entities to distinguish between genuine and fabricated media content, emphasizing audio, video, and image manipulation. Through their Application Programming Interface (API) and web application, Reality Defender delivers real-time scanning capabilities, risk assessment, and comprehensive PDF reports to equip clients with the tools needed to counteract the rising threat of deepfakes.


What are deepfakes?

Deepfakes, which refer to AI-generated media content that convincingly portrays individuals saying or doing things they never did, have become increasingly sophisticated and accessible. This trend poses a significant challenge, as even seasoned observers may need help identifying certain deepfakes. The implications are far-reaching, affecting financial transactions, personal and corporate reputations, public perception, and national security.


Who are the founders of Reality Defender?

The founders of Reality Defender, Benjamin Colman, Gaurav Bharaj, and Ali Shahriyari, possess advanced degrees from esteemed institutions such as Harvard, NYU, and UCLA, along with substantial experience at prestigious organizations like Goldman Sachs, Google, the CIA, the FDIC, the Department of Defense, and Harvard University, specializing in the intersection of machine learning and cybersecurity. Interestingly, the inception of Reality Defender stemmed from an unconventional project involving Deepak Chopra. While collaborating with Chopra on creating a lifelike deepfake for real-time conversations, the founders realized the alarming realism achievable with deepfake technology and the dire need for reliable detection methods.


A model of models approach

To address this gap, Reality Defender adopts a unique approach. Unlike companies attempting a one-size-fits-all "silver bullet" solution, they advocate for a "multi-model" strategy, where various deep-learning detection models focus on specific features, and their scores integrate into an aggregate "model of models." This approach acknowledges the evolving nature of adversaries and technologies in the deepfake landscape.


Their commitment to collaboration and innovation has attracted contributions from partners such as Microsoft, UC Berkeley, and Harvard, leading to the development of a scalable platform. Early users have sought Reality Defender's assistance in diverse deepfake scenarios, from disinformation campaigns targeting international relations to forged audio for financial fraud and manipulated video of public figures engaged in scandalous behavior. The platform has consistently identified these fabrications with high accuracy.


One central challenge that deepfakes pose is their ability to adapt and circumvent existing security measures, requiring detection techniques to evolve in tandem. Reality Defender adopts this perspective by integrating external deepfake detection models and collaborating with a network of researchers. This iterative approach ensures that the platform remains ahead of emerging threats. In addition to audio and video deepfake detection, Reality Defender has expanded its capabilities to include AI-generated text detection. This diversification allows them to address a broader range of AI-generated content threats.


The company is also exploring subscription-based pricing plans, allowing clients to scan at least 250 media assets monthly. This approach reflects their dedication to affordability and accessibility while accommodating high computing costs associated with running multiple real-time models.


As with computer viruses, deepfake technology will persistently evolve to evade current security measures, necessitating a continual arms race. Reality Defender positions itself as a dynamic, adaptable defender against deepfake threats. By embracing collaboration, innovation, and proactive detection, they strive to protect clients from misinformation and deception, all while advancing the field of deepfake detection.


Arguments in favor of Investing in Reality Defender:


Cutting-Edge Technology: Reality Defender leads the battle against deepfake threats through a dynamic "model of models" approach and active collaborations, ensuring they remain at the forefront of detection technology.

Collaborative Ecosystem: The platform's network of contributors and partners fosters innovation and provides access to diverse expertise, enhancing detection quality.

Proactive Development: Reality Defender identifies existing deepfake threats and anticipates and counters future challenges, offering clients a comprehensive defense strategy.

Comprehensive Scanning: The platform's holistic approach covers audio, video, and image deepfake detection, offering a more comprehensive solution for clients.

Actionable Insights: Reality Defender provides clients with probabilistic percentage scores, enabling informed content moderation decisions rather than binary yes/no results.

Affordability: Their subscription-based pricing model aims to make deepfake detection accessible, balancing affordability and computing costs.


Arguments against Investing in Reality Defender:


Limited Consumer Access: Reality Defender primarily serves large enterprises, government agencies, and organizations, potentially limiting access for individual consumers who also require protection from deepfake threats.

Subscription-Based Model: While affordability is a goal, the subscription-based pricing model may still pose financial challenges for clients who require continuous access to the service.

Regulatory Uncertainty: The absence of comprehensive regulatory frameworks for deepfake detection may create uncertainties in the market's long-term viability.

Dependency on Collaboration: Reality Defender's success relies on collaborators' willingness to share detection methods and models, which may only sometimes align with their interests or priorities.

Competitive Landscape: The deepfake detection market is highly competitive, with established players and emerging technologies challenging the company's competitive edge. The following is a list of competitors already in this space.

Truepic, c2pa.org, ZeroGPT, Hugging Face Image Detector, Deep Media, DeepWar, USENIX, FakeCatcher, Microsoft's Video Authenticator, Hive Moderation, Mayachitra, Copyleaks, Counter.social, Deeptrace, Sensity.ai.

Just like there is no single software virus protection company enabling unbreakable defense against future viruses, expecting one company with a more substantial moat of resistance against competitors and other established entities in this space will be a difficult race to the bottom of enterprise consumer value.

A company needs to differentiate itself by being as close as possible to be in a market of one. Designing deepfake technology is not that market, and Reality Defender is not that company.


Sources -








7 views0 comments

Recent Posts

See All

Comentários


bottom of page