20. December 2022
AI and machine learning are shaping a smarter kind of future. We see their impact in ground-breaking innovations as varied as autonomous vehicles and digital therapeutics, law enforcement, and scientific research. AI also plays an equally vital role in business success, enabling ever-smarter ways to manufacture and connect. As these technologies require substantial amounts of video data, more and more are concerned about how to protect privacy in AI and machine learning project.
Innovation accelerates growth
In fact, the results of pursuing innovation on bottom line growth speak for themselves. For example, the Booz & Co. (now Strategy&) 2011 Global Innovation 1000 report showed that innovative organizations achieve 22% higher EBITDA growth than their less creative counterparts.
A private matter
Yet training algorithms to perform their tasks may often involve vast amounts of high-quality image and video data. And if you do run AI and machine learning projects, you will also undoubtedly be aware of the need to comply with stringent privacy regulations, including CCPA in the US, PIPL in China, and the EU’s General Data Protection Regulation (GDPR).
A fine state of affairs
You will also know that the consequences of breaching these standards can be extremely severe. By October 2022 alone, enforcementtracker.com estimates that fines for violating privacy standards will amount to 555 million euros.
A delicate dilemma
Since the collection, use, and storage of data requires the express written consent of subjects, it is clearly unrealistic to expect you to track down data subjects to gain this consent. So, innovation-driven companies face a dilemma: how to continue running AI and machine learning projects that require video data while remaining compliant with privacy regulations.
Privacy threatens to stifle innovation
According to Bitkom, Germany’s digital association, over 75% of the 502 companies surveyed agreed that innovation projects have failed due to the legal obligations imposed by the GDPR. And 86% have halted projects due to uncertainties in dealing with the regulation.
Quality and quantity
The conventional approach is to turn to anonymization technologies such as blurring, pixelation and black bars. Yet this is far from perfect. These methods are inherently unable to preserve the accuracy and integrity of the original data. Yet in most cases, you need to retain that original quality to optimize algorithms and ensure systems work as intended.
DNAT ensures compliant video data
Deep Natural Anonymization (DNAT), however, resolves this dilemma. Based on generative AI, this unique technology creates synthetic faces and replica license plates that prevent the original subjects from being recognized. By preserving the quality of the original data while ensuring compliance to global standards, it eliminates the compromise between privacy and innovation, and it is able to protect privacy in AI and machine learning projects.
In other words, DNAT allows you to run AI and machine learning projects on video data in line with global privacy standards.
This anonymization technique is much more valuable than simply blurring faces and license plates. Facial features and physical attributes can still be recognized, and data can be used to train machine learning models. Importantly, this approach ensures that video recordings remain compliant with strict data protection guidelines.
Philipp Wende, Senior Consultant Automotive & Innovation Program Lead, DXC
A whole new layer of privacy
DNAT automatically detects faces and other identifiable elements such as license plates in the original images and videos. It then randomly generates artificial replacements that accurately reflect the original attributes, such as facial features, expressions, gender, emotions, intent, or age. And it then applies these non-reversible overlays to the original, ensuring that re-identification by facial recognition technology is impossible.
Maintaining quality. Guaranteeing privacy.
In doing so, DNAT preserves semantic segmentation, making it the only anonymization technique capable of powering analytics and machine learning. DNAT solves the dilemma between high quality video data and data privacy.
The benefits of DNAT
DNAT is safe, because it makes re-identification by facial recognition technology impossible: synthetic faces are randomly generated and non-reversible. It is also highly accurate: age, gender, race, emotions, facing direction, and intention are retained for AI and machine learning development. And finally, DNAT is compliant, holding EuroPriSe certification for privacy-compliant IT products.
Taken together, Deep Natural Anonymization allows you to safely use videos and images to optimize AI and machine learning, yet without the threat of receiving heavy fines or halting innovation.
