In today’s dynamic digital landscape, the battle against illegal and harmful content online comes with all sorts of challenges. As organizations and platforms strive to identify and address such deleterious material, such as child sexual abuse material (CSAM) and copyright infringing content, they are faced with the urgency of efficiently identifying and addressing this problematic content.
At the forefront of this issue stands Videntifier Technologies, a company with an impressive track record in multimedia search infrastructure since its establishment in 2008. They aim to provide platform owners, law enforcement agencies, and Child Sexual Abuse Material hotlines with state-of-the-art video identification tools to detect and remove harmful content and supercharge their current tech stack to improve efficiency.
How Videntifier visual fingerprinting works
An integral part of Videntifier’s edge is its innovative approach to hash matching—a method of visual fingerprinting often employed for image detection purposes—which addresses some of the inherent limitations of cryptographic and perceptual hashes frequently employed. Cryptographic hashes can only identify exact matches, proving inadequate when even the slightest file alteration occurs. Perceptual hashes on the other hand are able to detect minor modifications such as resizing, compression, and contrast changes, but cannot cope with significant modifications such as cropping or embedding images. These widely-used methods fall short, allowing harmful or illegal content to persist online.
To bridge this gap, Videntifier utilizes local descriptor hashes, which offer unique properties that cannot be achieved with other hash types. Using these specialized hashes enables the identification of altered videos despite such modifications as cropping, bordering, embedding, and picture-in-picture. With this sophisticated and powerful feature, local descriptors have become invaluable tools for platforms wishing to monitor harmful and copyrighted content as well as ads.
Overcoming scalability and efficiency challenges, implementing local descriptor hashes has been a pivotal undertaking for Videntifier. The company’s skilled engineers have merged their in-depth visual fingerprinting algorithm with their patent-protected NV-Tree vector database through a two-pronged approach.
A remarkably flexible solution to content identification, Videntifier offers trust and safety teams with various search query types and versatile match options. As an efficient tool for extracting visual descriptors from video frames, the descriptor extraction component can process video frames and extract visual descriptors as images. Using fine-tuned extraction parameters, Videntifier ensures that the most suitable frames, descriptors, and levels of detail are selected for target identification. The proprietary NV-Tree vector database structure, purpose-built to accommodate billions of descriptors, plays a vital role in their ability to excel in this area. This capability enables rapid searches within vast libraries of video and images, thus ensuring local descriptor searches’ practicality and effectiveness.
Aiding the fight against harmful content and copyright infringement
Videntifier’s cutting-edge video identification solution has garnered high praise from trust and safety teams. Within seconds, moderators can query and pinpoint content with remarkable precision, all thanks to the company’s technology that allows for rapid and accurate search capabilities. The system also takes great care to ensure the confidentiality of client and user data, employing a one-way encoding method for queries to ensure user privacy throughout the process. Added to their repertoire of powerful tools is Videntifier Nano, a premier harmful content risk assessment tool that enables platforms to evaluate their vulnerability to hosting harmful content while staying compliant with evolving safety regulations. This potent amalgamation of features empowers platforms to combat harmful content and ensure users’ well-being effectively.
Blue Efficiency, an anti-piracy solution provider with a prominent reputation for its anti-piracy technologies, offers an informative case study illustrating the impact of Videntifier’s technology. After successfully testing Videntifier fingerprints on select works by Blue Efficiency in late 2022, they’ve made Videntifier fingerprints available to all their clients. With over 2000 fingerprints generated for a diverse range of content, including movies, reality shows, fiction, and entertainment programs, the collaboration yielded impressive results. Together with Videntifier, Blue Efficiency successfully detected and removed over 30,000 videos from TikTok in just six months. The seamless integration of Videntifier’s fingerprinting technology proved invaluable in Blue Efficiency’s relentless efforts to combat piracy and safeguard digital content.
Throughout these collaborations, Videntifier has established its central role in identifying and removing harmful and illegal content, ensuring a safer internet experience. Undeterred in its pursuit of advancements, the company continues to reshape the landscape of video identification by consistently providing solutions that safeguard users, protect data, and promote a more secure online environment.
To learn more about Videntifier’s work and expertise, visit their website at https://www.videntifier.com/