PrivaSee AI

Safeguarding your Privacy, Pixel by Pixel

Comment 1

Inspiration

While people are constantly taught how to avoid danger and risks in daily activities, cyber risks remain hidden from people’s eyesight. When people post pictures or videos, sometimes they overlook what can cause information to leak in their background. This can not only potentially cause personal data leaks, but can also threaten one’s life. Our team wants to design a solution that helps protect people’s privacy while allowing them to enjoy expressing themselves online.

What it does

Our cybersecurity service employs AI visual detection to safeguard private information within photos, videos, or the live camera on the website. Whether users are submitting media or engaging in real-time interactions, the system actively scans for potential leaks of sensitive information, including numbers, words, or on-screen content. If a potential privacy breach is detected, the website promptly notifies the user, delivering real-time alerts for immediate user action. This comprehensive approach not only covers conventional forms of private information but also extends to diverse data formats, ensuring a secure and vigilant online experience for users.

How we built it

The inception of our cybersecurity service involved an intensive process that began with brainstorming ideas and conducting external research to comprehend the intricacies of AI models and image recognition tools. After a careful evaluation, we opted to implement YOLOv8 for its robust image recognition capabilities. The development phase continued with the utilization of Roboflow, a tool that facilitated the pre-training and labelling of our computer vision model through meticulous annotation of each hand-crafted image. The model was then trained with meticulously cleaned data, and subjected to thorough testing using various sources such as webcams, pictures, and videos.

Challenges we ran into

Embarking on the creation of our cybersecurity service posed its own set of challenges. Initially, the process of learning AI models from scratch proved both exhilarating and daunting, as we navigated through the excitement of discovery while grappling with frustrations in understanding fundamental concepts and establishing the initial environment. Setting up the framework involved extensive troubleshooting, with trial and error on different machines, following diverse instructions to identify and rectify issues. Additionally, the adoption of cutting-edge technology introduced its own hurdles, as the tools and frameworks we utilized were relatively new, lacking comprehensive support systems. The latest version of Ultralytics, for instance, was still in the process of patching, leading us to encounter unresolved bugs when configuring the code. Despite these challenges, our commitment to overcoming obstacles and refining our cybersecurity service underscores our dedication to delivering a reliable and innovative solution.

Accomplishments that we're proud of

Our journey in creating the cybersecurity service was marked by notable accomplishments as a cohesive team. Efficient team organization played a pivotal role, fostering strong morale during the hacking process as we adeptly distributed tasks and assumed individual responsibilities. Ensuring psychological safety within our team was equally paramount, allowing everyone adequate time for sleep, meals, and short breaks. Beyond the shared objective of creating the cybersecurity service, we fostered a sense of camaraderie through engaging activities like red bull pong and collaborative slime-making. One of our most significant achievements was overcoming the initial uncertainty about our ability to construct an AI model for the project. The fear of this daunting obstacle was transformed into a shared dedication to learning. Witnessing our hard work materialize into a tangible AI model, gradually taking shape with the data we meticulously collected, was truly gratifying. The realization of turning something we once hesitated to imagine into a reality marked a triumphant culmination of our team's efforts.

What we learned

The process of creating our cybersecurity service has been a profound learning experience for our team, yielding insights into crucial areas. One notable revelation was the realization of the underemphasized significance of cybersecurity, particularly in its potential impact on individuals. Through our external research, we discovered a scarcity of products in the market designed to assist people in preventing information leaks, despite the widespread use of social media. This underscored the importance of prioritizing cyber security to safeguard personal information online. Another significant learning curve for our team involved delving into the realm of AI and machine learning. Prior to this hackathon, none of us possessed experience in training AI models. This endeavor proved invaluable, not only enhancing our understanding of neural network concepts but also providing a comprehensive learning experience in the intricate process of training AI models, starting from fundamental stages such as data gathering and cleansing. The knowledge gained in both cybersecurity and AI has not only enriched our skill set but has also deepened our awareness of the vital role technology plays in securing personal information.

What's next for PrivaSee AI

The next phase for our cybersecurity company involves an expansion of features to further enhance user privacy. We are introducing the ability to automatically blur flagged portions of images, providing users with an added layer of control over their shared content. Additionally, our system will now incorporate facial detection technology specifically designed to identify and protect children's faces within images. The introduction of automatic, real-time blurring ensures immediate action in response to potential privacy breaches. Furthermore, our enhanced capabilities extend to the detection of explicit content, offering a more comprehensive and proactive approach to maintaining a secure online environment for our users.

Built With

Share this project:

Updates