Inspiration

We just saved the Internet

1 out of 3 boys of the age of 13 has been exposed to internet porn. Most of them, by accident. By the time they reach that age, they may not have held someone's hand.

Kids should not be exposed to Sex, Violence, Verbal Abuse and Death like an adult is. There is a large difference for a child to see Porn for the first time than for an adult. Educating the youth on those subject takes time but we can help the development by leveraging AI. Delaying the exposition to some sexual and gore content will give more time and resources to the mentors to teach kids about those tough subjects.

What it does

Soteria leverage state-of-the-art models and machine learning techniques to understand the content and meaning of a website. With this information in mind, Soteria will obfuscate parts of the website that are considered sexual, violent, racist, abusive or as a distraction.

Whether you are a professional or student who wants to get more work done or the parent of a young child, Soteria will deliver you value:

Professionals: The Internet contains more distractions than anywhere else on the planet. Remove those distractions and become more productive by enabling Soteria. All content considered NSFW will be blurred out!

Students: One distraction can take up to 20 minutes delay in getting a task done. Unfortunately, students use tools such as facebook and Reddit for communications and research which are a huge cause of distraction. Remove those distractions with Soteria!

Parents: Safe your child from bad influences and sexual/gore content with Soteria. Studies show that teaching those subjects by an adult is more valuable for their development. [1] Plus, with our NLP technology, we can also obfuscate all cyberbullying comments from any webpages!

How I built it

Python was super helpful! The front-end chrome extension was developed in JavaScript. The chrome extension interacts with a back-end server developed in Python. The machine learning models were trained in a cloud environment and their weights are downloaded to perform local inference. The back-end server consumes the machine learning models when inference is needed and returns to the chrome extension information on what needs to be censored.

Challenges I ran into

Integrating so many moving pieces together :slightly_smiling_face: ! The front-end, back-end, and machine learning tech were developed in different libraries, languages and by different people.

Accomplishments that I'm proud of

Developed an end-to-end machine learning product who's parameters can be adjusted by the end-user.

What I learned

Learned Flask back-end development and got to practice some deep learning training and deployment

What's next for Soteria

Use the technology we developed for cyber-bullying prevention and future social-good technology applications

[1] : https://www.youtube.com/watch?v=TCY2dOf2eMs&t=1016s

Built With

+ 7 more
Share this project:
×

Updates