About FiB

"Your 2nd pair of eyes on the Internet"

Mountain View

FiB is a chrome extension for Facebook which detects Fake News on Social Media. As the user scrolls through his/her newsfeed, FiB marks every post as verified or non-verified, verified being not fake news. This chrome extension scraps the User's Facebook Page, while he is going through his/her timeline and obtains links from the posts as well as the pictures. For each link, our algorithm first checks if its a malware or phishing or clickbait link or not, if so then marks it as dangerous, else obtains a score based on how authentic the link is, between 0 to 100. If the score is good enough, it marks it as verified. If not, it obtains the header as well as the keywords from the link, checks on Google/Bing if anything on this content was ever shared by other sites. Each links returned by search bar goes through the same scrutiny of getting a score. If the score is good enough, we take that first verified link, get the content and show to the user as more trustable content. Our chrome extension works for images as well. Whenever someone posts a twitter snapshot, it converts the image to text, grabs the username, goes on twitter and checks if that post was ever posted by the user. If so, it tags the post as verified else non-verified. We also tag adult content/racy content/ pornographic content as non-verified, to reduce sharing of mms scandals.

View post on imgur.com

Founders :

Team Picture

Anant Goel, Nabanita De, Qinglin Chen, Mark Craft (left to right)

Team Picture

Qinglin Chen, Mark Craft, Anant Goel, Nabanita De (left to right). The Google Moonshot win at HackPrinceton 2016

Ideas Page for GSOC 2017 :

Team Picture

We are open to new ideas, however we encourage developers to check out our Github page and get started from there. Look at our code vulnerabilities and get in touch with our team, on what you want to work in. Given below are some of the possible topic ideas -

1. Implementation of hover option on the verified/non-verified tags

Summary : On hovering over the verified/non-verified button, the summarized content from verified sources would show up. The summarized content is already processed in backend, it just needs to be linked to front end.

Mentors : Mark Craft (Mscraft2@illinois.edu) , Qinglin Chen (qchen50@illinois.edu)

Skills: Javascript, CSS

2. Building a Fake news detector using Machine Learning

Summary : This would involve collection of fake news from different sources and building a classifier which will detect fake news.

Mentors : Nabanita De (nabanita@cs.umass.edu)

Skills : Python, TensorFlow

3. Scaling servers for accomodating more users and building a less throttled service

Summary : Right now, our product was released only in the US but the no. of user requests per second was phenomenal - around 50,000. Thus we need to scale our servers which is originaly done using Heroku. It would include either moving to AWS or Azure. It would also involve creating a database of processed links, so that it can be fetched easily while serving new users.

Mentors : Anant Goel (anantdgoel@gmail.com)

Skills : Javascript, Python, Flask, AWS, Azure

4. Writing APIs to provide Fake news detection as a Service

Summary : Right now, our product is just a chrome extension. Our next steps would be trying to create our backend of processing links to identify if verified news or not - building an API service out of it. We would want to provide our services in Javascript and Java as well, so our backend code in python would be needed to be converted to Javascript and Java.

Mentors : Nabanita De (nabanita@cs.umass.edu)

Skills : Java, Python, Javascript

5. Re-ramping our Website.

Summary: Creating a more dynamic website for our product, as compared to our initial version.

Mentors : Anant Goel (anantdgoel@gmail.com)

Skills : Javascript, HTML, CSS

6. Building a Fake news detector Chat bot

Summary : This Bot would accept links, texts or pictures as input and would tell the user if it is verified news or not. If it cannot verify the news, it will use our algorithm to show content from verified sources.

Mentors : Nabanita De (nabanita@cs.umass.edu)

Skills : Java, Python, Microsoft Bot Framework, Facebook Bot

Email us at nabanita@cs.umass.edu for more details.

The general format for applying for the proposal would be :

1. Name

2. University

3. Relevant Background

4. Skills

5. Interested projects

6. Hours you can contribute, per week. Also include any prior commitments you have over the summer

7. Your contribution to our extension

8. Proposal

You can also join our slack channel to talk to us about next steps.

Our Slack Channel »