Explore machine learning in browser

Teachable Machine allows you to study machine learning straight from the browser, without any coding
05 October 2017   1639

Recently Google Creative Labs released the Teachable Machine - online application, that allows everyone to get familiar with basics of machine learning and neural networks.

Teachable Machine is an experiment that makes it easier for anyone to explore machine learning.

You can teach a machine to using your camera, live in the browser – no coding required. You train a neural network locally on your device, without sending any images to a server. That’s how it responds so quickly to you. 

Developers recommend to capture at least 30 images per class. Be aware of when you’re pressing and releasing the button (that’s when it starts/stops capturing images). And you might need to capture lots of angles or variations of whatever it is you want your machine to recognize.

Hype.codes team made a small research and explored this services. As a result, when a author of this article shaked his hand, a cute cat animation came up; when shaking a head - a dog animation came up. This took less than 2 minutes and was quit funny.

Teachable Machine
Teachable Machine 

As you can see on a screenshot above, the neural network can learn 3 different moves from a webcam and bind it to a 3 different types of GIF animations, sounds or speechs. Also, when performing a move, there is a "Confidence" bar, that shows the condifence level of machine. 

Learn more at GitHub and check the website to explore the world of neural networks. 

Third Party Apps Could Read Twitter Messaging

According to the company, no one used this vulnerability and the issues is now solved
18 December 2018   665

Until the beginning of December, third-party applications could access Twitter private messages. According to the company, no one took advantage of this vulnerability. Terence Eden, who found it, was paid almost $ 3,000 under the Bug Bounty program.

In 2013, there was a leak of keys to the Twitter API - so applications could access the interface bypassing the social network. To protect users, Twitter implemented an application authorization mechanism through predefined addresses (Callback URL), but it didn’t suit everyone.

Applications that do not support Callback URLs could authenticate using PIN codes. With this authorization, a window pops up that lists which data the user opens to access. The window did not request access to private messages, but in fact the application received it.

On December 6, Twitter reported that it had solved the problem. Judging by the statement of the company on the HackerOne website, no one had time to take advantage of this vulnerability.

This is not the first social network security error related to the API. In September, Twitter discovered a bug in AAAPI (Account Activity API): the system sent a copy of the user's personal message to a random recipient.