Concept of using images for website input verification

The verification methods we are currently using are mainly verification codes, allowing users to enter verification Characters Based on images. This method is still safe, but it will bring inconvenience and troubles to users, for example, the Yahoo verification code:

This is highly secure and cannot be identified by machines or people.

 

In fact, we need to make people understand and machines dizzy, as long as we come up with our human strengths-image recognition, imagine using images for verification will be good:

The above example demonstrates the image verification interface. It is easy for users to perform verification. They only need to click the category of the image and enjoy it; however, it is difficult for a machine to understand the image content when its head is too big?

 

Some people say that they can use complex techniques such as pupil recognition and facial recognition to determine whether they belong to the category. I think that's just a category, and they have to take photos of the front face, in addition, it is sometimes possible to classify the gorilla as a human. If we further classify the person into two categories: male and female, the machine will become bigger.

 

If you are not at ease, you can add interference to the output image to fuel the machine:

As a result, the machine will suffer from almost no increased recognition difficulty for humans.

 

If you reset random interference regularly, you can also avoid repeated calls after the machine includes images.

 

Another way to Prevent indexing is to constantly change and expand the image library. However, generally, lazy webmasters do not like to do this frequently. Here I suggest various websites that provide image service.(For examplePicasa)To take up this task, provide the image verification service to external users, and use a wide range of image resources to serve other websites. If users find pictures interesting, you will also click the "Show Original graph" link to visit the website that provides services.

 

Of course, apart from image recognition, the human brain is also good at many other jobs, such as voice recognition, word association, and content understanding. You can also think about better verification solutions, this allows websites to enhance security while providing users with a more user-friendly experience.

 

Download the PDF version of this article: http://uushare.com/user/icesee/file/1265130

 

 

Q &:

-- Reference --------------------------------------------------
Yangq: raise the following questions:
1. How many times is the time and bandwidth used to load such a large image as a verification code?
2. It is easy to use machines to brute force crack the platform. Even if you do not submit the platform once, you can change the image to verify it. The probability that the machine is right is quite high ~
3. The key issue is that the images must be manually processed and then input into the database, and the options must also be manually input, because it is difficult for the machine itself to identify who are people and what are things.

How can we solve these problems?
--------------------------------------------------------
All of you have similar questions. Let me answer them:
1. Why is the bandwidth overhead not abnormal when the image size is not necessarily so big? Normally, there are also a lot of images on your website that have been accessed. For example, forums and image sites are much larger than the overhead. In addition, if a third party provides services as I said, there will be no problem.
2. options can be added, and the image library can be expanded constantly. If 100 random images and random interference are added, the machine cannot know which image has appeared. Why? Chaos can be guessed, but explosion-proof methods are certainly there, such as Retry Interval and maximum number of retries, especially this is a very simple verification method for humans, the retry requirements can be much stricter and will not affect the use.
3. manual input is not a big deal of labor. It is nothing more than looking for an image to upload and select a category. If you get two images every day, it will take a considerable amount of time to cope with most situations.