Future development plans for Xevil Captcha solver and Xrumer in 2020
I know many of the Asia Virtual Solutions customers use captcha solvers with their tools, and many of you own a copy of Xrumer and using their awesome Xevil Captcha solver. Since you maybe know by now, Xrumer is developed by a Russian company called Botmasterlabs, and although they do have a English forum with some information, the bulk of the focus is on the Russian forum of Botmasterlabs, and that is where most of the latest news are posted first and where you will find the most guidelines and technical discussions. So if you cant read Russian or have not figured out yet how easy it is to use Google translate, you will not be able to read the breaking news about Xrumer and its Xevil captcha solver.
Too help the English speaking users of Xrumer and in particular Xevil, I thought it will be use full to know about a post the was made in the Russian forum about the future development plans for Xevil in 2020, so the below post is a translation of the post earlier this week in the Botmasterlabs Russian forum ( Original Russian Post here ) . I have translated the text and also the diagrams for you.
The Post from Botmasterlabs:
Since the beginning of this year, our team is actively working on a number of innovations, as well as on the preparation of the new XEvil 5.0, and in this current we would like to outline our plans for the current year – first of all, to make it clear: what exactly and when ( approximately) will appear this year.
New XEvil 5.0
The key and most important advantage of XEvil 5.0 ahead of XEvil 4.0 will be deep optimization of ReCapchi-2 recognition: as previously reported, Since the beginning of this year, we have started work on a browser-less reCapcha-2 processing mechanism based on our own JS solution, which will reduce the load on CPU and memory consumption at least 10-20 times – and as a result, significantly increase the ceiling of streams to work with ReCaptcha-2: on the “average” iron will be possible work in 200-300 streams, on a more powerful – 500 and more powerful – in 500 and more flows. At the same time, after this innovation, the mechanism of recognition of the images reCapchi-2 will give a greater load, rather than “heavy” processing of clicks on these pictures – as the flow of processed images in a unit of time will be greater. This is the first stage that is being worked on right now.
The first beta version of XEvil 5.0 (where this phase 1 feature is implemented) will be available approximately in May, or later. The XEvil 5.0 early access mechanism will be similar to gaining access to XEvil 4.0 – and more details on this issue will be published in the coming days. Let me remind you: no single purchase of XEvil 5.0 isnot discussed, it will only require the introduction of subscription.
The next stage (presumably July – August) is to expand the ability to recognize conventional caps. The imaging module is planned to be put into a separate 64-bit module (a similar procedure has already been done in XRumer-e), which will significantly reduce the flow of resources, and as a result – will increase the stability of work. Thanks to this “unloading”, it will be possible to expand the range of loaded neural networks – and will be added a neural network with sensitivity to the register of letters, as well as Cyrillic neural network. Plus – separately, the pre-neural network under the updated capsBing, Rambler and a number of others.
In the third phase, GPU support will be implemented, and as a result, NVidia video card owners with CUDA technology will be able to further improve XEvil’s performance by order, by switching the processing of neural networks from CPU to GPU. In addition- will be increased ergonomics and informative interface, optimized other functionality.
And, of course, before the release of the first beta version of XEvil 5.0, there should be a general release of XEvil 4.0 – this version will finally be available to all license holders Standard and Business, without any additional payment.
One of the main drawbacks in the use of our complex today is the need to launch XRumer to work with XEvil and SocPlugin, even those who do not need XRumer. This is due to the historical mechanism of protecting our complex from hacking, where all protection is tied to XRumer. It has shown its high reliability – to this day nowhere in the network is it impossible to find either the hacked full version of XEvil, nor XRumer (versions of the last 5 years), and yet – requires a complete restructuring. First of all – for the ability to run and use XEvil separately from XRumer.
Work is already underway to prepare a cloud solution for the authorization of our software, and the first options for updated authorization should be ready for the summer. Before the start of the software will require the launch of a separate mini-application (Launcher), through which will also automatically download and install updates, which will also increase the convenience of use.
Another advantage of this innovation is the increase in stability and independence from the operation of our servers, as the cloud solution is a priority more reliable depending on one or two servers. The iron binding will also be abolished (this mechanism is outdated and strongly hinders the use of XRumer-and with the changing configuration of iron) – but, at the same time, will be implemented control sessions, which will stop attempts to resell/transfer software in third hands, and unauthorized sharing.
As the experience of using crowdsourcing platforms for learning new textcaps (logical anti-bot-issues – “Name the capital of Britain” etc., used in trust forums, the new approach is very effective: last year we managed to improve the efficiency of XRumer-a on ordersof magnitude. Of course, this effect is short-lived, because the administrators of popular resources change text captchas almost immediately, when their resources are a stream of spam. However, for a few months such an update gives a very good increase in the return on the software – and this is 100% justified, because 1 backlink from the trust resource is valued higher than 1000 backlinks from the spammed.
In the last big update – XRumer 19 – software through a new crowd sourcing platform was trained 45,000 text captchas. Now we have collected material on more than 80,000 text captchas, and it continues to be replenished. All that’s left to do is train them. However, this is a very low-cost process (we have spent more than $7,000 in the summer of 2019), and in addition – not fast (those who participated in our contests on learning text captchas,well remember – that to find an answer to the next question can take up to 10-20 minutes, study Wikipedia and other sources).
Therefore, it is possible that we will organize a kind of crowd-funding to speed up learning new text-caps, and participants will be able to get an improved version earlier. This approach was offered to us by several active XRumer users, and we can balance it so that those who participate in crowd-funding, and the rest are not deprived of the benefit
In addition, literally in recent days, we have begun experimental work on automatic training XRumer-and new text captchas “on the fly.” How it works: It turned out that answers of more than 70-80% of captchas are quite possible to find in Yandex/Google. The biggest challenge here is finding the right answer in the issuance: it’s not enough just to take the first result of the issue, or the Wikipedia headline, it will in most cases give an error. That is algorithmically it is almost an insoluble task. However, recent experience with neural networks shows that this task is quite solvable – through a special configuration of the neural network, in which as a support will be a ready linguistic neural network, as well as other introductory parameters. This approach by the end of this year will increase the return from XRumer by order,and in fact will instill new life in this tool.
Applying neural networks to XRumer and SocPlugin logic
As many have noticed, we are increasingly focusing on neural network programming as the most promising direction for the coming years. This experience can also be applied to related tasks, in addition to captcha recognition, such as:
- “Meaningful” comments in XRumer (similar to“Antispam”but even more human-like)
- filtering target profiles in SocPlugin, for example by appearance in a photo
- and other areas
Of course, in addition to this, other tools of our complex will be updated, in particular in the next 1-2 weeks there will be an update of Hrefer, in addition to the next updates XRumer-a, XEvil-a, and SocPlugin.
Work in crisis
Our world has entered the culmination of a global economic restructuring – and it will touch (already touched) absolutely everyone. Trade chains, economic and political relations will undergo significant changes. For those who are “in the subject”, this should not be a surprise – as the coming economic collapse warned absolutely all experts in 2019,including the most optimistic. There is no need to be afraid of this, because the old one will be replaced by a new, much more interesting one.
Our online sphere will be the least affected by this – because we are already working in a “new” economic paradigm, without realizing it. Although all sorts of “grey” and “black” schemes will continue to lose their relevance – on all fronts will continue to tighten the nuts against money laundering, the opportunities of stock players and many other “employees” whose craft yesterday will decrease seemed normal.
Rebuild your business in more “white” and useful to society, help family and friends who have been affected by the crisis the most.
Now, when many people lose their jobs, a good alternative can be earnings online – and the relevance of online courses, mentoring, referral schemes will grow. Recruit students, show them a new way and new opportunities – but please, let it not be dying doorways and breathing on incense black SEO, and much more promising and useful directions.