How to host your first online hackathon without breaking the bank: expectation and reality
Last winter we tried out the hackathon format for the first time. Due to restrictions on mass events, it was held online, but this did not affect our expectations in any way. We have long wanted to increase employer brand awareness outside the cybersecurity market – among developers and other IT specialists who often have no idea about us. In addition, we wanted to understand whether it is possible to meet promising specialists at the hackathon, and we were also very interested in learning the inner workings of professional organizers of such events in order to learn how to conduct them ourselves.
If you have been haunted by the idea of holding a hackathon for a long time, but the veins are shaking, and you do not know where to start, welcome to cat. We will tell you how it happened with us, how to get around our rake, and how expectations can diverge from reality, as well as whether it is possible to recreate the atmosphere of an offline hackathon online.
How did we get the idea to host the first online hackathon?
In addition to everything that we have listed above, we were faced with the main task – to solve a practical problem that one of our teams could not approach for a long time. We had a text corpus, and on its basis it was necessary to create an algorithm that would select from this corpus named entities related to the threat intelligence domain: threat actors, malwares, indicators of compromise and others. The nuance was that the text corpus was in English. The task was non-trivial, interesting, but we didn’t get our hands on it for a very long time.
Why a hackathon when there are many other tools?
In general, to solve the main task, it was possible to hire outsourcers, and to search for candidates, it was possible to more actively involve HR and concentrate on recruiting. But suddenly a representative of the Pystech.Genesis company, which specializes in hackathons for IT specialists, contacted us. We started discussing the organizational process and eventually decided to test a joint online hackathon with the support of a professional organizer of such events. Since we ourselves had plans to hold hackathons in the future, we wanted to watch live how specialists with rich experience behind them do it.
We have selected a large number of experts from our company in order to be able to observe at every stage what and how our organizers are doing. So we managed to collect information on how to properly organize this process. It turned out that there is nothing difficult in this, but you need a sufficient amount of time and, most importantly, human resources – people who are ready to immerse themselves as much as possible in the organization of such a movement. We discussed the conditions with the organizers and began to prepare for the first joint online hackathon.
How did you choose the task?
The task selection process itself was fun. It was attended by employees of two divisions at once: the Center of Expertise and the Development Department. The first helps clients and partners to build information security processes based on our solutions, and the second is divided into teams based on the products being created. In fact, departments and individual teams competed with each other, and they were all interested in choosing exactly their task.
We held a separate meeting, at which those who wished to pitch each of their tasks and explained to those present why their task should be completed at the hackathon. For example, the Center of Expertise proposed the development of a new license generator for the company as a task. There were many interesting tasks, but in the end the task from the R-Vision Threat Intelligence Platform (TIP) product team won, which is a platform for managing cyber intelligence data.
The essence of the task
Initially, the task in the formulation of the TIP team was specific and too specific for the information security sphere. This was not very successful, since the hackathon assumes a certain leeway for the team. Therefore, we decided to split the problem into two parts.
-
The main task was quite broad: to build an interface for collecting, describing and analyzing threats, indicators of compromise based on real documents.
-
Further, the teams were given complete freedom to create a product. The result could be a program, an ML-model, or another implementation that would allow solving applied problems of collecting and analyzing unstructured data about information security threats. The plan was to give preference to the most complete solutions that would cover the end-to-end problem.
The surprises we faced
When we finally approved the assignment, the organizers began to engage the audience. Unexpectedly for us, applications were submitted by many well-trained teams that specialize in participating in hackathons and are interested in financial rewards. We wanted more amateur hackathons to come to us, because it was among them that we hoped to meet potential applicants.
Initially, we counted on about 60 participants, that is, no more than 10 teams. But due to the fact that this was the first online hackathon in Russia this year, the number of applicants exceeded 200 people. As a result, we were forced to carry out the screening, because from the point of view of the organization, we simply could not stand such a number of teams. In the end, we selected 13 teams, of which 9 reached the actual implementation – the prototype of the solution, the rest dropped out at intermediate stages.
How did the preparation for the hackathon begin?
Our assignment had a certain information security specificity, we realized that we need to explain the task to the participants who do not know at all what information security and indicators of compromise are. To do this, Anton Solovey aka @likeafreedom, product manager of R-Vision TIP and concurrently “toastmaster” of our hackathon, was responsible for interaction with the participants: preparing assignments, presentations, pitching with teams together with other experts of our company, – doing everything necessary, for the hackathon to take place. To introduce developers and designers who are far from the field of information security, we decided to move away from the positioning of our products and not explain what IRP, SGRC, TIP, SENSE and Deception are, therefore, in the presentation about the company, we visualized our solutions in the form of mascots.
Many people were involved in preparing the presentation – they read syntax errors, suggested how to improve the presentation, which slides to insert in order to correctly present our company.
As a result, we came to the conclusion that it is necessary to involve experts from our side in the hackathon, who would participate in the teams’ pitches, accept assignments and evaluate them, answer technical questions of the participants. We have assembled a large team of experts: it included product experts, datasinter, frontend and backend, testers, project manager and even the head of the development department.
Hackathon in numbers
-
Duration 45 hours…
-
More 220 applications for participation, competition 4.5 people per seat…
-
13 teams, 9 of them reached a ready-made solution, 6 made it to the final, of which there are 3 winners…
Hackathon stages
-
Friday evening. Acquaintance. Task story, presentation by R-Vision.
-
Friday night. Development of a prototype.
-
Saturday morning. Pitch decisions: what we managed to do overnight. 5 minutes per pitch.
-
Saturday, all day and night. Second pitch: a relatively ready-made version.
-
Sunday morning. A ready-made solution for expert evaluation.
Decisions that the commands executed
The participating teams worked out the presentations of their solutions and the design of the tool. Among the interesting ones, we can highlight the solution of the Digital Rover team, which came up with something like Alice for infobez – a virtual assistant based on artificial intelligence for describing threats based on real documents. The solution based on the ruGPT3 language model is a bot for Telegram that allows you to write something in the native language in text or send a voice request. The tool has our entire corpus of texts, which it translates into English. Then it translates the received text or audio message into English, analyzes the correspondence of the request to the content of the document corpus and provides the user with relevant information. For example, you can write or say: “Tell me about such and such a vulnerability” or “Write more information about such and such an indicator of compromise”, and the bot will respond with extracts from the text corpus.
We had a questionnaire, where each expert put estimates for his block on a 10-point scale, at each stage the experts set intermediate marks. Initially, one winner was conceived, but after the final summation of the weights (points) for each examination, it turned out that the three teams had such a minimal difference in points that we decided to determine three winners who shared the prize place. All of them had very interesting solutions, it was simply impossible to single out one of them. As a result, the following teams became the winners:
-
ARES. Development of a product for analysts in the field of information security – a knowledge base that aggregates data sources with the analysis of computer security threats, reference books with terms, relationships between entities of the subject area. Tasks that were embodied in the product:
-
Creation of a system for filtering analytics sources by keywords
-
Annotation of new reports to highlight terms related to information security analytics and add a document to the knowledge base
-
Connecting help systems (MITER and MISP)
-
Visualization of relationships between entities of the information security domain
-
-
Stellar. The R-Vision Assistant project is a service for collecting, visualizing and analyzing data on cybersecurity threats. The service features a convenient and understandable UX / UI interface, the ability to work with various sources of data on cyber threats and add new ones. The solution contains many options for visualization and analytics.
-
DEV Labs. An interface for semi-automated parsing of basic entities from unstructured data sources, within which an information security specialist analyzes vulnerabilities. The program aims to improve the quality of data stored in the knowledge base.
Resourcefulness is welcome
The organizers created a common Telegram channel to communicate with experts. But some participants, whether in order not to disclose their idea, or to gain a competitive advantage, began to write to our experts in a personal, for example, clarified how interesting this or that idea would be as a solution. We saw nothing wrong with this, since it is also part of the work on the product and it was not prohibited by the rules of the hackathon. In order not to create too much competitive advantage in the end, we, of course, tried to transfer all communications to a common channel, but we still appreciated this call positively, because this is a completely natural story for interaction with the customer.
conclusions
What did we manage in the end
We were able to use the hackathon as an opportunity to tell about us as a company and as a result selected 15 people for further communication about employment. However, the specialists who seemed interesting to us were not interested in changing employers, but were only ready to work on individual projects.
What could have been done better
We have allocated too many experts from our company, it took a lot of our employees’ time. Next time, we will reduce the number of experts by at least half.
It also took us a lot of time to deploy products at our stands: 10-15 minutes were spent on each team. For the future, we decided to give the teams a pre-formed stand so that they had the opportunity to deploy the ready-made solution themselves.
Future plans
We have gained useful practical experience and will definitely use it in the future. Now we plan to hold our own internal hackathons once a quarter and periodically organize external ones with the involvement of contractors. It is possible to do this on your own, but the support from a professional hackathon organizer allows you to reduce the number of experts from the company to 2-3 people, which means saving employees’ working time.
