Fighting cybercrime with chatbots

GreenShoot Labs


GreenShoot Labs helped Cyber Helpline, a non-profit that provides free, expert support from cybersecurity experts, build a chatbot that can navigate the complex world of cybercrime to help victims contain threats and recover.

The challenge:

Creating a bot to guide users through a complex knowledge base 

If you experience a cyber crime you want help fast and at any time, day or night. Cyber Helpline approached GreenShoot Labs with a complex problem: how to build a chatbot that could translate a complex and technical knowledge base into frictionless natural language interactions, both to help victims get the advice they needed more quickly, and to reduce the caseload of the organisation’s volunteer-staffed helpline. GreenShoot Labs’ initial research into how people report and describe cyber crimes showed clearly that people did not talk in hacking or crime terminology and don’t understand how their personal devices operate at a low level.

The solution:

Combining NLP and machine learning to enable natural interactions

Most help systems ask users to go through a series of questions to aid slow self diagnosis, but the team at GreenShoot Labs knew that wasn’t effective enough. Seeking to provide a system that allowed people to provide as much detail as possible in the most natural way, their solution was to allow the user to write freely, letting them explain in their own language what has happened, and then to use Natural Language Processing and Machine Learning to interpret, categorise and diagnose the attack. Knowledge graphs were used to store possible attack information, and allowed the cybersecurity experts to train the platform over time.


0 %
Successful Diagnostic Rate

The outcome:

Satisfied users, reduced caseload for volunteers

After submitting their issue, the system provides users with details on the most likely attack and the next steps they should take to ensure protection. If the first diagnosis is not correct, the system offers the next most likely cause and learns from this data. Around 85% of users confirmed that the tool had provided a correct diagnosis, allowing the cybercrime experts to focus on the cases that need help the most. With only 1 in 10 cases now reaching volunteers, the helpline can deal with a higher volume of cases with minimal cost and resource requirements.


Get In Touch

If you are interested in the work above and think we can help you and your business, drop us a note. We will get straight back to you.