Autism Disclosures and Cybercrime Discourse on a Large Underground Forum

Prior research on the relationship between autism and cybercrime has been inconclusive. While some research suggests those who had an autism diagnosis are more likely to engage in cybercrime than those without, other evidence indicates a diagnosis of autism is associated with a lower risk of cybercrime offending. Prior research has primarily relied on self-report survey data. To the best of our knowledge, data gathered from cybercrime-related conversations on underground forums has not previously been used to study the relationship between autism and cybercrime offending. This research applies natural language processing (NLP) techniques to a large underground cybercrime forum data. We developed two NLP classifiers to automatically categorise the context and content of forum posts. We find that terms related to autism were mostly used in a negative context, primarily to insult other users. We find that actors who self-declare as autistic, post more frequently on the forum than those who do not disclose to be autistic. Despite the increased frequency of their activity, we find those who disclose they are autistic are less likely to discuss cybercrime-related matters, compared to a matched sample of users with similar posting activity.