Predicting cyber offenders and victims and their offense and damage time from routine chat times and online social network activities

• Published in: Computers in human behavior 2022-03, Vol.128, p.107099, Article 107099
• Main Authors: Yokotani, Kenji, Takano, Masanori
• Format: Article
• Language: English
• Subjects: Anomaly detection; Cyber offenders and victims; Damage; Internet; Machine learning; Predicting offense and damage time; Routine activity theory; Social contagion; Social networks
• ISSN: 0747-5632; 1873-7692
• DOI: 10.1016/j.chb.2021.107099

Predicting offenders and victims and the timing of their offenses and damages has become a perennial challenge in global cyberspace. This study aimed to solve this challenge through analyses of routine chat times and online social networks. We sampled more than 550,000 online users over 6 months. We also used unsupervised and supervised machine learning to predict cyber offenders and victims and their offense and damage times, respectively. Our predictors, based on routine chat times and online social network inputs, identified future cyber offenders and victims within 2 months. Furthermore, we predicted the hours and days of their offenses and damages within a week. Extraction of routine chat times and online social networks from chat data can help predict future cyber offenders and victims. Our cyber offense and damage time predictor could help prevent future cyber offenses and damages and promote a safer cyberspace for everyone. •We tracked more than 550,000 online Pigg Party users for 6 months.•Our predictors identified future cyber offenders and victims within 2 months.•We predicted the hours and days of their offenses and damages within a week.•Chat data were useful for extracting routine chat times and online social networks.•Routine chat times were useful for predicting future offense and damage times.