Learning from failure: Trend analysis for compromised accounts and passwords

点击翻译按钮获取中文摘要

Computer accounts get compromised for reasons: some used weak passwords; some created a digital exhaustmuch higher than normal; some were at high-risk positions (e.g., executives, accounting) and were exposed totoo many attacks. Learning lessons from compromised accounts and password use would help people designbetter defense technologies. Following this principle, this Engage project is to develop a set of new algorithms,which automatically learn the trends in compromised accounts and password use and identify potential threatsearlier. The project plans to answer the following questions related to computer security: how and where toretrieve relevant information from publicly available data on compromised accounts? How insecure are mostpasswords? How long would it take to brute force crack/guess them? Do the passwords being used follow anybest practice (e.g., letters + numbers)? Are there correlations with password security and the user's professionalpositions (e.g., executives, managers, and accounting)? Who are the accounts most likely to get compromisedand what industry are they in? Do the compromised accounts have a higher than normal digital exhaustpresence? With the patterns learnt from trend analysis, could we build a model to predict which accounts arelikely to get compromised? Could the model also give recommendations on how to patch the security problemsof accounts that are likely to get compromised?The output of this project includes: (1) a data collection module that can retrieve and store information frompublicly available data on compromised accounts from dark web, (2) a set of algorithms that can answer theabove-raised questions in trend analysis, and (3) a prediction model that can predict a user's likelihood to becompromised.