Odysseus

Phishing has become a major problem in recent years. Phishing is the criminally fraudulent process of stealing personal / financial information from internet users. Usually the phisher's modus operandus follows the same method. A spam (bulk email) system is used to send phishing emails to many people. Often these emails threaten that the user's account at Paypal / Ebay / AIB etc is going to expire and asks the user to log onto the website to update their details. When the user clicks on the link embedded in the email they are brought to a site that mimics the real one and they submit their personal information. This can then be used by the phisher to access their financial information.

Phishing detection has been studied for many years. These systems commonly use machine learning techniques in order to classify an email as phishing or non-phishing. Another area of interest is to be able to extract the target from the email body. By this we mean that we want to extract automatically the corporation that is being targeted by the scam in an attempt to give early warning to the corporation.

The tasks that will need to completed are:
• Mark-up a dataset: Current datasets on phishing emails do not contain target information. Emails will need to be read in order to get this information.
• Perform background research on information extraction.
• Build a baseline system for target extraction (based on a simple technique such as word frequency)
• Build an IE system to recognise the target automatically.
• Compare the performance of your IE system with the baseline.
• Produce a poster describing your work.

Informal enquiries can be directed to Fergus Toolan (fergus.toolan@ucd.ie). Applications should be
completed in the normal manner.

Pre-reqs: Knowledge of Java or Perl required.