Wenyin Liu,Xiaotie Deng,
Guanglin Huang,
and Anthony Y.Fu
City University of Hong Kong
An Antiphishing
Strategy Based on Visual
Similarity Assessment
The authors’ proposed antiphishing strategy uses visual characteristics to identify
potential phishing sites and measure suspicious pages’ similarity to actual sites
registered with the system. The first of two sequential processes in the
SiteWatcher system runs on local email servers and monitors emails for keywords
and suspicious URLs. The second process then compares the potential phishing
pages against actual pages and assesses visual similarities between them in terms
of key regions, page layouts, and overall styles.The approach is designed to be
part of an enterprise antiphishing solution.
P
hishing attacks are increasing in fre-
quency and sophistication. The Anti-
Phishing Working Group (APWG;
www.antiphishing.org) recently reported
that the number of attacks is growing by
50 percent per month, with roughly 5
percent of recipients falling victim to
them. Phishing Web pages generally use
similar page layouts, styles (font families,
sizes, and so on), key regions, and blocks
to mimic genuine pages in an effort to
convince Internet users to divulge per-
sonal information, such as bank account
numbers and passwords.
To confront those challenges, we de-
veloped an antiphishing strategy that
uses a visual approach to detect bogus
Web pages. To monitor phishing attacks,
site owners can register their true URLs
and associated keywords with our Site-
Watcher system (currently running as a
prototype at http://antiphishing.cs.cityu.
edu.hk). Given that most phishing attacks
are initiated via email, SiteWatcher is
designed to run on mail servers and mon-
itor and analyze both incoming and out-
going messages for potential phishing
URLs. (We can deploy the antiphishing
service with providers that agree to joint-
ly support it with us.) If a message con-
tains keywords associated with our client
sites, the system considers all the URLs
embedded in the message to be suspicious
and flags them for further investigation,
comparing the Web pages at the suspi-
cious URLs against those protected pages
designated by the keywords.
We’ve built a prototype, which is run-
ning on one of the City University of Hong
Kong’s internal email servers. Our experi-
58 MARCH • APRIL 2006 Published by the IEEE Computer Society 1089-7801/06/$20.00 © 2006 IEEE IEEE INTERNET COMPUTING
Computer Security