COMPA: Detecting Compromised Accounts on Social Networks

Author(s): Manuel Egele, Gianluca Stringhini, Christopher Kruegel, and Giovanni Vigna

Download: Paper (PDF)

Date: 23 Apr 2013

Document Type: Presentations

Additional Documents: Slides

Associated Event: NDSS Symposium 2013

Abstract:

COMPA detects compromised, yet legitimate social network accounts by building behavioral profiles for inpidual accounts and matching new messages against the extracted behavior. Should multiple similar messages violate their users’ behavioral profiles, COMPA declares these users as compromised. COMPA demonstrated high precision when evaluated on large-scale real-world datasets (1.4B Twitter, 106M Facebook messages).