Slide 1Summary & ObjectiveMLE (Gaussian)K-Nearest NeighborReason for Mixed ResultsNext StepsVirtual Private NetworkPattern ClassificationJoe MaddenFall 2010ECE/CS/ME 539Summary & Objective•Create multiple training/testing files from .csv file containing Outgoing, Incoming packets •Determine best pattern classification algorithm presented in class•Compare with basic statistical measures•Set a conservative threshold for VPN activity to ensure Quality of ServiceMLE (Gaussian)•Negative Log Likelihood •61% Classified1000 2000 3000 4000 5000 60000100020003000400050006000700080009000Outgoing PacketsIncoming PacketsIncoming vs Outgoing Packets OutgoingIncomingClassified Point AttemptK-Nearest Neighbor0 5 10 15 20 25051015202530354045Classification error rate vs. kk -nearest neighbors% classification error1 1.5 2 2.5 3 3.5 4 4.5 5051015202530354045Classification error rate vs. kk -nearest neighbors% classification errorReason for Mixed Results•Data follows Log-Normal CurveNext Steps•Use a standard normal distribution for generating class vectors•Attempt 3-way cross
View Full Document
Unlocking...