Table 1. Document type classification accuracies in the MPIIDPEye dataset using differentially private eye movement features with majority voting.

	Document type classification accuracies (k-NN|SVM|DT|RF)
Method	ϵ = 0.48	ϵ = 2.4	ϵ = 4.8	ϵ = 24	ϵ = 48
FPA	0.50|0.63|0.82|0.87	0.51|0.63|0.81|0.87	0.5|0.61|0.81|0.87	0.52|0.63|0.82|0.87	0.52|0.64|0.83|0.88
CFPA-32	0.39|0.37|0.45|0.44	0.40|0.38|0.45|0.44	0.40|0.44|0.46|0.44	0.58|0.58|0.55|0.60	0.71|0.69|0.66|0.66
CFPA-64	0.41|0.36|0.45|0.45	0.40|0.37|0.44|0.45	0.40|0.41|0.44|0.45	0.57|0.59|0.55|0.59	0.70|0.70|0.66|0.66
CFPA-128	0.36|0.33|0.45|0.45	0.36|0.33|0.44|0.44	0.37|0.35|0.44|0.45	0.52|0.56|0.52|0.57	0.69|0.68|0.64|0.66
DCFPA-32	0.51|0.37|0.46|0.44	0.51|0.36|0.47|0.42	0.47|0.35|0.47|0.43	0.49|0.37|0.46|0.44	0.48|0.36|0.47|0.45
DCFPA-64	0.61|0.45|0.43|0.41	0.55|0.35|0.43|0.41	0.56|0.41|0.43|0.41	0.60|0.43|0.45|0.42	0.59|0.40|0.44|0.43
DCFPA-128	0.64|0.45|0.46|0.48	0.62|0.42|0.45|0.46	0.69|0.50|0.44|0.46	0.57|0.45|0.45|0.46	0.60|0.42|0.45|0.46