%PDF-1.5 % 5 0 obj << /S /GoTo /D (section*.1) >> endobj 8 0 obj (Illuminating the Dark or how to recover what should not be seen in FE-based classifiers) endobj 9 0 obj << /S /GoTo /D (section.1) >> endobj 12 0 obj (1 Scenario and threat model) endobj 13 0 obj << /S /GoTo /D (section.2) >> endobj 16 0 obj (2 Functional encryption) endobj 17 0 obj << /S /GoTo /D (section.3) >> endobj 20 0 obj (3 Information leakage in typical FE deployment scenarios) endobj 21 0 obj << /S /GoTo /D (subsection.3.1) >> endobj 24 0 obj (3.1 Classification use case) endobj 25 0 obj << /S /GoTo /D (subsection.3.2) >> endobj 28 0 obj (3.2 Information leakage) endobj 29 0 obj << /S /GoTo /D (subsection.3.3) >> endobj 32 0 obj (3.3 Leakage estimation) endobj 33 0 obj << /S /GoTo /D (section.4) >> endobj 36 0 obj (4 Experiments) endobj 37 0 obj << /S /GoTo /D (subsection.4.1) >> endobj 40 0 obj (4.1 Datasets) endobj 41 0 obj << /S /GoTo /D (subsection.4.2) >> endobj 44 0 obj (4.2 Neural networks structure) endobj 45 0 obj << /S /GoTo /D (subsubsection.4.2.1) >> endobj 48 0 obj (4.2.1 Prediction NN) endobj 49 0 obj << /S /GoTo /D (subsubsection.4.2.2) >> endobj 52 0 obj (4.2.2 Attack NN) endobj 53 0 obj << /S /GoTo /D (subsection.4.3) >> endobj 56 0 obj (4.3 Results) endobj 57 0 obj << /S /GoTo /D (section.5) >> endobj 60 0 obj (5 Conclusion and future works) endobj 61 0 obj << /S /GoTo /D (appendix.A) >> endobj 64 0 obj (A Information leakage on Census Income dataset) endobj 65 0 obj << /S /GoTo /D [66 0 R /Fit] >> endobj 75 0 obj << /Length 4380 /Filter /FlateDecode >> stream xڕZIwϯ[yjœ(Vb;/"j>I4ʯOmr@a*alw ~tI#$wH2gF GWAaQ&7LojUϼw|u6n Ga:m}ߵ]xe