Do dummies pay off? Limits of dummy traffic protection in anonymous - PowerPoint PPT Presentation

Do dummies pay off? Limits of dummy traffic protection in anonymous communication systems Simon Oya , Carmela Troncoso and Fernando Pérez-González

Introduction • Anonymous channels hide correspondences (input/output). • Dummies are a common protection mechanism. 2

Motivation 3

System Model • Basic pool-mix scenario • Firing condition: triggers the flushing of messages. • Batching strategy: how messages are chosen from the pool. • Sender-based dummies: die at the mix. • Mix-based dummies: generated by the mix. • (This was one communication round ) 4

System Model. Adversary. • Goal: infer • She cannot see the content of the messages… • … but she can see the number of messages sent/received. • She is aware of the system parameters: observes ρ • Mix behavior communication • Dummy policy rounds 5

Least-Squares Estimator • Goal: infer all • Information: ρ rounds • Idea: minimize output error: ρ rounds Pool + Dummies :probabilities • Expanding • Rem. Constraints 6

Performance Analysis. MSE. • Derive an expression for: • Assumptions: • MSE formula:  Receiver Observations Pool Sender 7

Performance Analysis. Interpretation (I). 8

Performance analysis. Interpretation (II). 9

Performance analysis. Interpretation (II). 10

Performance analysis. Interpretation (II). 11

Designing dummy traffic strategies. • We understand the effect of the dummies in the MSE. Dummy Strategy! • Privacy objective PICK YOUR FAVOURITE PRIVACY OBJECTIVE!! • As an example: we pick two objectives. • Assumptions: • Return channel • Budget of dummies 12

Dummy Strategy 1. • Objective : increase the protection of all the relations by a constant factor. 13

Dummy Strategy 1. • Solution : assign dummies proportionally to message rates. • Experimental results: 14

Design Strategy 2. • Objective : provide a minimum protection as large as possible. Receiver-based dummies Unaffected Sender-based dummies Both kinds of dummies 15

Dummy Strategy 2. • Solution : waterfilling-like algorithm. • Experimental Results 16

Conclusions. Methodology. 17

Thanks! simonoya@gts.uvigo.es 18