question: What is the trade-off between privacy and convergence performance when using artificial noise obscuring in federated learning? option 1: Increasing the noise variance improves privacy but degrades convergence option 2: Increasing the noise variance improves privacy and convergence option 3: Increasing the noise variance does not affect privacy or convergence option 4: Increasing the noise variance degrades privacy and convergence option 5: Increasing the noise variance improves privacy but has no effect on convergence 