From: Privacy-preserving semi-parallel logistic regression training with fully homomorphic encryption
Step | Level | Type | n/N | stdv | Security | Sample size |
---|---|---|---|---|---|---|
1 | L0 | LWE | n=612 | 2−15 | ≈128 | 4.8 KB |
L1 | RLWE | N=2048 | 2−53 | ≈128 | 32 KB | |
L2 | RLWE | N=8192 | 2−53 | ≈128 | 128 KB | |
2 | L0 | LWE | n=612 | 2−15 | ≫128 | 4.8 KB |
L1 | RLWE | N=4096 | 2−32 | ≫128 | 32 KB | |
L2 | RLWE | N=4096 | 2−48 | ≫128 | 32 KB | |
L3 | RLWE | N=4096 | 2−64 | ≫128 | 64 KB | |
L4 | RLWE | N=4096 | 2−80 | ≫128 | 64 KB | |
L5 | RLWE | N=4096 | 2−105 | ≈130 | 64 KB | |
3 | L0 | RLWE | N=4096 | 3.2/q with q=232 | ≫128 | 32KB |
L1 | RLWE | N=32768 | 3.2/q with q=2581 | ≈128 | 4.5MB |