Φ-Hiding ⇒ Lossy TDF (Kiltz–O’Neill–Smith)
Statement
If the Φ-hiding assumption holds, RSA is a lossy trapdoor permutation — KOS10. In the injective mode and is the RSA permutation with trapdoor ; in the lossy mode and the map is at least -to- on ; the two modes are indistinguishable exactly by Φ-hiding. RSA acts on rather than on bit strings — AKPS19. In the injective mode it permutes all of , so its restriction to the inputs in is an injective function on whose lossy mode has lossiness at least — standard. When , for instance for close to , this is a lossy trapdoor function; for smaller , the product of independent instances multiplies the lossiness by and is one — standard.
Sketch
For prime , the map has images on , so at most images on ; the restricted domain has elements, which gives the lossiness — standard.
Notes
- In the lossy mode is at most -to- on , so for the lossiness is below , the residual leakage exceeds half the input length, and the function is not highly lossy — standard.
- A domain-extension technique of FGKRS10 makes RSA a lossy trapdoor permutation on a fixed domain, with considerably reduced lossiness; AKPS19 give Φ-hiding-based lossy trapdoor permutations with improved lossiness and the first all-but-one lossy trapdoor permutation from Φ-hiding — AKPS19.
- KOS10 use the lossiness of RSA to prove RSA-OAEP IND-CPA-secure in the standard model, with the first round’s hash function -wise independent — KOS10.