Jump to content

ISISf: Difference between revisions

From Lattice Assumption Zoo
First step
 
First draft
Line 9: Line 9:
* <math>\mathbf{s}_i \leftarrow \mathbf{A}_s^{-1}(f(x_i))</math>
* <math>\mathbf{s}_i \leftarrow \mathbf{A}_s^{-1}(f(x_i))</math>


Given the matrix <math>\mathbf{A}</math>, the function <math>f</math>, and the set of hints <math>\{ (x_i, \mathbf{s}_i) \}_{i \in [k]}</math>, the adversary is asked to find a new tuple <math>(x^*, \mathbf{s}^*) \in [N] \times \mathcal{R}^m</math> satisfying <math>\mathbf{A} \cdot \mathbf{s}^* = f\left(x^*\right) \bmod q \land 0 < \lVert \mathbf{s}^* \rVert \leq \beta \land (x^*, \mathbf{s}^*) \in \left\{ (x_i, \mathbf{s}_i)_{i \in [k]} \right\}.</math>
Given the matrix <math>\mathbf{A}</math>, the function <math>f</math>, and the set of hints <math>\{ (x_i, \mathbf{s}_i) \}_{i \in [k]}</math>, the adversary is asked to find a new tuple <math>(x^*, \mathbf{s}^*) \in [N] \times \mathcal{R}^m</math> satisfying <math>\mathbf{A} \cdot \mathbf{s}^* = f\left(x^*\right) \bmod q \land 0 < \lVert \mathbf{s}^* \rVert \leq \beta \land (x^*, \mathbf{s}^*) \notin \left\{ (x_i, \mathbf{s}_i)_{i \in [k]} \right\}.</math>


'''Intuition.''' ISIS<sub>f</sub> essentially expects the adversary to either successfully solve [[Short Integer Solution#ISISn,m,q,β - Inhomogeneous SIS EQUIVALENT|ISIS]] or compute a preimage of the function <math>f</math>. Thus, the hardness of ISIS<sub>f</sub>  depends on the choice of <math>f</math>. We list few examples for insecure choices of <math>f</math>.
'''Intuition.''' ISIS<sub>f</sub> essentially expects the adversary to either successfully solve [[Short Integer Solution#ISISn,m,q,β - Inhomogeneous SIS EQUIVALENT|ISIS]] or compute a preimage of the function <math>f</math>. Thus, the hardness of ISIS<sub>f</sub>  depends on the choice of <math>f</math>. We list few examples for insecure choices of <math>f</math>.
Line 18: Line 18:


=== Interactive ISIS<sub>f</sub> ===
=== Interactive ISIS<sub>f</sub> ===
TODO
Let <math>(n,m,d,\ell_m, \ell_r,q,\beta_s, \beta_m,s,N)</math> be public parameters, matrices <math>\mathbf{A} \in \mathcal{R}_q^{n \times m}, \mathbf{C} \in \mathcal{R}_q^{n \times (\ell_m + \ell_r)}</math> be chosen uniformly at random, <math>f</math> be a specified function <math>f: [N] \rightarrow \mathcal{R}_q^n</math>, and <math>\mathcal{M} = \emptyset</math>. Given the matrices <math>\mathbf{A}, \mathbf{C}</math>, and the function <math>f</math>, an adversary is able to query an oracle <math>O_\text{pre}</math> adaptively, which on input <math>\mathbf{m} \in \mathcal{R}^{\ell_m}, \mathbf{r} \in \mathcal{R}^{\ell_r}</math>  proceeds as follows.
 
# if <math>\lVert (\mathbf{m}, \mathbf{r}) \rVert > \beta_m</math> then return <math>\perp</math>
# <math>x \leftarrow [N]</math>
# <math>\mathbf{s} \leftarrow \mathbf{A}_s^{-1}\left( f(x) + \mathbf{C} \cdot \begin{bmatrix} \mathbf{m} \\ \mathbf{r} \end{bmatrix} \right)</math>
# <math>\mathcal{M} \leftarrow \mathcal{M} \cup \{ \mathbf{m} \}</math>
# return <math>(x,\mathbf{s})</math>
 
An adversary is asked to find a new tuple <math>(x^*, \mathbf{s}^*, \mathbf{m}^*, \mathbf{r}^*)</math> satisfying <math>\mathbf{m}^* \notin \mathcal{M} \land \mathbf{A} \cdot \mathbf{s}^* = f( x^* ) + \mathbf{C} \cdot \begin{bmatrix} \mathbf{m} \\ \mathbf{r} \end{bmatrix} \land 0 < \lVert \mathbf{s}^* \rVert \leq \beta_s \land \lVert (\mathbf{m}^*, \mathbf{r}^*) \rVert \leq \beta_m.</math>
 
'''Intuition.''' The interactive version expands ISIS<sub>f</sub> by an Ajtai commitment and enables adaptive queries, where an adversary controls the input to the Ajtai commitment. Notice that the adversary is not capable of influencing the choice of <math>x \in [N]</math>, which can be thought of as salt in the context of a signature whereas <math>\mathbf{s} \in \mathcal{R}^m</math> is the signature.


== Hardness of ISIS<sub>f</sub> and its interactive version ==
== Hardness of ISIS<sub>f</sub> and its interactive version ==
TODO
If <math>f</math> is a random oracle then the problem, in the ROM, is at least as hard as SIS.<ref>Gentry, Craig, Chris Peikert, and Vinod Vaikuntanathan. [https://eprint.iacr.org/2007/432.pdf Trapdoors for hard lattices and new cryptographic constructions]. ''Proceedings of the fortieth annual ACM symposium on Theory of computing''. 2008.</ref>
 
Bootle et al.<ref name=":0" /> set <math>f</math> to be <math>f(x) = \mathbf{B} \cdot \text{bin}(x)</math>, where <math>\text{bin}: [N] \rightarrow \mathbb{Z}^{\lceil \log(N) \rceil}</math> outputs the binary encoding of <math>x \in [N]</math>. They call this problem ISIS<sub>bin</sub>. The authors analyse direct lattice reduction as well as exploiting relations on the image space for ISIS<sub>bin</sub>.
 
In Theorem 3.3, Bootle et al.<ref name=":0" /> show that interactive ISIS<sub>f</sub> is at least as hard as ISIS<sub>f</sub>. The reduction uses <math>\mathbf{C} = \mathbf{A} \cdot \mathbf{R}</math> for a uniformly chosen <math>\mathbf{R} \leftarrow \{0,1\}^{m \times (\ell_m + \ell_r)}</math>, rejection sampling, the entropy that <math>x \in [N]</math> and <math>\mathbf{s} \in \mathcal{R}^m</math> are sampled with and introduces a polynomial loss factor depending on the number of allowed queries to <math>O_\text{pre}</math>.


== Constructions based on ISIS<sub>f</sub> ==
== Constructions based on ISIS<sub>f</sub> ==
Bootle et al.<ref name=":0" /> provide a framework to generically build the following constructions from any ISIS<sub>f</sub> instance.
Bootle et al.<ref name=":0" /> provide a framework to generically build the following constructions from any ''interactive'' ISIS<sub>f</sub> instance.


* Signatures<ref name=":0" />
* Signatures<ref name=":0" />

Revision as of 13:34, 25 July 2025

The ISISf assumption was introduced by Bootle, Lyubashevsky, Nguyen, and Sorniotti in 2023.[1] It introduces a function f that removes the requirement of a static target vector and passes additional hints to the adversary.

Formal Definition

ISISf

Let (n,m,d,q,β,k,s,N) be public parameters, matrix 𝐀qn×m be chosen uniformly at random and f be a specified function f:[N]qn. The challenger generates k hints (xi,𝐬i) in the following way.

  • xi[N]
  • 𝐬i𝐀s1(f(xi))

Given the matrix 𝐀, the function f, and the set of hints {(xi,𝐬i)}i[k], the adversary is asked to find a new tuple (x*,𝐬*)[N]×m satisfying 𝐀𝐬*=f(x*)modq0<𝐬*β(x*,𝐬*){(xi,𝐬i)i[k]}.

Intuition. ISISf essentially expects the adversary to either successfully solve ISIS or compute a preimage of the function f. Thus, the hardness of ISISf depends on the choice of f. We list few examples for insecure choices of f.

  • Additively homomorphic functions imply trivial solutions by adding or subtracting two hints.
  • Any efficiently invertible function using public information enables choosing 𝐬*m short and finding a preimage of 𝐀𝐬*.
  • Assume f is a linear function and the domain of f was mapped to N. Then, any hint (xi,𝐬i) can be used to generate a valid ISISf solution (xi,𝐬i).

Interactive ISISf

Let (n,m,d,m,r,q,βs,βm,s,N) be public parameters, matrices 𝐀qn×m,𝐂qn×(m+r) be chosen uniformly at random, f be a specified function f:[N]qn, and =. Given the matrices 𝐀,𝐂, and the function f, an adversary is able to query an oracle Opre adaptively, which on input 𝐦m,𝐫r proceeds as follows.

  1. if (𝐦,𝐫)>βm then return
  2. x[N]
  3. 𝐬𝐀s1(f(x)+𝐂[𝐦𝐫])
  4. {𝐦}
  5. return (x,𝐬)

An adversary is asked to find a new tuple (x*,𝐬*,𝐦*,𝐫*) satisfying 𝐦*𝐀𝐬*=f(x*)+𝐂[𝐦𝐫]0<𝐬*βs(𝐦*,𝐫*)βm.

Intuition. The interactive version expands ISISf by an Ajtai commitment and enables adaptive queries, where an adversary controls the input to the Ajtai commitment. Notice that the adversary is not capable of influencing the choice of x[N], which can be thought of as salt in the context of a signature whereas 𝐬m is the signature.

Hardness of ISISf and its interactive version

If f is a random oracle then the problem, in the ROM, is at least as hard as SIS.[2]

Bootle et al.[1] set f to be f(x)=𝐁bin(x), where bin:[N]log(N) outputs the binary encoding of x[N]. They call this problem ISISbin. The authors analyse direct lattice reduction as well as exploiting relations on the image space for ISISbin.

In Theorem 3.3, Bootle et al.[1] show that interactive ISISf is at least as hard as ISISf. The reduction uses 𝐂=𝐀𝐑 for a uniformly chosen 𝐑{0,1}m×(m+r), rejection sampling, the entropy that x[N] and 𝐬m are sampled with and introduces a polynomial loss factor depending on the number of allowed queries to Opre.

Constructions based on ISISf

Bootle et al.[1] provide a framework to generically build the following constructions from any interactive ISISf instance.

Related Assumptions

  • Generalised ISISf
  • One-More-ISIS
  • Randomised One-More-ISIS

References

  1. 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 Bootle, Jonathan, et al. A framework for practical anonymous credentials from lattices. Annual International Cryptology Conference. Cham: Springer Nature Switzerland, 2023.
  2. Gentry, Craig, Chris Peikert, and Vinod Vaikuntanathan. Trapdoors for hard lattices and new cryptographic constructions. Proceedings of the fortieth annual ACM symposium on Theory of computing. 2008.
  3. 3.0 3.1 Lyubashevsky, Vadim, Gregor Seiler, and Patrick Steuer. The LaZer library: Lattice-based zero knowledge and succinct proofs for quantum-safe privacy. Proceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security. 2024.