Short Integer Solution: Difference between revisions
First steps |
Add ring + module SIS |
||
(8 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
Short Integer Solution (SIS) is an average-case problem, which was introduced in 1996 by Miklós Ajtai.<ref>Ajtai, Miklós. [https://dl.acm.org/doi/pdf/10.1145/237814.237838 Generating hard instances of lattice problems]. ''Proceedings of the twenty-eighth annual ACM symposium on Theory of computing''. 1996.</ref> He introduced a family of one-way functions based on SIS and showed that SIS is hard to solve on average if a version of the [[wikipedia:Lattice_problem#Shortest_vector_problem_(SVP)|shortest vector problem]] is hard in a worst-case scenario. | Short Integer Solution (SIS) is an average-case problem, which was introduced in 1996 by Miklós Ajtai.<ref name=":0">Ajtai, Miklós. [https://dl.acm.org/doi/pdf/10.1145/237814.237838 Generating hard instances of lattice problems]. ''Proceedings of the twenty-eighth annual ACM symposium on Theory of computing''. 1996.</ref> He introduced a family of one-way functions based on SIS and showed that SIS is hard to solve on average if a version of the [[wikipedia:Lattice_problem#Shortest_vector_problem_(SVP)|shortest vector problem]] is hard in a worst-case scenario. | ||
== Formal Definition == | == Formal Definition == | ||
=== SIS<sub>n,m,q,β</sub | === SIS<sub>n,m,q,β</sub> === | ||
Let matrix <math>\mathbf{A} \in \mathbb{Z}_q^{n \times m}</math> be chosen uniformly at random. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathbb{Z}^m</math>satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{0} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | Let matrix <math>\mathbf{A} \in \mathbb{Z}_q^{n \times m}</math> be chosen uniformly at random. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathbb{Z}^m</math>satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{0} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | ||
SIS intuitively states that it is hard to find a short vector in the kernel of matrix <math>\mathbf{A}</math>. | |||
A solution to SIS without the condition <math>\lVert \mathbf{s} \rVert \leq \beta</math> can be found using [[wikipedia:Gaussian_elimination|Gaussian elimination]]. Thus, the condition <math>\beta < q</math> is required as otherwise <math>(q, 0,\dots,0) \in \mathbb{Z}^m</math> is a trivial solution. | |||
The inhomogeneous version of SIS introduces a target vector <math>\mathbf{t} \in \mathbb{Z}_q^n</math>, which is chosen uniformly at random. The probability of ending up in the homogeneous case with <math>\mathbf{t} = \mathbf{0}</math> happens with probability <math>q^{-n}</math>, which allows removing the condition of <math>\mathbf{s}</math> being non-zero. | === Ring-SIS<sub>m,d,q,β</sub> === | ||
Let matrix <math>\mathbf{a} \in \mathcal{R}_q^{m}</math> be chosen uniformly at random. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathcal{R}^m</math>satisfying <math>\mathbf{a}^T \cdot \mathbf{s} = \mathbf{0} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | |||
Let <math>\mathcal{R}</math> denote a polynomial ring <math>\mathbb{Z}_q[X] / (f(X))</math>. The function <math>f(X)</math> is usually chosen as <math>(X^d + 1)</math> with special interest in <math>d</math> being a power of 2.<ref name=":4">Lyubashevsky, Vadim, Chris Peikert, and Oded Regev. [https://www.iacr.org/archive/eurocrypt2010/66320288/66320288.pdf On ideal lattices and learning with errors over rings]. ''Annual international conference on the theory and applications of cryptographic techniques''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2010.</ref> However, the Ring SIS (R-SIS) problem has also been studied for other choices such as <math>f(X) = (X^d - 1)</math>.<ref name=":5">Micciancio, Daniele. [https://doi.org/10.1109/SFCS.2002.1181960 Generalized compact knapsacks, cyclic lattices, and efficient one-way functions from worst-case complexity assumptions]. ''The 43rd Annual IEEE Symposium on Foundations of Computer Science, 2002. Proceedings.''. IEEE, 2002.</ref><ref name=":6">Peikert, Chris, and Alon Rosen. [https://iacr.org/archive/tcc2006/38760145/38760145.pdf Efficient collision-resistant hashing from worst-case assumptions on cyclic lattices]. ''Theory of Cryptography Conference''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2006.</ref><ref name=":7">Lyubashevsky, Vadim, and Daniele Micciancio. [https://doi.org/10.1007/11787006_13 Generalized compact knapsacks are collision resistant]. ''International Colloquium on Automata, Languages, and Programming''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2006.</ref> | |||
=== Module-SIS<sub>n,m,d,q,β</sub> === | |||
Let matrix <math>\mathbf{A} \in \mathcal{R}_q^{n \times m}</math> be chosen uniformly at random. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathcal{R}^m</math>satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{0} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | |||
While M-SIS is a less compact variant of SIS than R-SIS, the M-SIS problem is asymptotically at least as hard as R-SIS and therefore gives a tighter bound on the hardness assumption of SIS. This makes assuming the hardness of M-SIS a safer, but less efficient underlying assumption when compared to R-SIS.<ref name=":8">Langlois, Adeline, and Damien Stehlé. [https://eprint.iacr.org/2012/090.pdf Worst-case to average-case reductions for module lattices]. ''Designs, Codes and Cryptography'' 75.3 (2015): 565-599.</ref> | |||
== Versions of SIS == | |||
=== Inhomogeneous SIS<sub>n,m,q,β</sub> === | |||
Let matrix <math>\mathbf{A} \in \mathbb{Z}_q^{n \times m}</math>and target vector <math>\mathbf{t} \in \mathbb{Z}_q^n</math> be chosen uniformly at random. An adversary is asked to find a short vector <math>\mathbf{s} \in \mathbb{Z}^m</math> satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{t} \bmod q \land \lVert \mathbf{s} \rVert \leq \beta</math>. | |||
The inhomogeneous version of SIS (ISIS) introduces a target vector <math>\mathbf{t} \in \mathbb{Z}_q^n</math>, which is chosen uniformly at random. The probability of ending up in the homogeneous case with <math>\mathbf{t} = \mathbf{0}</math> happens with probability <math>q^{-n}</math>, which allows removing the condition of <math>\mathbf{s}</math> being non-zero. | |||
ISIS is as hard as SIS. A SIS instance can be reduced to ISIS using the last column of <math>\mathbf{A}</math> as target vector for ISIS. Any solution <math>\mathbf{s} \in \mathbb{Z}^{m-1}</math> to the ISIS instance with challenge matrix <math>\mathbf{A}_{[1:m-1]}</math> and target vector <math>\mathbf{a}_m</math> yields a valid SIS solution <math>(\mathbf{s}^T, -1) \in \mathbb{Z}^m</math> of slightly larger norm. The reduction from ISIS to SIS requires index guessing a non-zero entry in the SIS solution and embedding the target vector at this position in the challenge matrix <math>\mathbf{A}</math>. | ISIS is as hard as SIS. A SIS instance can be reduced to ISIS using the last column of <math>\mathbf{A}</math> as target vector for ISIS. Any solution <math>\mathbf{s} \in \mathbb{Z}^{m-1}</math> to the ISIS instance with challenge matrix <math>\mathbf{A}_{[1:m-1]}</math> and target vector <math>\mathbf{a}_m</math> yields a valid SIS solution <math>(\mathbf{s}^T, -1) \in \mathbb{Z}^m</math> of slightly larger norm. The reduction from ISIS to SIS requires index guessing a non-zero entry in the SIS solution and embedding the target vector at this position in the challenge matrix <math>\mathbf{A}</math>. | ||
=== | === Normal Form SIS<sub>n,m,q,β</sub> === | ||
Let matrix <math>\bar{\mathbf{A}} \in \mathbb{Z}_q^{n \times m-n}</math> be chosen uniformly at random and define <math>\mathbf{A} = \begin{bmatrix} \mathbf{I}_n &\bar{\mathbf{A}} \end{bmatrix}</math>. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathbb{Z}^m</math>satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{t} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | Let matrix <math>\bar{\mathbf{A}} \in \mathbb{Z}_q^{n \times m-n}</math> be chosen uniformly at random and define <math>\mathbf{A} = \begin{bmatrix} \mathbf{I}_n &\bar{\mathbf{A}} \end{bmatrix}</math>. An adversary is asked to find a short non-zero vector <math>\mathbf{s} \in \mathbb{Z}^m</math>satisfying <math>\mathbf{A} \cdot \mathbf{s} = \mathbf{t} \bmod q \land 0 < \lVert \mathbf{s} \rVert \leq \beta</math>. | ||
Normal Form SIS (NFSIS) is related to the Hermite normal form of a uniformly random matrix <math>\mathbf{A} \in \mathbb{Z}_q^{n \times m}</math>. The normal form version of SIS is often used to reduce public key sizes by size <math>n</math> as the static part of the matrix, the identity matrix <math>\mathbf{I}_n</math>, can be omitted for data transmission. | |||
A SIS instance can be reduced to a NFSIS instance if the first <math>n | |||
</math> columns of its challenge matrix <math>\mathbf{A}</math> are invertible over <math>\mathbb{Z}_q</math>. Assuming this is the case, denote the first <math>n</math> columns of <math>\mathbf{A}</math> by <math>\mathbf{A}_0</math> and define the NFSIS challenge matrix by <math>\mathbf{A}_0^{-1} \cdot \mathbf{A}</math>. Then, any solution of the NFSIS instance is a solution of the SIS instance and vice versa. | |||
=== Further versions === | |||
* SIS with infinity norm<ref name=":1">Lyubashevsky, Vadim. [https://eprint.iacr.org/2011/537.pdf Lattice signatures without trapdoors]. ''Annual International Conference on the Theory and Applications of Cryptographic Techniques''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012.</ref> | |||
* Decision SIS<ref name=":1" /> | |||
== Hardness of SIS == | |||
The initial hardness results of Ajtai<ref name=":0" /> in 1996 were later refined by a series of works<ref>Micciancio, Daniele, and Oded Regev. [https://doi.org/10.1137/S0097539705447360 Worst-case to average-case reductions based on Gaussian measures]. ''SIAM journal on computing'' 37.1 (2007): 267-302.</ref><ref name=":2">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><ref>Micciancio, Daniele, and Chris Peikert. [https://eprint.iacr.org/2013/069.pdf Hardness of SIS and LWE with small parameters]. ''Annual cryptology conference''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2013.</ref>. All results follow are instances of the following theorem. | |||
'''Theorem'''<ref name=":3">Peikert, Chris. [https://eprint.iacr.org/2015/939.pdf A decade of lattice cryptography]. ''Foundations and trends® in theoretical computer science'' 10.4 (2016): 283-424.</ref> For any m = poly(n), any β > 0, and any sufficiently large q ≥ β · poly(n), solving SIS<sub>n,m,q,β</sub> with non-negligible probability is at least as hard as solving the decisional approximate shortest vector problem GapSVP<sub>γ</sub> and the approximate shortest independent vectors problems SIVP<sub>γ</sub> (among others) on arbitrary n-dimensional lattices (i.e., in the worst case) with overwhelming probability, for some γ = β · poly(n). | |||
Similar reductions exist for R-SIS and M-SIS but their hardness relies on the worst-case hardness of SIVP over ideal and module lattices respectively.<ref name=":4" /><ref name=":8" /> R-SIS as defined above is broken for cyclic lattices, i.e. <math>\mathcal{R} = \mathbb{Z}_q[X] / (X^d - 1)</math>, but there are refined versions for cyclic lattices with worst-case to average-case reductions.<ref name=":5" /><ref name=":6" /><ref name=":7" /> | |||
== Constructions based on SIS == | == Constructions based on SIS == | ||
This is a non-exhaustive list of constructions, whose security is or can be based on SIS (or R-SIS and M-SIS). | |||
* Signatures | * One-way function<ref name=":0" /> | ||
* | * Collision-resistant hash function | ||
* Preimage Sampleable Function<ref name=":2" /> | |||
* Signatures<ref name=":1" /><ref name=":2" /><ref>Boyen, Xavier. [https://link.springer.com/chapter/10.1007/978-3-642-13013-7_29 Lattice mixing and vanishing trapdoors: A framework for fully secure short signatures and more]. ''International workshop on public key cryptography''. Berlin, Heidelberg: Springer Berlin Heidelberg, 2010.</ref> | |||
* Commitments<ref>Baum, C., Damgård, I., Lyubashevsky, V., Oechsner, S. and Peikert, C. [https://eprint.iacr.org/2016/997.pdf More efficient commitments from structured lattice assumptions]. ''International conference on security and cryptography for networks''. Cham: Springer International Publishing, 2018.</ref><ref>Lyubashevsky, Vadim, Ngoc Khanh Nguyen, and Maxime Plançon. [https://eprint.iacr.org/2022/284.pdf Lattice-based zero-knowledge proofs and applications: shorter, simpler, and more general]. ''Annual International Cryptology Conference''. Cham: Springer Nature Switzerland, 2022.</ref> | |||
* Vector and Functional Commitments<ref>Peikert, Chris, Zachary Pepin, and Chad Sharp. [https://eprint.iacr.org/2021/1254.pdf Vector and functional commitments from lattices]. ''Theory of Cryptography Conference''. Cham: Springer International Publishing, 2021.</ref> | |||
== Related Assumptions == | == Related Assumptions == | ||
* | * [[Approximate SIS]] | ||
* | * [[Learning with Errors]] | ||
== | == Further reading suggestions == | ||
* | * [https://eprint.iacr.org/2015/939.pdf#page=20 Section 4.1] in ''A decade of lattice cryptography''<ref name=":3" /> can help providing further intuition about SIS | ||
* [https://people.csail.mit.edu/vinodv/CS294/ Lecture Notes] by Vinod Vaikuntanathan | |||
** Lecture 3 on ''Smoothing Parameter and Worst-case to Average-case Reduction for SIS'' | |||
** Lecture 10 on ''Ideal Lattices and Ring Learning with Errors'' | |||
== References == | == References == |
Latest revision as of 11:43, 28 August 2025
Short Integer Solution (SIS) is an average-case problem, which was introduced in 1996 by Miklós Ajtai.[1] He introduced a family of one-way functions based on SIS and showed that SIS is hard to solve on average if a version of the shortest vector problem is hard in a worst-case scenario.
Formal Definition
SISn,m,q,β
Let matrix be chosen uniformly at random. An adversary is asked to find a short non-zero vector satisfying .
SIS intuitively states that it is hard to find a short vector in the kernel of matrix .
A solution to SIS without the condition can be found using Gaussian elimination. Thus, the condition is required as otherwise is a trivial solution.
Ring-SISm,d,q,β
Let matrix be chosen uniformly at random. An adversary is asked to find a short non-zero vector satisfying .
Let denote a polynomial ring . The function is usually chosen as with special interest in being a power of 2.[2] However, the Ring SIS (R-SIS) problem has also been studied for other choices such as .[3][4][5]
Module-SISn,m,d,q,β
Let matrix be chosen uniformly at random. An adversary is asked to find a short non-zero vector satisfying .
While M-SIS is a less compact variant of SIS than R-SIS, the M-SIS problem is asymptotically at least as hard as R-SIS and therefore gives a tighter bound on the hardness assumption of SIS. This makes assuming the hardness of M-SIS a safer, but less efficient underlying assumption when compared to R-SIS.[6]
Versions of SIS
Inhomogeneous SISn,m,q,β
Let matrix and target vector be chosen uniformly at random. An adversary is asked to find a short vector satisfying .
The inhomogeneous version of SIS (ISIS) introduces a target vector , which is chosen uniformly at random. The probability of ending up in the homogeneous case with happens with probability , which allows removing the condition of being non-zero.
ISIS is as hard as SIS. A SIS instance can be reduced to ISIS using the last column of as target vector for ISIS. Any solution to the ISIS instance with challenge matrix and target vector yields a valid SIS solution of slightly larger norm. The reduction from ISIS to SIS requires index guessing a non-zero entry in the SIS solution and embedding the target vector at this position in the challenge matrix .
Normal Form SISn,m,q,β
Let matrix be chosen uniformly at random and define . An adversary is asked to find a short non-zero vector satisfying .
Normal Form SIS (NFSIS) is related to the Hermite normal form of a uniformly random matrix . The normal form version of SIS is often used to reduce public key sizes by size as the static part of the matrix, the identity matrix , can be omitted for data transmission.
A SIS instance can be reduced to a NFSIS instance if the first columns of its challenge matrix are invertible over . Assuming this is the case, denote the first columns of by and define the NFSIS challenge matrix by . Then, any solution of the NFSIS instance is a solution of the SIS instance and vice versa.
Further versions
Hardness of SIS
The initial hardness results of Ajtai[1] in 1996 were later refined by a series of works[8][9][10]. All results follow are instances of the following theorem.
Theorem[11] For any m = poly(n), any β > 0, and any sufficiently large q ≥ β · poly(n), solving SISn,m,q,β with non-negligible probability is at least as hard as solving the decisional approximate shortest vector problem GapSVPγ and the approximate shortest independent vectors problems SIVPγ (among others) on arbitrary n-dimensional lattices (i.e., in the worst case) with overwhelming probability, for some γ = β · poly(n).
Similar reductions exist for R-SIS and M-SIS but their hardness relies on the worst-case hardness of SIVP over ideal and module lattices respectively.[2][6] R-SIS as defined above is broken for cyclic lattices, i.e. , but there are refined versions for cyclic lattices with worst-case to average-case reductions.[3][4][5]
Constructions based on SIS
This is a non-exhaustive list of constructions, whose security is or can be based on SIS (or R-SIS and M-SIS).
- One-way function[1]
- Collision-resistant hash function
- Preimage Sampleable Function[9]
- Signatures[7][9][12]
- Commitments[13][14]
- Vector and Functional Commitments[15]
Related Assumptions
Further reading suggestions
- Section 4.1 in A decade of lattice cryptography[11] can help providing further intuition about SIS
- Lecture Notes by Vinod Vaikuntanathan
- Lecture 3 on Smoothing Parameter and Worst-case to Average-case Reduction for SIS
- Lecture 10 on Ideal Lattices and Ring Learning with Errors
References
- ↑ 1.0 1.1 1.2 Ajtai, Miklós. Generating hard instances of lattice problems. Proceedings of the twenty-eighth annual ACM symposium on Theory of computing. 1996.
- ↑ 2.0 2.1 Lyubashevsky, Vadim, Chris Peikert, and Oded Regev. On ideal lattices and learning with errors over rings. Annual international conference on the theory and applications of cryptographic techniques. Berlin, Heidelberg: Springer Berlin Heidelberg, 2010.
- ↑ 3.0 3.1 Micciancio, Daniele. Generalized compact knapsacks, cyclic lattices, and efficient one-way functions from worst-case complexity assumptions. The 43rd Annual IEEE Symposium on Foundations of Computer Science, 2002. Proceedings.. IEEE, 2002.
- ↑ 4.0 4.1 Peikert, Chris, and Alon Rosen. Efficient collision-resistant hashing from worst-case assumptions on cyclic lattices. Theory of Cryptography Conference. Berlin, Heidelberg: Springer Berlin Heidelberg, 2006.
- ↑ 5.0 5.1 Lyubashevsky, Vadim, and Daniele Micciancio. Generalized compact knapsacks are collision resistant. International Colloquium on Automata, Languages, and Programming. Berlin, Heidelberg: Springer Berlin Heidelberg, 2006.
- ↑ 6.0 6.1 Langlois, Adeline, and Damien Stehlé. Worst-case to average-case reductions for module lattices. Designs, Codes and Cryptography 75.3 (2015): 565-599.
- ↑ 7.0 7.1 7.2 Lyubashevsky, Vadim. Lattice signatures without trapdoors. Annual International Conference on the Theory and Applications of Cryptographic Techniques. Berlin, Heidelberg: Springer Berlin Heidelberg, 2012.
- ↑ Micciancio, Daniele, and Oded Regev. Worst-case to average-case reductions based on Gaussian measures. SIAM journal on computing 37.1 (2007): 267-302.
- ↑ 9.0 9.1 9.2 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.
- ↑ Micciancio, Daniele, and Chris Peikert. Hardness of SIS and LWE with small parameters. Annual cryptology conference. Berlin, Heidelberg: Springer Berlin Heidelberg, 2013.
- ↑ 11.0 11.1 Peikert, Chris. A decade of lattice cryptography. Foundations and trends® in theoretical computer science 10.4 (2016): 283-424.
- ↑ Boyen, Xavier. Lattice mixing and vanishing trapdoors: A framework for fully secure short signatures and more. International workshop on public key cryptography. Berlin, Heidelberg: Springer Berlin Heidelberg, 2010.
- ↑ Baum, C., Damgård, I., Lyubashevsky, V., Oechsner, S. and Peikert, C. More efficient commitments from structured lattice assumptions. International conference on security and cryptography for networks. Cham: Springer International Publishing, 2018.
- ↑ Lyubashevsky, Vadim, Ngoc Khanh Nguyen, and Maxime Plançon. Lattice-based zero-knowledge proofs and applications: shorter, simpler, and more general. Annual International Cryptology Conference. Cham: Springer Nature Switzerland, 2022.
- ↑ Peikert, Chris, Zachary Pepin, and Chad Sharp. Vector and functional commitments from lattices. Theory of Cryptography Conference. Cham: Springer International Publishing, 2021.