arXiv subjects

Shucheng Yu

Publications and source records attributed to Shucheng Yu.

At least 19 recordsLinked to original sources

arXiv2026

Zero-one laws for uniform approximation via Gaussian and Eisenstein integers

We establish two distinct zero-one laws for the uniform Diophantine approximation of complex numbers by quotients of Gaussian integers and by quotients of Eisenstein integers. Using tools from homogeneous dynamics, we study this problem by reducing to a shrinking target problem on certain homogeneous spaces of $\mathrm{SL}_2(\mathbb{C})$. The main novel ingredients include measure estimates on a certain family of neighborhoods of the corresponding critical loci, as well as new disjointness statements to control the short-range mixing contribution. Due to the different nature of the critical loci in the Gaussian and Eisenstein cases, these measure estimates are obtained by rather different arguments.

math.DS↗

arXiv2026

A zero-one law for improvements to Dirichlet's theorem in arbitrary dimension

Let $\psi$ be a continuous decreasing function defined on all large positive real numbers. We say that a real $m\times n$ matrix $A$ is $\psi$-Dirichlet if for every sufficiently large real number $t$ one can find $\mathbf{p} \in \mathbb{Z}^m$, $\mathbf{q} \in \mathbb{Z}^n\setminus\{\mathbf{0}\}$ satisfying $\|A\mathbf{q}-\mathbf{p}\|^m< \psi(t)$ and $\|\mathbf{q}\|^n<t$. By removing a technical condition from a partial zero-one law proved by Kleinbock-Str\"ombergsson-Yu, we prove a zero-one law for the Lebesgue measure of the set of $\psi$-Dirichlet matrices provided that $\psi(t)<1/t$ and $t\psi(t)$ is increasing. In fact, we prove the zero-one law in a more general situation with the monotonicity assumption on $t\psi(t)$ replaced by a weaker condition. Our proof follows the dynamical approach of Kleinbock-Str\"ombergsson-Yu in reducing the question to a shrinking target problem in the space of lattices. The key new ingredient is a family of carefully chosen subsets of the shrinking targets studied by Kleinbock-Str\"ombergsson-Yu, together with a short-range mixing estimate for the associated hitting events. Our method also works for the analogous weighted problem where the relevant supremum norms are replaced by certain weighted quasi-norms.

math.NT↗

arXiv2025

Extreme events and impact statistics for unipotent actions on the space of lattices

This paper extends a recent extreme value law for horocycle flows on the space of two-dimensional lattices, due to Kirsebom and Mallahi-Karai, to the simplest examples of rank-$k$ unipotent actions on the space of $n$-dimensional lattices. We analyse the problem in terms of the hitting time and impact statistics for the unipotent action with respect to a shrinking surface of section, following the strategy of Pollicott and the first named author in the case of hyperbolic surfaces. If $k=n-1$, the limit law is given by directional statistics of Euclidean lattices, whilst for $k<n-1$ we observe new distributions for which we derive precise tail asymptotics.

math.DS↗

arXiv2024

Asymptotic estimates of large gaps between directions in certain planar quasicrystals

For quasicrystals of cut-and-project type in $\mathbb{R}^d$, it was proved by Marklof and Str\"ombergsson that the limit local statistical properties of the directions to the points in the set are described by certain $\operatorname{SL}_d(\mathbb{R})$-invariant point processes. In the present paper we make a detailed study of the tail asymptotics of the limiting gap statistics of the directions, for certain specific classes of planar quasicrystals.

math.NT↗

arXiv2024

GridSE: Towards Practical Secure Geographic Search via Prefix Symmetric Searchable Encryption (Full Version)

The proliferation of location-based services and applications has brought significant attention to data and location privacy. While general secure computation and privacy-enhancing techniques can partially address this problem, one outstanding challenge is to provide near latency-free search and compatibility with mainstream geographic search techniques, especially the Discrete Global Grid Systems (DGGS). This paper proposes a new construction, namely GridSE, for efficient and DGGS-compatible Secure Geographic Search (SGS) with both backward and forward privacy. We first formulate the notion of a semantic-secure primitive called \textit{symmetric prefix predicate encryption} (SP$^2$E), for predicting whether or not a keyword contains a given prefix, and provide a construction. Then we extend SP$^2$E for dynamic \textit{prefix symmetric searchable encryption} (pSSE), namely GridSE, which supports both backward and forward privacy. GridSE only uses lightweight primitives including cryptographic hash and XOR operations and is extremely efficient. Furthermore, we provide a generic pSSE framework that enables prefix search for traditional dynamic SSE that supports only full keyword search. Experimental results over real-world geographic databases of sizes (by the number of entries) from $10^3$ to $10^7$ and mainstream DGGS techniques show that GridSE achieves a speedup of $150\times$ - $5000\times$ on search latency and a saving of $99\%$ on communication overhead as compared to the state-of-the-art. Interestingly, even compared to plaintext search, GridSE introduces only $1.4\times$ extra computational cost and $0.9\times$ additional communication cost. Source code of our scheme is available at https://github.com/rykieguo1771/GridSE-RAM.

cs.CR↗

arXiv2022

Over-the-Air Federated Learning with Enhanced Privacy

Federated learning (FL) has emerged as a promising learning paradigm in which only local model parameters (gradients) are shared. Private user data never leaves the local devices thus preserving data privacy. However, recent research has shown that even when local data is never shared by a user, exchanging model parameters without protection can also leak private information. Moreover, in wireless systems, the frequent transmission of model parameters can cause tremendous bandwidth consumption and network congestion when the model is large. To address this problem, we propose a new FL framework with efficient over-the-air parameter aggregation and strong privacy protection of both user data and models. We achieve this by introducing pairwise cancellable random artificial noises (PCR-ANs) on end devices. As compared to existing over-the-air computation (AirComp) based FL schemes, our design provides stronger privacy protection. We analytically show the secrecy capacity and the convergence rate of the proposed wireless FL aggregation algorithm.

cs.CR↗

arXiv2022

Second moment of the light-cone Siegel transform and applications

We study the light-cone Siegel transform, transforming functions on the light cone of a rational indefinite quadratic form $Q$ to a function on the homogenous space $\text{SO}^+_Q(\mathbb{Z})\backslash \text{SO}^+_Q(\mathbb{R})$. In particular, we prove a second moment formula for this transform for forms of signature $(n+1,1)$, and show how it can be used for various applications for counting integer points on the light cone. In particular, we prove some new results on intrinsic Diophantine approximations on ellipsoids as well as on the distribution of values of random linear and quadratic forms on the light cone.

math.NT↗

arXiv2022

Fourier expansion of light-cone Eisenstein series

In this work we give an explicit formula for the Fourier coefficients of Eisenstein series corresponding to certain arithmetic lattices acting on hyperbolic n+1-space. As a consequence we obtain results on location of all poles of these Eisenstein series as well as their supremum norms. We use this information to get new results on counting rational points on spheres.

math.NT↗

arXiv2021

A measure estimate in geometry of numbers and improvements to Dirichlet's theorem

Let $\psi$ be a continuous decreasing function defined on all large positive real numbers. We say that a real $m\times n$ matrix $A$ is $\psi$-Dirichlet if for every sufficiently large real number $t$ one can find $\boldsymbol{p} \in \mathbb{Z}^m$, $\boldsymbol{q} \in \mathbb{Z}^n\smallsetminus\{\boldsymbol{0}\}$ satisfying $\|A\boldsymbol{q}-\boldsymbol{p}\|^m< \psi({t})$ and $\|\boldsymbol{q}\|^n<{t}$. This property was introduced by Kleinbock and Wadleigh in 2018, generalizing the property of $A$ being Dirichlet improvable which dates back to Davenport and Schmidt (1969). In the present paper, we give sufficient conditions on $\psi$ to ensure that the set of $\psi$-Dirichlet matrices has zero or full Lebesgue measure. Our proof is dynamical and relies on the effective equidistribution and doubly mixing of certain expanding horospheres in the space of lattices. Another main ingredient of our proof is an asymptotic measure estimate for certain compact neighborhoods of the critical locus (with respect to the supremum norm) in the space of lattices. Our method also works for the analogous weighted problem where the relevant supremum norms are replaced by certain weighted quasi-norms.

math.NT↗

arXiv2021

BO-DBA: Query-Efficient Decision-Based Adversarial Attacks via Bayesian Optimization

Decision-based attacks (DBA), wherein attackers perturb inputs to spoof learning algorithms by observing solely the output labels, are a type of severe adversarial attacks against Deep Neural Networks (DNNs) requiring minimal knowledge of attackers. State-of-the-art DBA attacks relying on zeroth-order gradient estimation require an excessive number of queries. Recently, Bayesian optimization (BO) has shown promising in reducing the number of queries in score-based attacks (SBA), in which attackers need to observe real-valued probability scores as outputs. However, extending BO to the setting of DBA is nontrivial because in DBA only output labels instead of real-valued scores, as needed by BO, are available to attackers. In this paper, we close this gap by proposing an efficient DBA attack, namely BO-DBA. Different from existing approaches, BO-DBA generates adversarial examples by searching so-called \emph{directions of perturbations}. It then formulates the problem as a BO problem that minimizes the real-valued distortion of perturbations. With the optimized perturbation generation process, BO-DBA converges much faster than the state-of-the-art DBA techniques. Experimental results on pre-trained ImageNet classifiers show that BO-DBA converges within 200 queries while the state-of-the-art DBA techniques need over 15,000 queries to achieve the same level of perturbation distortion. BO-DBA also shows similar attack success rates even as compared to BO-based SBA attacks but with less distortion.

cs.LG↗

arXiv2021

SAFELearning: Enable Backdoor Detectability In Federated Learning With Secure Aggregation

For model privacy, local model parameters in federated learning shall be obfuscated before sent to the remote aggregator. This technique is referred to as \emph{secure aggregation}. However, secure aggregation makes model poisoning attacks such backdooring more convenient considering that existing anomaly detection methods mostly require access to plaintext local models. This paper proposes SAFELearning which supports backdoor detection for secure aggregation. We achieve this through two new primitives - \emph{oblivious random grouping (ORG)} and \emph{partial parameter disclosure (PPD)}. ORG partitions participants into one-time random subgroups with group configurations oblivious to participants; PPD allows secure partial disclosure of aggregated subgroup models for anomaly detection without leaking individual model privacy. SAFELearning can significantly reduce backdoor model accuracy without jeopardizing the main task accuracy under common backdoor strategies. Extensive experiments show SAFELearning is robust against malicious and faulty participants, whilst being more efficient than the state-of-art secure aggregation protocol in terms of both communication and computation costs.

cs.CR↗

arXiv2020

Quantitative Diophantine approximation with congruence conditions

In this short paper we prove a quantitative version of the Khintchine-Groshev Theorem with congruence conditions. Our argument relies on a classical argument of Schmidt on counting generic lattice points, which in turn relies on a certain variance bound on the space of lattices.

math.NT↗

arXiv2020

Translates of rational points along expanding closed horocycles on the modular surface

We study the limiting distribution of the rational points under a horizontal translation along a sequence of expanding closed horocycles on the modular surface. Using spectral methods we confirm equidistribution of these sample points for any translate when the sequence of horocycles expands within a certain polynomial range. We show that the equidistribution fails for generic translates and a slightly faster expanding rate. We also prove both equidistribution and non-equidistribution results by obtaining explicit limiting measures while allowing the sequence of horocycles to expand arbitrarily fast. Similar results are also obtained for translates of primitive rational points.

math.DS↗

arXiv2020

Effective density for inhomogeneous quadratic forms II: fixed forms and generic shifts

We establish effective versions of Oppenheim's conjecture for generic inhomogeneous quadratic forms. We prove such results for fixed quadratic forms and generic shifts. Our results complement our companion paper where we considered generic forms and fixed shifts. In this paper, we use ergodic theorems and in particular we establish a strong spectral gap with effective bounds for some representations of orthogonal groups which do not possess Kazhdan's property (T).

math.NT↗

arXiv2019

Effective density for inhomogeneous quadratic forms I: generic forms and fixed shifts

We establish effective versions of Oppenheim's conjecture for generic inhomogeneous quadratic forms. We prove such results for fixed shift vectors and generic quadratic forms. When the shift is rational we prove a counting result which implies the optimal density for values of generic inhomogeneous forms. We also obtain a similar density result for fixed irrational shifts satisfying an explicit Diophantine condition. The main technical tool is a formula for the second moment of Siegel transforms on certain congruence quotients of $\operatorname{SL}_n(\mathbb{R})$ which we believe to be of independent interest. In a sequel, we use different techniques to treat the companion problem concerning generic shifts and fixed quadratic forms.

math.NT↗

arXiv2019

A dynamical Borel-Cantelli lemma via improvements to Dirichlet's theorem

Let $X\cong \operatorname{SL}_2(\mathbb R)/\operatorname{SL}_2(\mathbb Z)$ be the space of unimodular lattices in $\mathbb R^2$, and for any $r\ge 0$ denote by $K_r\subset X$ the set of lattices such that all its nonzero vectors have supremum norm at least $e^{-r}$. These are compact nested subset{s} of $X$, with $K_0 = {\bigcap}_{r}K_r$ being the union of two closed horocycles. We use an explicit second moment formula for the Siegel transform of the indicator functions of squares in $\mathbb R^2$ centered at the origin to derive an asymptotic formula for the volume of sets $K_r$ as $r\to 0$. Combined with a zero-one law for the set of the $\psi$-Dirichlet numbers established by Kleinbock and Wadleigh, this gives a new dynamical Borel-Cantelli lemma for the geodesic flow on $X$ with respect to the family of shrinking targets $\{K_r\}$.

math.DS↗

arXiv2019

LEP-CNN: A Lightweight Edge Device Assisted Privacy-preserving CNN Inference Solution for IoT

Supporting convolutional neural network (CNN) inference on resource-constrained IoT devices in a timely manner has been an outstanding challenge for emerging smart systems. To mitigate the burden on IoT devices, the prevailing solution is to offload the CNN inference task, which is usually composed of billions of operations, to public cloud. However, the "offloading-to-cloud" solution may cause privacy breach while moving sensitive data to cloud. For privacy protection, the research community has resorted to advanced cryptographic primitives and approximation techniques to support CNN inference on encrypted data. Consequently, these attempts cause impractical computational overhead on IoT devices and degrade the performance of CNNs. Moreover, relying on the remote cloud can cause additional network latency and even make the system dysfunction when network connection is off. We proposes an extremely lightweight edge device assisted private CNN inference solution for IoT devices, namely LEP-CNN. The main design of LEP-CNN is based on a novel online/offline encryption scheme. The decryption of LEP-CNN is pre-computed offline via utilizing the linear property of the most time-consuming operations of CNNs. As a result, LEP-CNN allows IoT devices to securely offload over 99% CNN operations, and edge devices to execute CNN inference on encrypted data as efficient as on plaintext. LEP-CNN also provides an integrity check option to help IoT devices detect error results with a successful rate over 99%. Experiments on AlexNet show that LEP-CNN can speed up the CNN inference for more than 35 times for resource constrained IoT devices. A homomorphic encryption based AlexNet using CryptoNets is implemented to compare with LEP-CNN to demonstrate that LEP-CNN has a better performance than homomorphic encryption based privacy preserving neural networks under time-sensitive scenarios.

cs.CR↗

arXiv2018

Values of random polynomials in shrinking targets

Relying on the classical second moment formula of Rogers we give an effective asymptotic formula for the number of integer vectors $v$ in a ball of radius $t$, with value $Q(v)$ in a shrinking interval of size $t^{-\kappa}$, that is valid for almost all indefinite quadratic forms in $n$ variables for any $\kappa<n-2$. This implies in particular, the existence of such integer solutions establishing the prediction made by Ghosh Gorodnik and Nevo. We also obtain similar results for random polynomials of higher degree.

math.NT↗