arXiv subjects

David Freeman

Publications and source records attributed to David Freeman.

12 recordsLinked to original sources

arXiv2025

Predictive Response Optimization: Using Reinforcement Learning to Fight Online Social Network Abuse

Detecting phishing, spam, fake accounts, data scraping, and other malicious activity in online social networks (OSNs) is a problem that has been studied for well over a decade, with a number of important results. Nearly all existing works on abuse detection have as their goal producing the best possible binary classifier; i.e., one that labels unseen examples as "benign" or "malicious" with high precision and recall. However, no prior published work considers what comes next: what does the service actually do after it detects abuse? In this paper, we argue that detection as described in previous work is not the goal of those who are fighting OSN abuse. Rather, we believe the goal to be selecting actions (e.g., ban the user, block the request, show a CAPTCHA, or "collect more evidence") that optimize a tradeoff between harm caused by abuse and impact on benign users. With this framing, we see that enlarging the set of possible actions allows us to move the Pareto frontier in a way that is unattainable by simply tuning the threshold of a binary classifier. To demonstrate the potential of our approach, we present Predictive Response Optimization (PRO), a system based on reinforcement learning that utilizes available contextual information to predict future abuse and user-experience metrics conditioned on each possible action, and select actions that optimize a multi-dimensional tradeoff between abuse/harm and impact on user experience. We deployed versions of PRO targeted at stopping automated activity on Instagram and Facebook. In both cases our experiments showed that PRO outperforms a baseline classification system, reducing abuse volume by 59% and 4.5% (respectively) with no negative impact to users. We also present several case studies that demonstrate how PRO can quickly and automatically adapt to changes in business constraints, system behavior, and/or adversarial tactics.

cs.LG↗

arXiv2022

"Real Attackers Don't Compute Gradients": Bridging the Gap Between Adversarial ML Research and Practice

Recent years have seen a proliferation of research on adversarial machine learning. Numerous papers demonstrate powerful algorithmic attacks against a wide variety of machine learning (ML) models, and numerous other papers propose defenses that can withstand most attacks. However, abundant real-world evidence suggests that actual attackers use simple tactics to subvert ML-driven systems, and as a result security practitioners have not prioritized adversarial ML defenses. Motivated by the apparent gap between researchers and practitioners, this position paper aims to bridge the two domains. We first present three real-world case studies from which we can glean practical insights unknown or neglected in research. Next we analyze all adversarial ML papers recently published in top security conferences, highlighting positive trends and blind spots. Finally, we state positions on precise and cost-driven threat modeling, collaboration between industry and academia, and reproducible research. We believe that our positions, if adopted, will increase the real-world impact of future endeavours in adversarial ML, bringing both researchers and practitioners closer to their shared goal of improving the security of ML systems.

cs.CR↗

arXiv2022

Propagation effects in high-harmonic generation from dielectric thin films

Theoretical investigation is conducted of high-order harmonic generation (HHG) in silicon thin films to elucidate the effect of light propagation in reflected and transmitted waves. The first-principles simulations are performed of the process in which an intense pulsed light irradiates silicon thin films up to 3 $\mu$m thickness. Our simulations are carried within the time-dependent density functional theory (TDDFT) with the account of coupled dynamics of the electromagnetic fields and the electronic motion. It was found that the intensity of transmission HHG gradually decreases with the thickness, while the reflection HHG becomes constant from a certain thickness. Detailed analyses show that transmission HHG have two origins: the HHG generated near the front edge and propagating to the back surface, and that generated near the back edge and emitted directly. The dominating mechanism of the transmission HHG is found to depend on the thickness of the thin film and the frequency of the HHG. At the film thickness of 1 $\mu$m, the transmission HHG with the frequency below 20 eV is generated near the back edge, while that with the frequency above 20 eV is generated near the front edge and propagates from there to the back surface.

cond-mat.mtrl-sci↗

arXiv2022

Quantum Mechanics for Closure of Dynamical Systems

We propose a scheme for data-driven parameterization of unresolved dimensions of dynamical systems based on the mathematical framework of quantum mechanics and Koopman operator theory. Given a system in which some components of the state are unknown, this method involves defining a surrogate system in a time-dependent quantum state which determines the fluxes from the unresolved degrees of freedom at each timestep. The quantum state is a density operator on a finite-dimensional Hilbert space of classical observables and evolves over time under an action induced by the Koopman operator. The quantum state also updates with new values of the resolved variables according to a quantum Bayes' law, implemented via an operator-valued feature map. Kernel methods are utilized to learn data-driven basis functions and represent quantum states, observables, and evolution operators as matrices. The resulting computational schemes are automatically positivity-preserving, aiding in the physical consistency of the parameterized system. We analyze the results of two different modalities of this methodology applied to the Lorenz 63 and Lorenz 96 multiscale systems, and show how this approach preserves important statistical and qualitative properties of the underlying chaotic systems.

math.DS↗

arXiv2022

Data Assimilation in Operator Algebras

We develop an algebraic framework for sequential data assimilation of partially observed dynamical systems. In this framework, Bayesian data assimilation is embedded in a non-abelian operator algebra, which provides a representation of observables by multiplication operators and probability densities by density operators (quantum states). In the algebraic approach, the forecast step of data assimilation is represented by a quantum operation induced by the Koopman operator of the dynamical system. Moreover, the analysis step is described by a quantum effect, which generalizes the Bayesian observational update rule. Projecting this formulation to finite-dimensional matrix algebras leads to new computational data assimilation schemes that are (i) automatically positivity-preserving; and (ii) amenable to consistent data-driven approximation using kernel methods for machine learning. Moreover, these methods are natural candidates for implementation on quantum computers. Applications to data assimilation of the Lorenz 96 multiscale system and the El Nino Southern Oscillation in a climate model show promising results in terms of forecast skill and uncertainty quantification.

math.ST↗

arXiv2021

High order harmonic generation in semiconductors driven at near- and mid-IR wavelengths

We study high order harmonics generation (HHG) in crystalline silicon and diamond subjected to near and mid-infrared laser pulses. We employ time-dependent density functional theory and solve the time-dependent Kohn-Sham equation in the single-cell geometry. We demonstrate that clear and clean HHG spectra can be generated with careful selection of the pulse duration. In addition, we simulate dephasing effects in a large silicon super-cell through displacement of atomic positions prepared by a molecular dynamics simulation. We compare our results with the previous calculations by Floss et al. [arXiv:1705.10707] [Phys. Rev. A 97, 011401(R) (2018)] on Diamond at 800 nm and by Tancogne-Dejean et al. [arXiv:1609.09298] [Phys. Rev. Lett. 118, 087403 (2017)] on Si at 3000 nm.

cond-mat.mtrl-sci↗

arXiv2019

Correlation enhancement of high-order harmonics generation in Xe

We consider the process of high-order harmonics generation (HHG) in the xenon atom enhanced by the inter-shell correlation between the valence 5p and inner 4d shells. We derive the HHG spectrum from a numerical solution of the one-electron time-dependent Schr\"odinger equation multiplied by the enhancement factor taken as the photoionization cross-sections ratio calculated with and without the inter-shell correlation. Such a simplified approach is adequate to describe the experimental HHG spectrum reported by Shiner et al [Nat. Phys. 7, 464 (2011)] generated by a single color IR laser. Similarly, we find good agreement when applied to the two-color driven HHG spectra reported by Facciala et al [Phys. Rev. Lett. 117, 093902 (2016)].

physics.atom-ph↗

arXiv2018

Toward a quasi-M\"obius characterization of Invertible Homogeneous Metric Spaces

We study locally compact metric spaces that enjoy various forms of homogeneity with respect to M\"obius self-homeomorphisms. We investigate connections between such homogeneity and the combination of isometric homogeneity with invertibility. In particular, we provide a new characterization of snowflakes of boundaries of rank-one symmetric spaces of non-compact type among locally compact and connected metric spaces. Furthermore, we investigate the metric implications of homogeneity with respect to uniformly strongly quasi-M\"obius self-homeomorphisms, connecting such homogeneity with the combination of uniform bi-Lipschitz homogeneity and quasi-invertibility. In this context we characterize spaces containing a cut point and provide several metric properties of spaces containing no cut points. These results are motivated by a desire to characterize the snowflakes of boundaries of rank-one symmetric spaces up to bi-Lipschitz equivalence.

math.MG↗

arXiv2008

Abelian Varieties with Prescribed Embedding Degree

We present an algorithm that, on input of a CM-field $K$, an integer $k\ge1$, and a prime $r \equiv 1 \bmod k$, constructs a $q$-Weil number $\pi \in \O_K$ corresponding to an ordinary, simple abelian variety $A$ over the field $\F$ of $q$ elements that has an $\F$-rational point of order $r$ and embedding degree $k$ with respect to $r$. We then discuss how CM-methods over $K$ can be used to explicitly construct $A$.

math.NT↗

arXiv2007

Computing endomorphism rings of Jacobians of genus 2 curves over finite fields

We present algorithms which, given a genus 2 curve $C$ defined over a finite field and a quartic CM field $K$, determine whether the endomorphism ring of the Jacobian $J$ of $C$ is the full ring of integers in $K$. In particular, we present probabilistic algorithms for computing the field of definition of, and the action of Frobenius on, the subgroups $J[\ell^d]$ for prime powers $\ell^d$. We use these algorithms to create the first implementation of Eisenträger and Lauter's algorithm for computing Igusa class polynomials via the Chinese Remainder Theorem \cite{el}, and we demonstrate the algorithm for a few small examples. We observe that in practice the running time of the CRT algorithm is dominated not by the endomorphism ring computation but rather by the need to compute $p^3$ curves for many small primes $p$.

math.NT↗

arXiv2006

Constructing pairing-friendly elliptic curves with embedding degree 10

We present a general framework for constructing families of elliptic curves of prime order with prescribed embedding degree. We demonstrate this method by constructing curves with embedding degree k = 10, which solves an open problem posed by Boneh, Lynn, and Shacham. We show that our framework incorporates existing constructions for k = 3, 4, 6, and 12, and we give evidence that the method is unlikely to produce infinite families of curves with embedding degree k > 12.

math.NT↗

arXiv2005

Pairing-based identification schemes

We propose four different identification schemes that make use of bilinear pairings, and prove their security under certain computational assumptions. Each of the schemes is more efficient and/or more secure than any known pairing-based identification scheme.

cs.CR↗