Search for: All records

Quantum key distribution (QKD) has established itself as a groundbreaking technology, showcasing inherent security features that are fundamentally proven. Qubit-based QKD protocols that rely on binary encoding encounter an inherent constraint related to the secret key capacity. This limitation restricts the maximum secret key capacity to one bit per photon. On the other hand, qudit-based QKD protocols have their advantages in scenarios where photons are scarce and noise is present, as they enable the transmission of more than one secret bit per photon. While proof-of-principle entangled-based qudit QKD systems have been successfully demonstrated over the years, the current limitation lies in the maximum distribution distance, which remains at 20 km fiber distance. Moreover, in these entangled high-dimensional QKD systems, the witness and distribution of quantum steering have not been shown before. Here we present a high-dimensional time-bin QKD protocol based on energy-time entanglement that generates a secure finite-length key capacity of 2.39 bit/coincidences and secure cryptographic finite-length keys at 0.24 Mbits s⁻¹in a 50 km optical fiber link. Our system is built entirely using readily available commercial off-the-shelf components, and secured by nonlocal dispersion cancellation technique against collective Gaussian attacks. Furthermore, we set new records for witnessing both energy-time entanglement and quantum steering over different fiber distances. When operating with a quantum channel loss of 39 dB, our system retains its inherent characteristic of utilizing large-alphabet. This enables us to achieve a secure key rate of 0.30 kbits s⁻¹and a secure key capacity of 1.10 bit/coincidences, considering finite-key effects. Our experimental results closely match the theoretical upper bound limit of secure cryptographic keys in high-dimensional time-bin QKD protocols (Moweret al2013Phys. Rev.A87062322; Zhanget al2014Phys. Rev. Lett.112120506), and outperform recent state-of-the-art qubit-based QKD protocols in terms of secure key throughput using commercial single-photon detectors (Wengerowskyet al2019Proc. Natl Acad. Sci.1166684; Wengerowskyet al2020npj Quantum Inf.65; Zhanget al2014Phys. Rev. Lett.112120506; Zhanget al2019Nat. Photon.13839; Liuet al2019Phys. Rev. Lett.122160501; Zhanget al2020Phys. Rev. Lett.125010502; Weiet al2020Phys. Rev.X10031030). The simple and robust entanglement-based high-dimensional time-bin protocol presented here provides potential for practical long-distance quantum steering and QKD with multiple secure bits-per-coincidence, and higher secure cryptographic keys compared to mature qubit-based QKD protocols.

 

Mode-locked biphoton frequency combs exhibit multiple discrete comblike temporal correlations from the Fourier transform of its phase-coherent frequency spectrum. Both temporal correlation and Franson interferometry are valuable tools for analyzing the joint properties of biphoton frequency combs, and the latter has proven to be essential for testing the fundamental quantum nature, the time-energy entanglement distribution, and the large-alphabet quantum key distributions. However, the Franson recurrence interference visibility in biphoton frequency combs unavoidably experiences a falloff that deteriorates the quality of time-energy entanglement and channel capacity for longer cavity round trips. In this paper, we provide a new method to address this problem towards optimum Franson interference recurrence. We first observe mode-locked temporal oscillations in a 5.03 GHz free-spectral range singly filtered biphoton frequency comb using only commercial detectors. Then, we observe similar falloff trend of time-energy entanglement in 15.15 GHz and 5.03 GHz free-spectral range singly filtered biphoton frequency combs, whereas, the optimum central time-bin accidental-subtracted visibility over 97% for both cavities. Here, we find that by increasing the cavity finesseF, we can enhance the detection probability in temporal correlations and towards optimum Franson interference recurrence in our singly filtered biphoton frequency combs. For the first time, via a higher cavity finesseFof 45.92 with a 15.11 GHz free-spectral range singly filtered biphoton frequency comb, we present an experimental ≈3.13-fold improvement of the Franson visibility compared to the Franson visibility with a cavity finesseFof 11.14 at the sixth time bin. Near optimum Franson interference recurrence and a time-bin Schmidt number near 16 effective modes in similar free-spectral range cavity are predicted with a finesseFof 200. Our configuration is versatile and robust against changes in cavity parameters that can be designed for various quantum applications, such as high-dimensional time-energy entanglement distributions, high-dimensional quantum key distributions, and wavelength-multiplexed quantum networks.

 

High-dimensional quantum entanglement is a cornerstone for advanced technology enabling large-scale noise-tolerant quantum systems, fault-tolerant quantum computing, and distributed quantum networks. The recently developed biphoton frequency comb (BFC) provides a powerful platform for high-dimensional quantum information processing in its spectral and temporal quantum modes. Here we propose and generate a singly-filtered high-dimensional BFC via spontaneous parametric down-conversion by spectrally shaping only the signal photons with a Fabry-Pérot cavity. High-dimensional energy-time entanglement is verified through Franson-interference recurrences and temporal correlation with low-jitter detectors. Frequency- and temporal- entanglement of our singly-filtered BFC is then quantified by Schmidt mode decomposition. Subsequently, we distribute the high-dimensional singly-filtered BFC state over a 10 km fiber link with a post-distribution time-bin dimension lower bounded to be at least 168. Our demonstrations of high-dimensional entanglement and entanglement distribution show the singly-filtered quantum frequency comb’s capability for high-efficiency quantum information processing and high-capacity quantum networks.

 

This tutorial targets researchers and practitioners who are interested in ML technologies for NLP from indirect supervision. In particular, we will present a diverse thread of indirect supervision studies that try to answer the following questions: (i) when and how can we provide supervision for a target task T, if all we have is data that corresponds to a “related” task T′? (ii) humans do not use exhaustive supervision; they rely on occasional feedback, and learn from incidental signals from various sources; how can we effectively incorporate such supervision in machine learning? (iii) how can we leverage multi-modal supervision to help NLP? To the end, we will discuss several lines of research that address those challenges, including (i) indirect supervision from T ′ that handles T with outputs spanning from a moderate size to an open space, (ii) the use of sparsely occurring and incidental signals, such as partial labels, noisy labels, knowledge-based constraints, and cross-domain or cross-task annotations—all having statistical associations with the task, (iii) principled ways to measure and understand why these incidental signals can contribute to our target tasks, and (iv) indirect supervision from vision-language signals. We will conclude the tutorial by outlining directions for further investigation. 

Adversarial Examples Detection (AED) is a crucial defense technique against adversarial attacks and has drawn increasing attention from the Natural Language Processing (NLP) community. Despite the surge of new AED methods, our studies show that existing methods heavily rely on a shortcut to achieve good performance. In other words, current search-based adversarial attacks in NLP stop once model predictions change, and thus most adversarial examples generated by those attacks are located near model decision boundaries. To surpass this shortcut and fairly evaluate AED methods, we propose to test AED methods with Far Boundary (FB) adversarial examples. Existing methods show worse than random guess performance under this scenario. To overcome this limitation, we propose a new technique, ADDMU, adversary detection with data and model uncertainty, which combines two types of uncertainty estimation for both regular and FB adversarial example detection. Our new method outperforms previous methods by 3.6 and 6.0 AUC points under each scenario. Finally, our analysis shows that the two types of uncertainty provided by ADDMU can be leveraged to characterize adversarialexamples and identify the ones that contribute most to model’s robustness in adversarial training.