The coded-aperture snapshot spectral imaging (CASSI) system has garnered significant attention as a promising spectral snapshot-based imaging technique. However, several challenges persist in the development of CASSI systems: 1. High hardware costs. 2. The difficulty of balancing the performance and computational efficiency of the reconstruction network. In this paper, we propose a dual-camera mask-free CASSI system to replace the expensive encoding hardware, thereby reducing the overall cost of CASSI. To address the issues of a high number of parameters and computational demands associated with Transformer-based methods, we design a hierarchical attention-based spectral-spatial transformer. This approach involves embedding local features on both the spectral and spatial channels, while also leveraging global information from these channels to enhance the network’s global learning capabilities. Our experimental results demonstrate the superiority of the mask-free approach, achieving a PSNR of 37.9 dB in a single camera and 44.7 dB in dual cameras. Furthermore, experiments indicate that the mask-free system outperforms state-of-the-art methods while requiring lower computational and memory costs.

In the secure two-party computation (2PC), an adversary is often categorized as semi-honest or malicious, depending on whether it follows the protocol specifications. Covert security (Aumann and Lindell, 2010) first looks into the “middle ground”, such that an active adversary who cheats will be caught with a predefined probability. Other security notions, such as publicly auditable security (Baum et al., 2014) and (robust) accountability family (Küsters et al., 2010; Graf et al., 2023; Rivinius et al., 2022), achieve public verifiability as a stronger security guarantee by relying on heavy offline and online constructions with zero knowledge proofs and (or) a bulletin board functionality. In this work, we propose a new security notion called honorific security, where an external arbiter can identify the cheater without a bulletin board. Specifically, we delay and outsource the verification steps to the arbiter, so that the original online computation is thus accelerated. We show that a maliciously secure garbled circuit (GC) (Yao, 1986) protocol can be constructed with only slightly more overhead than a passively secure protocol. Our construction performs up to 2.37 times and 13.30 times as fast as the state-of-the-art protocols with covert and malicious security, respectively.

Federated learning (FL) proposes to train a global machine learning model across distributed datasets. However, the aggregation protocol as the core component in FL is vulnerable to well-studied attacks, such as inference attacks, poisoning attacks [71] and malicious participants who try to deviate from the protocol [24]. Therefore, it is crucial to achieve both malicious security and poisoning resilience from cryptographic and FL perspectives, respectively. Prior works either achieve incomplete malicious security [76], address issues by using expensive cryptographic tools [22, 59] or assume the availability of a clean dataset on the server side [32]. In this work, we propose AlphaFL, a two-server secure aggregation protocol achieving both malicious security in the universal composability (UC) framework [19] and poisoning resilience in FL (thus malicious2) against a dishonest majority. We design maliciously secure multi-party computation (MPC) protocols [24, 26, 48] and introduce an efficient input commitment protocol tolerating server-client collusion (dishonest majority). We also propose an efficient input commitment protocol for the non-collusion case (honest majority), which triples the efficiency in time and quadruples that in communication, compared to the state-of-the-art solution in MP-SPDZ [46]. To achieve poisoning resilience, we carry out 𝐿∞ and 𝐿2-Norm checks with a dynamic L_2-Norm bound by introducing a novel silent select protocol, which improves the runtime by at least two times compared to the classic select protocol. Combining these, AlphaFL achieves malicious2 security at a cost of 25% − 79% more runtime overhead than the state-of-the-art semi-malicious counterpart Elsa [76], with even less communication cost.

As a well known machine learning model, Gradient Boosting Decision Tree (GBDT) is widely used in many real-world scenes such as online marketing, risk management, fraud detection and recommendation systems. Due to limited data resources, two data owners may collaborate with each other to jointly train a high-quality model. As privacy regulations such as HIPPA and GDPR come into force, Privacy-Preserving Machine Learning (PPML) has drawn increasingly higher attention. Recently, a line of works [3--6] studies function secret sharing (FSS) schemes in the preprocessing model, where the online stage of secure two-party computation (2PC) is significantly improved. While recent privacy-preserving GDBT frameworks mainly focus on improving the performance of a singular module (e.g. secure bucket aggregation), we propose SiGBDT, a globally silent two-party GBDT framework via function secret sharing on a vertically partitioned dataset. During the training process, we apply FSS schemes to construct efficient modular protocols, such as secure bucket aggregation, argmax computation and a node split approach. We run in-depth experiments and discover that SiGBDT completely outperforms state-of-the-art frameworks. The experiment results show that SiGBDT is at least 3.32 X faster in LAN and at least 6.4 X faster in WAN.

In this work, we leverage the pure skin color patch from the face image as the additional information to train an auxiliary skin color feature extractor and face recognition model in parallel to improve performance of state-of-the-art (SOTA) privacy-preserving face recognition (PPFR) systems. Our solution is robust against black-box attacking and well-established generative adversarial network (GAN) based image restoration. We analyze the potential risk in previous work, where the proposed cosine similarity computation might directly leak the protected precomputed embedding stored on the server side. We propose a Function Secret Sharing (FSS) based face embedding comparison protocol without any intermediate result leakage. In addition, we show in experiments that the proposed protocol is more efficient compared to the Secret Sharing (SS) based protocol.

Jiali Zeng, Yufan Jiang, Yongjing Yin, Yi Jing, Fandong Meng, Binghuai Lin, Yunbo Cao, Jie Zhou. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Unsupervised summarization methods have achieved remarkable results by incorporating representations from pre-trained language models. However, existing methods fail to consider efficiency and effectiveness at the same time when the input document is extremely long. To tackle this problem, in this paper, we proposed an efficient Coarse-to-Fine Facet-Aware Ranking (C2F-FAR) framework for unsupervised long document summarization, which is based on the semantic block. The semantic block refers to continuous sentences in the document that describe the same facet. Specifically, we address this problem by converting the one-step ranking method into the hierarchical multi-granularity two-stage ranking. In the coarse-level stage, we proposed a new segment algorithm to split the document into facet-aware semantic blocks and then filter insignificant blocks. In the fine-level stage, we select salient sentences in each block and then extract the final summary from selected sentences. We evaluate our framework on four long document summarization datasets: Gov-Report, BillSum, arXiv, and PubMed. Our C2F-FAR can achieve new state-of-the-art unsupervised summarization results on Gov-Report and BillSum. In addition, our method speeds up 4-28 times more than previous methods.

Recent research questions the importance of the dot-product self-attention in Transformer models and shows that most attention heads learn simple positional patterns.In this paper, we push further in this research line and propose a novel substitute mechanism for self-attention: Recurrent AtteNtion (RAN).RAN directly learns attention weights without any token-to-token interaction and further improves their capacity by layer-to-layer interaction.Across an extensive set of experiments on 10 machine translation tasks, we find that RAN models are competitive and outperform their Transformer counterpart in certain scenarios, with fewer parameters and inference time.Particularly, when apply RAN to the decoder of Transformer, there brings consistent improvements by about +0.5 BLEU on 6 translation tasks and +1.0 BLEU on Turkish-English translation task.In addition, we conduct extensive analysis on the attention weights of RAN to confirm their reasonableness.Our RAN is a promising alternative to build more effective and efficient NMT models.

Neural architecture search (NAS) has advanced significantly in recent years but most NAS systems restrict search to learning architectures of a recurrent or convolutional cell.In this paper, we extend the search space of NAS.In particular, we present a general approach to learn both intra-cell and inter-cell architectures (call it ESS).For a better search result, we design a joint learning method to perform intra-cell and inter-cell NAS simultaneously.We implement our model in a differentiable architecture search system.For recurrent neural language modeling, it outperforms a strong baseline significantly on the PTB and Wiki-Text data, with a new state-of-the-art on PTB.Moreover, the learned architectures show good transferability to other systems.E.g., they improve state-of-the-art systems on the CoNLL and WNUT named entity recognition (NER) tasks and CoNLL chunking task, indicating a promising line of research on large-scale prelearned architectures.

In encoder-decoder neural models, multiple encoders are in general used to represent the contextual information in addition to the individual sentence.In this paper, we investigate multi-encoder approaches in document-level neural machine translation (NMT).Surprisingly, we find that the context encoder does not only encode the surrounding sentences but also behaves as a noise generator.This makes us rethink the real benefits of multi-encoder in context-aware translation -some of the improvements come from robust training.We compare several methods that introduce noise and/or well-tuned dropout setup into the training of these encoders.Experimental results show that noisy training plays an important role in multi-encoder-based NMT, especially when the training data is small.Also, we establish a new state-of-the-art on IWSLT Fr-En task by careful use of noise generation and dropout methods.

Large amounts of data has made neural machine translation (NMT) a big success in recent years.But it is still a challenge if we train these models on small-scale corpora.In this case, the way of using data appears to be more important.Here, we investigate the effective use of training data for low-resource NMT.In particular, we propose a dynamic curriculum learning (DCL) method to reorder training samples in training.Unlike previous work, we do not use a static scoring function for reordering.Instead, the order of training samples is dynamically determined in two ways -loss decline and model competence.This eases training by highlighting easy samples that the current model has enough competence to learn.We test our DCL method in a Transformerbased system.Experimental results show that DCL outperforms several strong baselines on three low-resource machine translation benchmarks and different sized data of WMT'16 En-De.

Deep encoders have been proven to be effective in improving neural machine translation (NMT) systems, but training an extremely deep encoder is time consuming.Moreover, why deep models help NMT is an open question.In this paper, we investigate the behavior of a well-tuned deep Transformer system.We find that stacking layers is helpful in improving the representation ability of N-MT models and adjacent layers perform similarly.This inspires us to develop a shallowto-deep training method that learns deep models by stacking shallow models.In this way, we successfully train a Transformer system with a 54-layer encoder.Experimental results on WMT'16 English-German and WMT'14 English-French translation tasks show that it is 1.4 × faster than training from scratch, and achieves a BLEU score of 30.33 and 43.29 on two tasks.The code is publicly available at https://github.com/libeineu/ SDT-Training.

Yufan Jiang, Chi Hu, Tong Xiao, Chunliang Zhang, Jingbo Zhu. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.