Robust Markov Decision Processes (MDPs) offer a promising framework for computing reliable policies under model uncertainty. While policy gradient methods have gained increasing popularity in robust discounted MDPs, their application to the average-reward criterion remains largely unexplored. This paper proposes a Robust Projected Policy Gradient (RP2G), the first generic policy gradient method for robust average-reward MDPs (RAMDPs) that is applicable beyond the typical rectangularity assumption on transition ambiguity. In contrast to existing robust policy gradient algorithms, RP2G incorporates an adaptive decreasing tolerance mechanism for efficient policy updates at each iteration. We also present a comprehensive convergence analysis of RP2G for solving ergodic tabular RAMDPs. Furthermore, we establish the first study of the inner worst-case transition evaluation problem in RAMDPs, proposing two gradient-based algorithms tailored for rectangular and general ambiguity sets, each with provable convergence guarantees. Numerical experiments confirm the global convergence of our new algorithm and demonstrate its superior performance.

Content based video search services find extensive applications across various domains including video surveillance and object detection. In recent times, researchers have increasingly turned their attention towards enhancing the security of video search over outsourced encrypted videos. Nonetheless, prior researchers often leverage cost-expensive techniques like Homomorphic encryption or Order-preserving encryption to ensure privacy preservation. To reduce the overhead, Bloom Filter (BF)-encoded keyword search is a promising technology for retrieving encrypted videos with image queries. However, it generally suffers from serious data privacy leakage since it will reveal the inclusion relationship between “1” and “0” in the BF. Fortunately, the privacy-preserving bloom filter-based search scheme (PBKS) was recently proposed to achieve secure and effective search while protecting the values in BFs, but it still has two limitations. One is the size of a search token is very large in some cases and the other is the cloud server can infer the true value of each bit in the BF by doing a few operations. In this paper, we propose an efficient and confidentiality-preserving bloom filter-encoded video search (ECVS) scheme for retrieving encrypted videos with image queries. We first design a new CPRF (prefix-constrained pseudorandom function)-based token compression method to reduce the size of the search token and reduce the communication cost largely. Furthermore, we customize a periodic refresh mechanism to conceal the true value of each bit in the BF while avoiding excessive computational pressure on resource-limited users. Security analysis and experiments confirm the security and efficiency of our schemes.

Permissioned blockchain is a promising methodology to build zero-trust storage foundation with trusted data storage and sharing for the zero-trust network. However, the inherent full-backup feature of the permissioned blockchain poses potential data privacy risks and substantial storage costs, hindering its usage as a storage medium. These issues necessitate the usage of secure data deduplication technology to mitigate them. Unfortunately, current secure data deduplication schemes are predominantly designed with centralized cloud servers in mind and are not suitable for distributed blockchain systems. The reason is that the full backup feature of the permissioned blockchain renders a wide attack surface to offline brute-force and frequency analysis attacks. In response, we propose DedupChain, a secure blockchain-enabled storage system with deduplication for zero-trust networks. DedupChain employs a trusted execution environment (i.e., Inter SGX enclave) in conjunction with Oblivious RAM (ORAM) to offer a novel security guarantee namedoblivious data deduplication, which empowers DedupChain with the ability to defend offline brute-force and frequency analysis attacks. DedupChain also proposes several novel techniques to address the security and efficiency issues raised by the SGX enclave. We implemented a system prototype of DedupChain and evaluated its performance metrics. Our experimental results show that DedupChain exhibits satisfactory operational delays, throughput, and storage overhead. Security analysis shows that DedupChain is robust enough to withstand several types of attacks. To the best of our knowledge, we are the first to apply secure data deduplication techniques to address data privacy and storage cost issues raised by permissioned blockchain when used as a zero-trust storage medium.

Dynamic searchable symmetric encryption (DSSE) enables clients to store encrypted data on untrusted servers while retaining the ability to search and update the data efficiently. However, most existing DSSE schemes are vulnerable to incorrect update queries, such as duplicated insertions or invalid deletions, which can compromise both security and availability. Although existing robust schemes have made progress in addressing these issues, they still suffer from significant search inefficiencies, particularly when handling large numbers of updates. To overcome these limitations, we proposeRO(SE)2, a novel robust DSSE scheme that simultaneously achieves robustness, forward-and-Type-III-backward security, and optimal search performance.RO(SE)2introduces a hierarchical binary tree structure combined with an oblivious map (OMAP) to handle incorrect updates during the update phase, eliminating the need for filtering during search queries and significantly improving search efficiency. Additionally,RO(SE)2employs a two-layer encryption mechanism to ensure forward security and supports efficient search result verification through its verifiable extension,RO(SE)2-v. Rigorous security analysis proves thatRO(SE)2can achieve not only robustness, forward and backward security but optimal search efficiency as well. Comparative analysis reveals thatRO(SE)2outperforms existing robust schemes in terms of search performance, whileRO(SE)2-v outperforms the state-of-the-art verifiable robust schemes in verification performance.

A deep equilibrium model (DEQ) is implicitly defined through an equilibrium point of an infinite-depth weight-tied model with an input-injection. Instead of infinite computations, it solves an equilibrium point directly with root-finding and computes gradients with implicit differentiation. In this paper, the training dynamics of over-parameterized DEQs are investigated, and we propose a novel probabilistic framework to overcome the challenge arising from the weight-sharing and the infinite depth. By supposing a condition on the initial equilibrium point, we prove that the gradient descent converges to a globally optimal solution at a linear convergence rate for the quadratic loss function. We further perform a fine-grained non-asymptotic analysis about random DEQs and the corresponding weight-untied models, and show that the required initial condition is satisfied via mild over-parameterization. Moreover, we show that the unique equilibrium point always exists during the training.

Robust Markov decision processes (RMDPs) provide a promising framework for computing reliable policies in the face of model errors. Many successful reinforcement learning algorithms build on variations of policy-gradient methods, but adapting these methods to RMDPs has been challenging. As a result, the applicability of RMDPs to large, practical domains remains limited. This paper proposes a new Double-Loop Robust Policy Gradient (DRPG), the first generic policy gradient method for RMDPs. In contrast with prior robust policy gradient algorithms, DRPG monotonically reduces approximation errors to guarantee convergence to a globally optimal policy in tabular RMDPs. We introduce a novel parametric transition kernel and solve the inner loop robust policy via a gradient-based method. Finally, our numerical results demonstrate the utility of our new algorithm and confirm its global convergence properties.

As the integration of the Internet of Vehicles and social networks, vehicular social networks (VSNs) are promising to boost the realization of intelligent transportation system. Recently, vehicular data privacy has been paid increasing attention in data sharing. Searchable encryption as a promising cryptographic primitive can be utilized to ensure vehicular data confidentiality without sacrificing data searchability. However, most vehicular data sharing schemes rely on centralized cloud servers, which are vulnerable to the single point of failure and distributed denial of service (DDoS) attacks. In this paper, we propose VehShare, a decentralized framework for privacy-preserving vehicular data sharing. We resort to the smart contract to implement a trusted platform for vehicles to share their encrypted vehicular data. To provide efficient access control, we design an authorization-based on-chain access control scheme with a lightweight cryptographic primitive. Moreover, we design a time synchronization-based non-interactive search token generation scheme to achieve efficient privacy-preserving search queries, while satisfying forward and backward security. We formally analyze the security of VehShare and extensive experiments demonstrate the efficiency of VehShare.

To reveal the mystery behind deep neural networks (DNNs), optimization may offer a good perspective. There are already some clues showing the strong connection between DNNs and optimization problems, e.g., under a mild condition, DNN's activation function is indeed a proximal operator. In this paper, we are committed to providing a unified optimization induced interpretability for a special class of networks-equilibrium models, i.e., neural networks defined by fixed point equations, which have become increasingly attractive recently. To this end, we first decompose DNNs into a new class of unit layer that is the proximal operator of an implicit convex function while keeping its output unchanged. Then, the equilibrium model of the unit layer can be derived, we name it Optimization Induced Equilibrium Networks (OptEq). The equilibrium point of OptEq can be theoretically connected to the solution of a convex optimization problem with explicit objectives. Based on this, we can flexibly introduce prior properties to the equilibrium points: 1) modifying the underlying convex problems explicitly so as to change the architectures of OptEq; and 2) merging the information into the fixed point iteration, which guarantees to choose the desired equilibrium point when the fixed point set is non-singleton. We show that OptEq outperforms previous implicit models even with fewer parameters.

Incomplete multi-view clustering has attracted much attention in decade years. To date, most of the remarkable achievements, however, exploit shallow models to learn shared feature representations based on incomplete views. Although some deep learning methods have been proposed to solve this issue, the existing ones still have the following problems: 1) The consistency between views is ignored, which will have serious negative impacts on incomplete multi-view learning. 2) The learned features do not have sufficient cluster-friendliness, that is, the tightness within clusters and the repulsiveness between clusters are not fully considered. To tackle the above shortcomings, we propose a Dual Alignment Self-supervised Incomplete Multi-view Subspace Clustering network (DASIMSC) in this paper. Specifically, the manifold alignment constraint and consistency alignment constraint are integrated with the autoencoder to preserve the compact inherent local structure within the view and the consistency semantics between incomplete views, respectively. Moreover, a self-expression layer coupled with a spectral clustering module is designed to naturally separate different types of data, leveraging the current clustering results to supervise subspace learning, which excludes inter-cluster. Experimental results on several datasets show that our algorithm outperforms all compared state-of-the-arts.