Searcharxiv⌕ Search

arXiv subjects

Syed Mahbub Hafiz

Publications and source records attributed to Syed Mahbub Hafiz.

4 recordsLinked to original sources
arXiv2026

Lightweight, Practical Encrypted Face Recognition with GPU Support

Face recognition models operate in a client-server setting where a client extracts a compact face embedding and a server performs similarity search over a template database. This raises privacy concerns, as facial data is highly sensitive. To provide cryptographic privacy guarantees, one can use fully homomorphic encryption to perform end-to-end encrypted similarity search. However, existing FHE-based protocols are computationally costly and, impose high memory overhead. Building on prior work, HyDia (PoPETS 2025), we introduce algorithmic and system-level improvements targeting real-world deployment with resource-constrained clients. First, we propose BSGS-Diagonal, an algorithm delivering fast and memory-efficient similarity computation. BSGS-Diagonal substantially shrinks the rotation-key set, lowering both client and server memory requirements, and also improves practical server runtime. This yields a 91% reduction in the number of rotation keys, translating to approximately 14 GB less memory used on the client, and reducing overall CPU peak RAM from over 33 GB in the original HyDia to under 11 GB for databases up to size 1M. In addition, runtime is improved by up to 1.57x for the membership verification scenario and 1.43x for the identification scenario. Secondly, we introduce fully GPU-optimized similarity matrix computation kernels. The implementation is built upon FIDESlib, a CKKS-level GPU library based on OpenFHE. Rather than offloading individual CKKS primitives in isolation, the integrated kernels fuse operations to avoid repeated CPU-GPU ciphertext movement and costly FIDESlib/OpenFHE data-structure conversions. As a result, our GPU implementations of both HyDia and BSGS-Diagonal achieve up to 9x and 21x speedups, respectively, enabling sub-second encrypted face recognition for databases up to 32K entries while further reducing host memory usage.

cs.CR↗
arXiv2024

Private Aggregate Queries to Untrusted Databases

Private information retrieval (PIR), a privacy-preserving cryptographic tool, solves a simplified version of this problem by hiding the database item that a client accesses. Most PIR protocols require the client to know the exact row index of the intended database item, which cannot support the complicated aggregation-based statistical query in a similar setting. Some works in the PIR space contain keyword searching and SQL-like queries, but most need multiple interactions between the PIR client and PIR servers. Some schemes support searching SQL-like expressive queries in a single round but fail to enable aggregate queries. These schemes are the main focus of this paper. To bridge the gap, we have built a general-purpose novel information-theoretic PIR (IT-PIR) framework that permits a user to fetch the aggregated result, hiding all sensitive sections of the complex query from the hosting PIR server in a single round of interaction. In other words, the server will not know which records contribute to the aggregation. We then evaluate the feasibility of our protocol for both benchmarking and real-world application settings. For instance, in a complex aggregate query to the Twitter microblogging database of 1 million tweets, our protocol takes 0.014 seconds for a PIR server to generate the result when the user is interested in one of 3K user handles. In contrast, for a much-simplified task, not an aggregate but a positional query, Goldberg's regular IT-PIR (Oakland 2007) takes 1.13 seconds. For all possible user handles, 300K, it takes equal time compared to the regular IT-PIR. This example shows that complicated aggregate queries through our framework do not incur additional overhead if not less, compared to the conventional query.

cs.CR↗
arXiv2023

Benchmarking Adversarial Robustness of Compressed Deep Learning Models

The increasing size of Deep Neural Networks (DNNs) poses a pressing need for model compression, particularly when employed on resource constrained devices. Concurrently, the susceptibility of DNNs to adversarial attacks presents another significant hurdle. Despite substantial research on both model compression and adversarial robustness, their joint examination remains underexplored. Our study bridges this gap, seeking to understand the effect of adversarial inputs crafted for base models on their pruned versions. To examine this relationship, we have developed a comprehensive benchmark across diverse adversarial attacks and popular DNN models. We uniquely focus on models not previously exposed to adversarial training and apply pruning schemes optimized for accuracy and performance. Our findings reveal that while the benefits of pruning enhanced generalizability, compression, and faster inference times are preserved, adversarial robustness remains comparable to the base model. This suggests that model compression while offering its unique advantages, does not undermine adversarial robustness.

cs.LG↗