arXiv subjects

Junquan Deng

Publications and source records attributed to Junquan Deng.

5 recordsLinked to original sources

arXiv2026

Understanding the (In)Security of Vibe-Coded Applications

Recent advances in large language models (LLMs) have enabled vibe coding, an emerging software development paradigm in which users create applications primarily through natural-language interactions with AI agents. Due to its low barrier to entry, vibe coding is rapidly gaining adoption in practice. Unlike conventional AI-assisted programming, where developers remain responsible for implementation and code review, vibe coding delegates a substantial portion of the development process to AI systems. This shift raises a fundamental question: how (in)secure are applications developed through vibe coding? In this paper, we conduct a systematic study of the security of real-world vibe-coded applications. We collect 9,041 open-source applications developed using popular AI agents (Claude Code and Lovable), and audit 200 publicly deployed applications, uncovering 1,186 vulnerabilities. Our study of these applications and vulnerabilities reveals several key findings: (1) insecurity is the norm rather than the exception: 91.0\% of audited applications contain at least one vulnerability, and 65.77\% of identified vulnerabilities are rated Critical or High severity, concentrated in broken access control, injection, and authentication failures; (2) these vulnerabilities are traceable to eight recurring failure modes rooted in three systematic limitations of AI agents: memory defects, objective defects, and knowledge defects; and (3) while improved agent harness and prompting strategies can reduce the incidence of vulnerabilities, they do not eliminate the underlying security risks. Overall, our study provides an empirical understanding of the security landscape of vibe-coded applications and lays the groundwork for addressing the security risks in the growing delegation of software development to AI systems.

cs.CR↗

arXiv2022

A Hybrid Millimeter-wave Channel Simulator for Joint Communication and Localization

Joint communication and localization~(JCL) is envisioned to be a key feature in future millimeter-wave~(mmWave) wireless networks for context-aware applications. A map-based channel model considering both site-specific radio environment and statistical channel characteristics is essential to facilitate JCL research and to evaluate the performance of various JCL systems. To this end, this paper presents an open-source hybrid mmWave channel simulator called OmniSIM for site-specific JCL research, which uses digital map, network layout and user trajectories as inputs to predict the channel responses between users and base stations. A fast shooting-bouncing rays~(FSBR) algorithm combined with Computational Electromagnetic, has been developed to generate channel parameters relevant to JCL, considering mmWave reflection, diffusing, diffraction and scattering.

eess.SP↗

arXiv2022

Supervised Contrastive CSI Representation Learning for Massive MIMO Positioning

Similarity metric is crucial for massive MIMO positioning utilizing channel state information~(CSI). In this letter, we propose a novel massive MIMO CSI similarity learning method via deep convolutional neural network~(DCNN) and contrastive learning. A contrastive loss function is designed considering multiple positive and negative CSI samples drawn from a training dataset. The DCNN encoder is trained using the loss so that positive samples are mapped to points close to the anchor's encoding, while encodings of negative samples are kept away from the anchor's in the representation space. Evaluation results of fingerprint-based positioning on a real-world CSI dataset show that the learned similarity metric improves positioning accuracy significantly compared with other known state-of-the-art methods.

cs.IT↗

arXiv2021

Semi-supervised t-SNE for Millimeter-wave Wireless Localization

We consider the mobile localization problem in future millimeter-wave wireless networks with distributed Base Stations (BSs) based on multi-antenna channel state information (CSI). For this problem, we propose a Semi-supervised tdistributed Stochastic Neighbor Embedding (St-SNE) algorithm to directly embed the high-dimensional CSI samples into the 2D geographical map. We evaluate the performance of St-SNE in a simulated urban outdoor millimeter-wave radio access network. Our results show that St-SNE achieves a mean localization error of 6.8 m with only 5% of labeled CSI samples in a 200*200 m^2 area with a ray-tracing channel model. St-SNE does not require accurate synchronization among multiple BSs, and is promising for future large-scale millimeter-wave localization.

cs.LG↗

arXiv2018

MmWave Channel Estimation via Atomic Norm Minimization for Multi-User Hybrid Precoding

To perform multi-user multiple-input and multiple-output transmission in millimeter-wave(mmWave) cellular systems, the high-dimensional channels need to be estimated for designing the multi-user precoder. Conventional grid-based compressed sensing (CS) methods for mmWave channel estimation suffer from the basis mismatch problem, which prevents accurate channel reconstruction and degrades the precoding performance. This paper formulates mmWave channel estimation as an Atomic Norm Minimization (ANM) problem. In contrast to grid-based CS methods which use discrete dictionaries, ANM uses a continuous dictionary for representing the mmWave channel. We consider a continuous dictionary based on sub-sampling in the antenna domain via a small number of radio frequency chains. We show that mmWave channel estimation using ANM can be formulated as a semidefinite programming (SDP) problem, and the user channel can be accurately estimated via off-the-shelf SDP solvers in polynomial time. Simulation results indicate that ANM can achieve much better estimation accuracy compared to grid-based CS, and significantly improves the spectral efficiency provided by multi-user precoding.

cs.IT↗