arXiv subjects

Tianyu Lu

Publications and source records attributed to Tianyu Lu.

13 recordsLinked to original sources

arXiv2026

The physical foundation beneath protein generative modeling

Physical equations in protein modeling appear to have been replaced by generative models trained directly on structure data. By learning a mapping from noise to data, sampling de novo protein structures has become much more efficient. However, such models can also learn non-physical features and break down with out-of-distribution settings which are typical in protein design campaigns. In this review, we highlight where the physics persists in protein generative modeling pipelines and the issues that linger when physically relevant components of a macromolecular system are left unmodeled. We present a perspective that respecting the underlying physics of macromolecular systems, increasingly through learned representations that are physically grounded and updating generative models with experimental data, is foundational to generative modeling for functional protein design.

q-bio.BM↗

arXiv2026

Toward Security-Resilient Cell-Free Massive MIMO: A Multi-Stage Framework

This paper develops a robust security-resilient transmission framework for cell-free massive multiple-input multiple-output (CF-mMIMO) systems under active pilot spoofing attacks. As a baseline, system performance is characterized under attack-free conditions to establish the target user's pre-attack service level. Upon attack detection, the system enters an absorption phase, during which, power allocation is adaptively adjusted across a limited subset of access points (APs) using contaminated channel state information (CSI). This phase quickly compensates for performance degradation while maintaining low operational overhead. Once the secrecy spectral efficiency (SSE) recovers to a prescribed loss level, the resulting power allocation initializes the restoration phase. Here, the transmit powers of all APs are jointly optimized, and a protective partial zero-forcing (PPZF) strategy further improves secrecy. In parallel, artificial noise (AN) is incorporated under a worst-case eavesdropping scenario accounting for large-scale fading uncertainty. The resulting stage-dependent non-convex problems are formulated within a unified framework and solved using successive convex approximation (SCA). Numerical results demonstrate that the proposed scheme achieves an effective time-quality tradeoff while maintaining the highest recovered secrecy; in a representative setup, it achieves gains of up to 3.6%, 15.7%, and 56% over the respective baseline schemes, with similar improvements under eavesdropper's channel uncertainty.

eess.SP↗

arXiv2026

Resilient Cell-Free Massive MIMO Networks

This paper proposes a novel optimization framework for enhancing the security resilience of cell-free massive multiple-input multiple-output (CF-mMIMO) networks with multi-antenna access points (APs) and protective partial zero-forcing (PPZF) under active eavesdropping. Based on the main principles of absorption, adaptation, and recovery, we formulate a security-aware resilience metric to quantify the system performance during and after a security outage. A multi-user service priority-aware power allocation problem is formulated to minimize the mean squared error (MSE) between real-time and desired security efficiency, thereby enabling a trade-off between the target user's secrecy performance and multi-user quality of service (QoS). To solve this non-convex problem, a security-aware iterative algorithm based on the successive convex approximation (SCA) is employed. The proposed algorithm determines the optimal power allocation strategy by balancing solution quality against recovery time. At each iteration, it evaluates the overall resilience score and selects the strategy that achieves the highest value. Simulation results confirm that the proposed framework significantly improves the resilience of CF-mMIMO networks, allowing flexible adaptation between rapid recovery and high-quality recovery, depending on system requirements.

eess.SP↗

arXiv2026

Delay Alignment Modulation for Secure ISAC Systems

This paper introduces delay-alignment modulation (DAM) for secure integrated sensing and communication (ISAC). Due to the broadcast nature of multi-user downlinks, communications are vulnerable to eavesdropping. DAM applies controlled per-path symbol delays at the transmitter to coherently align the multipath components at the intended user, enhancing the received signal power, while simultaneously creating delay misalignment at the eavesdropper (Eve). To mitigate sensing degradation caused by multipath propagation, we propose a two-stage protocol that first estimates the angle and then the delay of the line-of-sight (LoS) path after suppressing multipath interference. We derive the secrecy spectral efficiency (SSE) and the Cramer-Rao (CRB) of the target delay. Finally, we develop a path-based zero-forcing (ZF) precoding framework and formulate a max-min SSE design under CRB and power constraints. Simulation results show DAM significantly outperforms the strongest-path (SP) benchmark in terms of SSE, while meeting sensing requirements, since intentional delay alignment at legitimate users degrades reception at Eve.

cs.IT↗

arXiv2025

Adversarial Learning-Based Radio Map Reconstruction for Fingerprinting Localization

This letter presents a feature-guided adversarial framework, namely ComGAN, which is designed to reconstruct an incomplete fingerprint database by inferring missing received signal strength (RSS) values at unmeasured reference points (RPs). An auxiliary subnetwork is integrated into a conditional generative adversarial network (cGAN) to enable spatial feature learning. An optimization method is then developed to refine the RSS predictions by aggregating multiple prediction sets, achieving an improved localization performance. Experimental results demonstrate that the proposed scheme achieves a root mean squared error (RMSE) comparable to the ground-truth measurements while outperforming state-of-the-art reconstruction methods. When the reconstructed fingerprint is combined with measured data for training, the fingerprinting localization achieves accuracy comparable to models trained on fully measured datasets.

eess.SP↗

arXiv2025

Your Harness is Not Secure: Benchmarking Real-world Threat of Command Line Interface Agent

Command-line interface (CLI) agents powered by large language models (LLMs) can interpret natural-language requests, plan multi-step tasks, execute shell commands, and modify files and system state. As these agents are increasingly used for operating-system (OS) workflows, it is important to evaluate whether they can be misused to carry out security-relevant operations. Existing benchmarks often lack an attacker-knowledge model grounded in tactics, techniques, and procedures (TTPs), provide limited coverage of end-to-end kill chains, rely on simplified single-host environments, or use LLM-as-a-judge for success evaluation. We introduce AdvCLI, an MITRE ATT&CK-aligned benchmark for evaluating OS-level misuse risks of CLI agents in a controlled multi-host sandbox. AdvCLI contains 140 tasks: 40 direct malicious requests, 74 TTP-based tasks, and 26 end-to-end kill chains. Each task is paired with deterministic hard-coded verification protocols that check whether the requested OS-level effect is realized. We evaluate seven CLI agents and products built on nine foundation models, including ReAct, OpenClaw, OpenAI Agent SDK, Claude Code, Gemini CLI, Cursor CLI, and Cursor IDE. Results show that current CLI agents frequently proceed beyond refusal and can complete a non-negligible fraction of malicious OS-level tasks, especially when requests include TTP-style attacker knowledge. AdvCLI provides a reproducible testbed for evaluating these risks and for developing stronger safety mechanisms for tool-using CLI agents in the future.

cs.CR↗

arXiv2025

GFS: A Preemption-aware Scheduling Framework for GPU Clusters with Predictive Spot Instance Management

The surge in large language models (LLMs) has fundamentally reshaped the landscape of GPU usage patterns, creating an urgent need for more efficient management strategies. While cloud providers employ spot instances to reduce costs for low-priority (LP) tasks, existing schedulers still grapple with high eviction rates and lengthy queuing times. To address these limitations, we present GFS, a novel preemptive scheduling framework that enhances service-level objective (SLO) compliance for high-priority (HP) tasks while minimizing preemptions to LP tasks. Firstly, GFS utilizes a lightweight forecasting model that predicts GPU demand among different tenants, enabling proactive resource management. Secondly, GFS employs a dynamic allocation mechanism to adjust the spot quota for LP tasks with guaranteed durations. Lastly, GFS incorporates a preemptive scheduling policy that prioritizes HP tasks while minimizing the impact on LP tasks. We demonstrate the effectiveness of GFS through both real-world implementation and simulations. The results show that GFS reduces eviction rates by 33.0\%, and cuts queuing delays by 44.1\% for LP tasks. Furthermore, GFS enhances the GPU allocation rate by up to 22.8\% in real production clusters. In a production cluster of more than 10,000 GPUs, GFS yields roughly \$459,715 in monthly benefits.

cs.DC↗

arXiv2025

Doxing via the Lens: Revealing Location-related Privacy Leakage on Multi-modal Large Reasoning Models

Recent advances in multi-modal large reasoning models (MLRMs) have shown significant ability to interpret complex visual content. While these models enable impressive reasoning capabilities, they also introduce novel and underexplored privacy risks. In this paper, we identify a novel category of privacy leakage in MLRMs: Adversaries can infer sensitive geolocation information, such as a user's home address or neighborhood, from user-generated images, including selfies captured in private settings. To formalize and evaluate these risks, we propose a three-level visual privacy risk framework that categorizes image content based on contextual sensitivity and potential for location inference. We further introduce DoxBench, a curated dataset of 500 real-world images reflecting diverse privacy scenarios. Our evaluation across 11 advanced MLRMs and MLLMs demonstrates that these models consistently outperform non-expert humans in geolocation inference and can effectively leak location-related private information. This significantly lowers the barrier for adversaries to obtain users' sensitive geolocation information. We further analyze and identify two primary factors contributing to this vulnerability: (1) MLRMs exhibit strong reasoning capabilities by leveraging visual clues in combination with their internal world knowledge; and (2) MLRMs frequently rely on privacy-related visual clues for inference without any built-in mechanisms to suppress or avoid such usage. To better understand and demonstrate real-world attack feasibility, we propose GeoMiner, a collaborative attack framework that decomposes the prediction process into two stages: clue extraction and reasoning to improve geolocation performance while introducing a novel attack perspective. Our findings highlight the urgent need to reassess inference-time privacy risks in MLRMs to better protect users' sensitive information.

cs.CR↗

arXiv2024

ESM All-Atom: Multi-scale Protein Language Model for Unified Molecular Modeling

Protein language models have demonstrated significant potential in the field of protein engineering. However, current protein language models primarily operate at the residue scale, which limits their ability to provide information at the atom level. This limitation prevents us from fully exploiting the capabilities of protein language models for applications involving both proteins and small molecules. In this paper, we propose ESM-AA (ESM All-Atom), a novel approach that enables atom-scale and residue-scale unified molecular modeling. ESM-AA achieves this by pre-training on multi-scale code-switch protein sequences and utilizing a multi-scale position encoding to capture relationships among residues and atoms. Experimental results indicate that ESM-AA surpasses previous methods in protein-molecule tasks, demonstrating the full utilization of protein language models. Further investigations reveal that through unified molecular modeling, ESM-AA not only gains molecular knowledge but also retains its understanding of proteins. The source codes of ESM-AA are publicly released at https://github.com/zhengkangjie/ESM-AA.

q-bio.BM↗

arXiv2023

Secret Key Generation for IRS-Assisted Multi-Antenna Systems: A Machine Learning-Based Approach

Physical-layer key generation (PKG) based on wireless channels is a lightweight technique to establish secure keys between legitimate communication nodes. Recently, intelligent reflecting surfaces (IRSs) have been leveraged to enhance the performance of PKG in terms of secret key rate (SKR), as it can reconfigure the wireless propagation environment and introduce more channel randomness. In this paper, we investigate an IRS-assisted PKG system, taking into account the channel spatial correlation at both the base station (BS) and the IRS. Based on the considered system model, the closed-form expression of SKR is derived analytically considering correlated eavesdropping channels. Aiming to maximise the SKR, a joint design problem of the BS precoding matrix and the IRS phase shift vector is formulated. To address this high-dimensional non-convex optimisation problem, we propose a novel unsupervised deep neural network (DNN)-based algorithm with a simple structure. Different from most previous works that adopt iterative optimisation to solve the problem, the proposed DNN-based algorithm directly obtains the BS precoding and IRS phase shifts as the output of the DNN. Simulation results reveal that the proposed DNN-based algorithm outperforms the benchmark methods with regard to SKR.

eess.SP↗

arXiv2023

Machine Learning-Based Secret Key Generation for IRS-assisted Multi-antenna Systems

Physical-layer key generation (PKG) based on wireless channels is a lightweight technique to establish secure keys between legitimate communication nodes. Recently, intelligent reflecting surfaces (IRSs) have been leveraged to enhance the performance of PKG in terms of secret key rate (SKR), as it can reconfigure the wireless propagation environment and introduce more channel randomness. In this paper, we investigate an IRS-assisted PKG system, taking into account the channel spatial correlation at both the base station (BS) and the IRS. Based on the considered system model, the closed form expression of SKR is derived analytically. Aiming to maximize the SKR, a joint design problem of the BS precoding matrix and the IRS reflecting coefficient vector is formulated. To address this high-dimensional non-convex optimization problem, we propose a novel unsupervised deep neural network (DNN) based algorithm with a simple structure. Different from most previous works that adopt the iterative optimization to solve the problem, the proposed DNN based algorithm directly obtains the BS precoding and IRS phase shifts as the output of the DNN. Simulation results reveal that the proposed DNN-based algorithm outperforms the benchmark methods with regard to SKR.

eess.SP↗

arXiv2022

Joint Precoding and Phase Shift Design in Reconfigurable Intelligent Surfaces-Assisted Secret Key Generation

Key generation is a promising technique to establish symmetric keys between resource-constrained legitimate users. However, key generation suffers from low secret key rate (SKR) in harsh environments where channel randomness is limited. To address the problem, reconfigurable intelligent surfaces (RISs) are introduced to reshape the channels by controlling massive reflecting elements, which can provide more channel diversity. In this paper, we design a channel probing protocol to fully extract the randomness from the cascaded channel, i.e., the channel through reflecting elements. We derive the analytical expressions of SKR and design a water-filling algorithm based on the Karush-Kuhn-Tucker (KKT) conditions to find the upper bound. To find the optimal precoding and phase shift matrices, we propose an algorithm based on the Grassmann manifold optimization methods. The system is evaluated in terms of SKR, bit disagreement rate (BDR) and randomness. Simulation results show that our protocols significantly improve the SKR as compared to existing protocol.

cs.IT↗

arXiv2021

Random Embeddings and Linear Regression can Predict Protein Function

Large self-supervised models pretrained on millions of protein sequences have recently gained popularity in generating embeddings of protein sequences for protein function prediction. However, the absence of random baselines makes it difficult to conclude whether pretraining has learned useful information for protein function prediction. Here we show that one-hot encoding and random embeddings, both of which do not require any pretraining, are strong baselines for protein function prediction across 14 diverse sequence-to-function tasks.

q-bio.BM↗