arXiv subjects

Andrew Arash Mahyari

Publications and source records attributed to Andrew Arash Mahyari.

5 recordsLinked to original sources

arXiv2026

Adapter-Based Few-Shot Continual Learning for Malicious Packet Recognition

The continual evolution of malware variants necessitates detection systems that can adapt to new threats without retraining from scratch. However, continually updating models on new data often leads to catastrophic forgetting, where previously learned knowledge is overwritten. While continual learning has been increasingly explored for malware detection, the specific setting of Few-Shot Class-Incremental Learning (FSCIL), where new malware classes must be learned from only a small number of labeled examples, remains comparatively underexplored. Therefore, this work investigates the FSCIL setting for malware classification. To address the stability-plasticity dilemma, we propose a hybrid framework that leverages a Self-Supervised Learning (SSL) backbone initialized through domain-specific pre-training on malware packets. Our method incorporates Low-Rank Adaptation (LoRA) to efficiently adapt the model during the base session while freezing the core backbone to preserve previously learned representations, alongside a prototype-based classification head for incremental sessions to establish robust decision boundaries from limited samples. Extensive experiments across several datasets demonstrate that our approach consistently outperforms prior malware FSCIL baselines and achieves state-of-the-art performance.

cs.CR↗

arXiv2026

Curvature-Guided Module Localization for Low-Rank Detoxification of Backdoored Large Language Models

Backdoor attacks pose a serious threat to large language models (LLMs) by causing otherwise benign systems to produce attacker-specified malicious behavior when a hidden trigger is present. In this work, we study post hoc detoxification of backdoored LLMs in a practical setting where the defender has access to the poisoned model but does not wish to retrain the full network from scratch. We propose a mechanistically guided weight-space repair framework that first localizes modules involved in propagating trigger-induced behavior using activation patching and Fisher/K-FAC curvature analysis, and then applies targeted low-rank repair to only the most influential modules. We evaluate the method on poisoned variants of \texttt{Llama-3.2-1B-Instruct} with triggers inserted at the beginning, middle, and end of otherwise benign prompts. Results show that the proposed approach substantially suppresses trigger-conditioned malicious responses while preserving benign model behavior. These findings suggest that backdoor removal in LLMs can be formulated as a localized structural repair problem rather than only a broad behavioral alignment problem.

cs.CR↗

arXiv2025

Adaptive Additive Parameter Updates of Vision Transformers for Few-Shot Continual Learning

Integrating new class information without losing previously acquired knowledge remains a central challenge in artificial intelligence, often referred to as catastrophic forgetting. Few-shot class incremental learning (FSCIL) addresses this by first training a model on a robust dataset of base classes and then incrementally adapting it in successive sessions using only a few labeled examples per novel class. However, this approach is prone to overfitting on the limited new data, which can compromise overall performance and exacerbate forgetting. In this work, we propose a simple yet effective novel FSCIL framework that leverages a frozen Vision Transformer (ViT) backbone augmented with parameter-efficient additive updates. Our approach freezes the pre-trained ViT parameters and selectively injects trainable weights into the self-attention modules via an additive update mechanism. This design updates only a small subset of parameters to accommodate new classes without sacrificing the representations learned during the base session. By fine-tuning a limited number of parameters, our method preserves the generalizable features in the frozen ViT while reducing the risk of overfitting. Furthermore, as most parameters remain fixed, the model avoids overwriting previously learned knowledge when small novel data batches are introduced. Extensive experiments on benchmark datasets demonstrate that our approach yields state-of-the-art performance compared to baseline FSCIL methods.

cs.CV↗

arXiv2025

Proactive Adversarial Defense: Harnessing Prompt Tuning in Vision-Language Models to Detect Unseen Backdoored Images

Backdoor attacks pose a critical threat by embedding hidden triggers into inputs, causing models to misclassify them into target labels. While extensive research has focused on mitigating these attacks in object recognition models through weight fine-tuning, much less attention has been given to detecting backdoored samples directly. Given the vast datasets used in training, manual inspection for backdoor triggers is impractical, and even state-of-the-art defense mechanisms fail to fully neutralize their impact. To address this gap, we introduce a groundbreaking method to detect unseen backdoored images during both training and inference. Leveraging the transformative success of prompt tuning in Vision Language Models (VLMs), our approach trains learnable text prompts to differentiate clean images from those with hidden backdoor triggers. Experiments demonstrate the exceptional efficacy of this method, achieving an impressive average accuracy of 86% across two renowned datasets for detecting unseen backdoor triggers, establishing a new standard in backdoor defense.

cs.CV↗

arXiv2024

Toward a Real-Time Digital Twin Framework for Infection Mitigation During Air Travel

Pedestrian dynamics simulates the fine-scaled trajectories of individuals in a crowd. It has been used to suggest public health interventions to reduce infection risk in important components of air travel, such as during boarding and in airport security lines. Due to inherent variability in human behavior, it is difficult to generalize simulation results to new geographic, cultural, or temporal contexts. A digital twin, relying on real-time data, such as video feeds, can resolve this limitation. This paper addresses the following critical gaps in knowledge required for a digital twin. (1) Pedestrian dynamics models currently lack accurate representations of collision avoidance behavior when two moving pedestrians try to avoid collisions. (2) It is not known whether data assimilation techniques designed for physical systems are effective for pedestrian dynamics. We address the first limitation by training a model with data from offline video feeds of collision avoidance to simulate these trajectories realistically, using symbolic regression to identify unknown functional forms. We address the second limitation by showing that pedestrian dynamics with data assimilation can predict pedestrian trajectories with sufficient accuracy. These results promise to enable the development of a digital twin for pedestrian movement in airports that can help with real-time crowd management to reduce health risks.

cs.CY↗