Abstract As semiconductor manufacturing becomes increasingly outsourced to untrusted entities, Hardware Trojan (HT) attacks pose a critical threat to the security and reliability of modern integrated circuits. Machine learning models have improved the effectiveness of HT detection using Ring Oscillator Network (RON) side-channel data, yet recent work shows that these models are highly vulnerable to adversarial attacks. This paper evaluates the robustness of the Support Vector Machine (SVM) classifier, a leading algorithm in state-of-the-art HT detection frameworks, under gradient-based adversarial attacks. The proposed work demonstrates that high nominal accuracy does not ensure security against these attacks, which can reduce recall to zero. To strengthen resilience, three data-augmentation methods are investigated: SMOTE, Conditional Tabular Generative Adversarial Network (CTGAN), and Tabular Variational Autoencoder (TVAE). TVAE produces high-fidelity synthetic samples and substantially improves robustness, maintaining over 91% accuracy for nominal performance and over 88% accuracy under strong adversarial perturbations that cause a 100% attack success rate for the surrogate model. The results highlight the need to reframe hardware security evaluations beyond nominal accuracy toward adversarial robustness.

Aerial imagery provides crucial insights for various fields, including remote monitoring, environmental assessment, and autonomous navigation. However, the availability of aerial image datasets is limited due to privacy concerns and imbalanced data distribution, impeding the development of robust deep learning models. Recent advancements in text-guided image synthesis offer a promising approach to enrich and diversify these datasets. Despite progress, existing generative models face challenges in synthesizing realistic aerial images due to the lack of paired text-aerial datasets, the complexity of densely packed objects, and the limitations of modeling object relationships. In this paper, we introduce AeroDiffusion, a novel framework designed to overcome these challenges by leveraging large language models (LLMs) for keypoint-aware text description generation and a feature-augmented diffusion process for realistic image synthesis. Our approach integrates region-level feature extraction to preserve small objects and multimodal feature alignment to improve textual descriptions of complex aerial scenes. AeroDiffusion is the first to extend deep generative models for high-resolution, text-guided aerial image generation, including the creation of images from novel viewpoints. We contribute a new paired text-aerial image dataset and demonstrate the effectiveness of our model, achieving an FID score of 78.15 across five benchmarks, significantly outperforming state-of-the-art models such as DDPM (217.95), Stable Diffusion (119.13), and ARLDM (111.59).

Hardware Trojans are deliberate malicious hardware modifications inserted in semiconductor Integrated Circuits (ICs) for the purpose of stealing or leaking sensitive information, as well as disrupting critical systems upon activation. Emerging hardware security research highlights the criticality of employing AI for effective detection within the semiconductor IC supply chain. The efficient detection of these malicious Trojan circuits is of utmost significance, as it holds paramount importance in cultivating trust within the semiconductor IC supply chain. However, prevailing detection methodologies, predominantly reliant on Side-Channel Analysis (SCA), often necessitate the utilization of golden chips for validation. This article heralds a new era in hardware Trojan detection, harnessing the prowess of unsupervised machine learning in conjunction with SCA to eliminate the need for golden data. Employing unsupervised clustering, the methodology not only showcased a superior false-positive rate but also demonstrated a comparable accuracy level when compared to supervised counterparts. Notably, the proposed model exhibited an impressive accuracy rate of 93%, particularly excelling in pinpointing diminutive Trojans triggered by concise events, surpassing the capabilities of preceding techniques. In conclusion, this research advances a paradigm in hardware Trojan detection, emphasizing its potential in enhancing the integrity of semiconductor IC supply chains.

Cyber–physical systems (CPSs) have developed rapidly in recent years, contributing to an efficient integration between the cyber and physical worlds in intelligent and connected city environments. However, efficient mobility in a CPS is not well solved. Here, we present a prototype for a privacy-aware secure human-centric mobility-aware (SHM) model proposed and tested to analyze physical and human domains in IoT-based wireless sensor networks (WSNs). The proposed SHM model involves five modules: 1) sensor advertisements; 2) mobile sensor recruitment; 3) load balancing; 4) transmission guarantee; and 5) privacy with data-sharing phases. The proposed model is also validated using an accurate testing method that involves software and hardware tools and mathematical modeling to confirm secure communication. The model provides a tradeoff between energy efficiency and Quality-of-Service (QoS) requirements and compares the performance with other known models/protocols. Our testing process continued for four days, demonstrating that the SHM model provides compelling features of a secure CPS based on actual testing results. In practice, our model can be used in hospitals, as evident from validation in a real-life environment following the protocols.

Spatial-temporal charging coordination mechanisms are developed to avoid electrical overload at the charging stations and extravagant waiting time for electric vehicle drivers. Though, attackers could attack these mechanisms by launching distributed attacks against charging stations to prevent legitimate drivers from charging their vehicles. To attack a charging station, an attacker can compromise a set of vehicles, e.g., by disseminating a malware, and instruct them to send fake charging requests simultaneously to reserve the available energy capacity that is provided to a charging station without having the intention for charging, and thus benign vehicles do not find charging slots. This paper introduces an anomaly-based detection technique to identify the charging stations under this denial of charge (DoC) attacks using convolutional neural networks. The main idea is that each charging station has a normal energy demand pattern and launching DoC attacks changes this pattern. To capture such anomalous pattern, we use convolutional neural model to capture the temporal features within the demand of the charging station. To train our anomaly detector, we first create a benign dataset that could be utilized in other research areas such as load forecast and energy management. Then, we introduce a group of attacks that are used to create the malicious dataset. Finally, we used the benign and malicious datasets to train and test the deep neural model to detect DoC attacks. Our experiments show that our detector has high detection and low false alarm rates.

Advancement in technology improved living standard. Several known and unknown threats are handled using emerging technology. However, burglary threat is not fully addressed. In this paper, we introduce burglary detection (BD) framework to reduce the house thievery crime rate. BD involves secure home application, and framework. Secure home application involves two modes: protected and unprotected. Protected mode is enabled when people are at property (e.g. home, apartment). Unprotected mode is initiated when people are not in property. In any illegitimate person tries to enter the property, the signals are generated and sent to the owners, this process helps capture the illegitimate person. The proposed BD framework is implemented using Arduino, Java platform and Android. BD is tested and obtained desired results.