Graph Neural Networks

Alrahis, Lilas; Knechtel, Johann; Sinanoglu, Ozgur

doi:10.1145/3566097.3568345

Cited by 12 publications

(3 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(vi) Hardware security of ICs. To tackle the circuitrelated tasks related to hardware security, a generic end-toend GNN-based pipeline is suggested in [82]. Concerns like implantation hardware Trojans, piracy of design intellectual property, and reverse-engineering are highlighted.…”

Section: Relevant Applications Of Gnns In Circuit Designmentioning

confidence: 99%

From Graph Theory to Graph Neural Networks (GNNs): The Opportunities of GNNs in Power Electronics

Li,

Xue,

Zargari

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Graph theory within power electronics, developed over a 50-year span, is continually evolving, necessitating ongoing research endeavors. Facing with the never-been-seen explosion of graphstructured data, the state-of-the-art deep learning technique-Graph Neural Networks (GNNs), becomes the leading trend in machine learning within just recent five years and demonstrated surprisingly broad and prominent benefits covering from new drug discovery to better IC design. However, its promising applications in Power Electronics are still rarely discussed and its full potential remains unexplored. Addressing this gap, this review paper is the first to outline GNNs' general workflow in power electronics, laying the groundwork and examining current GNN methodologies within the field. To bridge the gap in the sparse GNN literature within this domain, we also provide extended discussions on leveraging insights from GNN-aided circuit design to enrich power electronics research. Our work includes in-depth GNNbased case studies that demonstrate promising applications from converters to system-level power electronics, showcasing GNNs' unique benefits and untapped possibilities (e.g., accurate component design, voltage predictions on IEEE-13 bus and 118 bus systems). Additionally, we provide a comprehensive survey of GNNs' latest and successful applications, emphasizing their impact on energy-centric sectors, such as transportation electrification, smart grids. Considering the interdisciplinary nature of power electronics in modern energy systems, our review highlights the potential of GNNs emerge as a promising tool to decode the intricate behavior and dynamics of power electronics systems, and we hope such synergies between advanced AI methodologies like GNNs with the ever-evolving graph theory can lead to more powerful tools, novel methodologies, and advancements in the power electronics community.

show abstract

Section: Relevant Applications Of Gnns In Circuit Designmentioning

confidence: 99%

From Graph Theory to Graph Neural Networks (GNNs): The Opportunities of GNNs in Power Electronics

Li,

Xue,

Zargari

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The user accepts the model if it meets target accuracy value known as clean accuracy. 4 The adversary manipulates D T rain by injecting the backdoor trigger into selected input samples (i.e., circuits) to build a backdoored model θ adv . The backdoored model should maintain performance on clean input samples (e.g., D T est ) to avoid detection by the user.…”

Section: Poisonedgnn Threat Modelmentioning

confidence: 99%

“…G RAPH neural networks (GNNs) have attracted considerable attention owing to their superior performance in graph-based learning applications [1], [2]. Researchers have successfully utilized GNNs for several electronic design automation (EDA) tasks, such as floorplanning optimization, estimating routing congestion, and assessing circuit reliability [3], [4], to name a few. The outstanding success of GNNs in EDA is primarily because Boolean circuits can be naturally represented as graphs.…”

Section: Introductionmentioning

confidence: 99%

Embracing Graph Neural Networks for Hardware Security

Alrahis

Patnaik²,

Shafique

et al. 2022

Proceedings of the 41st IEEE/ACM International Conference on Computer-Aided Design

View full text Add to dashboard Cite

Graph neural networks (GNNs) have shown great success in detecting intellectual property (IP) piracy and hardware Trojans (HTs). However, the machine learning community has demonstrated that GNNs are susceptible to data poisoning attacks, which result in GNNs performing abnormally on graphs with pre-defined backdoor triggers (realized using crafted subgraphs). Thus, it is imperative to ensure that the adoption of GNNs should not introduce security vulnerabilities in critical security frameworks. Existing backdoor attacks on GNNs generate random subgraphs with specific sizes/densities to act as backdoor triggers. However, for Boolean circuits, backdoor triggers cannot be randomized since the added structures should not affect the functionality of a design. We explore this threat and develop PoisonedGNN as the first backdoor attack on GNNs in the context of hardware design. We design and inject backdoor triggers into the register-transfer-or the gate-level representation of a given design without affecting the functionality to evade some GNN-based detection procedures. To demonstrate the effectiveness of PoisonedGNN, we consider two case studies: (i) Hiding HTs and (ii) IP piracy. Our experiments on TrustHub datasets demonstrate that PoisonedGNN can hide HTs and IP piracy from advanced GNN-based detection platforms with an attack success rate of up to 100%.

show abstract