Digital Commons Network^™

Hybrid Material Geometric Waveguide, Kellie Shaw

Defensive Publications Series

A hybrid-material waveguide including a plurality of fold mirrors formed from a first material, and an output mirror formed from a second material.

Selectively Suppress Evpn Routes (Rt1, Rt2 & Rt3) During Rt Mismatch In Evpn Multihomed Topology With Rtc, Anonymous

Defensive Publications Series

The present disclosure describes a method for selectively suppressing Ethernet Virtual Private Network (EVPN) route advertisements (specifically Route Types RT1, RT2, and RT3) during Route Target (RT) mismatches in EVPN multihomed topologies employing Route Target Constraints (RTC). The method involves differentiating RT advertisements based on Ethernet Segment (ES) identifiers by including ES extended community information within RTC updates. Upon detecting RT mismatches between multihomed Provider Edge (PE) routers, selective suppression is applied exclusively to certain non-essential EVPN routes (RT1 with VLAN tags, RT2 MAC/IP advertisements, and RT3 Inclusive Multicast routes). Critical EVPN routes necessary for Designated Forwarder (DF) election, aliasing, …

The Tech Audit Revolution: Navigating The Next 50 Years For Industry And Technology Leadership And An End-To-End Audit Framework For Ads, Search, And Content Platforms, Ramkumar Bharathan

Defensive Publications Series

The internal audit function stands at a pivotal juncture, undergoing a profound transformation from its traditional role in compliance to a strategic enabler of business value. For leading technology companies, the ability to safely deploy cutting-edge innovations is a core competitive differentiator, making a proactive and technologically advanced audit function essential for secure and rapid market entry. This paper presents a comprehensive, revolutionary audit framework designed to navigate the complex risk environment of the next five decades, addressing both current challenges and future technological frontiers. This framework provides a deep dive into auditing emerging technologies such as Quantum Computing, …

Exploring And Improving Initialization For Deep Graph Neural Networks: A Signal Propagation Perspective

Defensive Publications Series

Graph Neural Networks (GNNs) often suffer from performance degradation as the network depth increases. This paper addresses this issue by introducing initialization methods that enhance signal propagation (SP) within GNNs. We propose three key metrics for effective SP in GNNs: forward propagation, backward propagation, and graph embedding variation (GEV). While the first two metrics derive from classical SP theory, the third is specifically designed for GNNs. We theoretically demonstrate that a broad range of commonly used initialization methods for GNNs, which exhibit performance degradation with increasing depth, fail to control these three metrics simultaneously. To deal with this limitation, a …

Intelligent Test Orchestrator For Dynamic Test Sequencing, Na

Defensive Publications Series

In hyperscale data centers, the sheer volume of installed devices necessitates a substantial daily volume of diagnostic testing. Running all tests can be time-consuming, leading to longer repair times (MTTR) and impacting data center efficiency. This disclosure describes an intelligent test orchestrator (ITO) that leverages machine learning to dynamically sequence tests to rapidly detect failure without executing entire test suites. The ITO dynamically sequences tests based on component lists for each device, diagnostic test coverage information, component failure rates, repair records, test coverage, etc. The described intelligent test orchestrator reduces the time needed to restore repaired devices to the production …

Scalability Matters: Overcoming Challenges In Instructglm With Similarity-Degree-Based Sampling

Defensive Publications Series

Large Language Models (LLMs) have demonstrated strong capabilities in various natural language processing tasks; however, their application to graph-related problems remains limited, primarily due to scalability constraints and the absence of dedicated mechanisms for processing graph structures. Existing approaches predominantly integrate LLMs with Graph Neural Networks (GNNs), using GNNs as feature encoders or auxiliary components. However, directly encoding graph structures within LLMs has been underexplored, particularly in the context of large-scale graphs where token limitations hinder effective representation. To address these challenges, we propose SDM-InstructGLM, a novel instruction-tuned Graph Language Model (InstructGLM) framework that enhances scalability and efficiency without relying …

Llm-Based Navigation Guidance Revisions Based On Conversational Context, Florian Hartmann, Matthew Sharifi

Defensive Publications Series

When a vehicle has multiple occupants engaged in a conversation, audio instructions from a digital map/ navigation application can be disruptive and/or unnecessary. This disclosure describes techniques, implemented with specific user permission to access and analyze data related to the presence of multiple users in a vehicle and the conversational context. With user permission, the ongoing conversation in the vehicle is analyzed locally using a large language model and the analysis is used to enhance the audio delivery of dynamic navigational guidance. A transcript of the conversation related to the ongoing drive is provided to an LLM along with a …

Uncertainty Estimation For Heterophilic Graphs Through The Lens Of Information Theory

Defensive Publications Series

While uncertainty estimation for graphs recently gained traction, most methods rely on homophily and deteriorate in heterophilic settings. We address this by analyzing message passing neural networks from an information-theoretic perspec tive and developing a suitable analog to data processing inequality to quantify information throughout the model’s layers. In contrast to non-graph domains, information about the node level prediction target can increase with model depth if a node’s features are semantically differ ent from its neighbors. Therefore, on heterophilic graphs, the latent embeddings of an MPNN each provide different information about the data distri bution – different from homophilic settings. …

Wide & Deep Learning For Node Classification

Defensive Publications Series

Wide & Deep, a simple yet effective learning architecture for recommendation systems developed by Google, has had a significant impact in both academia and industry due to its com bination of the memorization ability of general ized linear models and the generalization ability of deep models. Graph convolutional net works (GCNs) remain dominant in node classification tasks; however, recent studies have high lighted issues such as heterophily and expressive ness, which focus on graph structure while seemingly neglecting the potential role of node features. In this paper, we propose a flexible frame work GCNIII, which leverages the Wide & Deep …

Alphafold Database Debiasing For Robust Inverse Folding

Defensive Publications Series

The AlphaFold Protein Structure Database (AFDB) offers unparalleled structural coverage at near-experimental accuracy, positioning it as a valuable resource for data-driven protein design. However, its direct use in training deep models that are sensitive to fine-grained atomic geometry—such as inverse folding—exposes a critical limitation. Comparative analysis of structural feature distributions reveals that AFDB structures exhibit distinct statistical regularities, reflecting a systematic geometric bias that deviates from the conformational diversity found in experimentally determined structures from the Protein Data Bank (PDB). While AFDB structures are cleaner and more idealized, PDB structures capture the intrinsic variability and physical realism essential for generalization …

Contextual Explanation Of Local Differences On Digital Map User Interface, Florian Hartmann, Matthew Sharifi

Defensive Publications Series

Different regions of the world have different regulations, e.g., related to driving, parking, etc. as well as different social norms, e.g., tipping, acceptable use of a mobile device on public transport, etc. Such differences can trip up travelers when they are in unfamiliar locations. This disclosure describes techniques to automatically determine via offline processing by a large language model (LLM). The differences are stored in a database and are used to surface proactive contextual alerts to users at appropriate times by explaining relevant differences between rules or norms at their current location and their typical location. The LLM can be …

Implicit Graph Neural Networks With Flexible Propagation Operators

Defensive Publications Series

Due to the capability to capture high-order information of nodes and reduce memory consumption, implicit graph neural networks have become an explored hotspot in recent years. However, these implicit graph neural networks are limited by the static topology, which makes it difficult to handle heterophilic graph-structured data. Furthermore, the existing methods inspired by optimization objectives are limited by the explicit structure of graph neural networks, which makes it difficult to set an appropriate number of network layers to solve optimization problems. To address these issues, we propose an implicit graph neural network with flexible propagation operators in this paper. From …

Balanced Graphformer: Enhancing Graph Transformers With Localized Training Against Over-Globalization

Defensive Publications Series

As Transformers gain traction in graph machine learning, the issue of "over-globalization" has emerged, where their global attention mechanisms excessively emphasize distant vertices, potentially diluting vital local information. This is particularly detrimental in graphs where local neighborhoods hold significant predictive value. Existing methods often lack flexibility in local processing or fail to effectively integrate local and global contexts. This paper introduces BalancedGraphFormer, a novel framework designed to localize graph transformer training. It integrates a dedicated local module with a complementary global module. The local module captures fine-grained neighborhood patterns, while the global module integrates broader context without overshadowing local details. …

Datacenter Planning And Rack Placement Optimization Using Reinforcement Learning, Na

Defensive Publications Series

This disclosure describes the use of reinforcement-learning (RL) techniques to optimize datacenter planning and rack placement. An RL agent operates within a state space that includes physical infrastructure (rack space, power, cooling, network), resource availability (machine inventory, backup power), constraints (emissions, workload demands), etc. The RL agent takes actions such as allocating capacity and placing machines. A reward function guides the agent towards optimal solutions by rewarding utilization, balancing load, achieving compliance with service-level objectives, and penalizing resource stranding and costs. RL algorithms balance exploration and exploitation, adapt to dynamic environments, and scale to large datacenters. Advantages of the described …

Robust Graph Learning Through Spatial-Spectral Synergy Against Structural Attacks

Defensive Publications Series

Graph Convolutional Networks (GCNs) are powerful tools for learning from graph data but exhibit significant vulnerability to adversarial structural attacks that manipulate node connections. While various defense strategies focusing independently on the spatial or spectral domains exist, they often fail to leverage the complementary strengths of both perspectives. This paper introduces the Spatial- Spectral Graph Convolutional Network (S²-GCN), a novel framework designed to enhance GCN robustness against structural attacks by synergistically combining spatial and spectral defense mechanisms. S²-GCN comprises two core GCN-based modules operating in parallel. The spectral module utilizes a graph structure derived from learnable low-frequency spectral components, adaptively …

Learning Resilient Graph Structures In Heterophilic Settings

Defensive Publications Series

Graphs provide a fundamental way to model relationships between entities and are central to numerous machine learning tasks. Standard graph-based methods often assume the provided graph structure is both accurate and complete. However, real-world graphs frequently suffer from noise and sparsity, negatively impacting downstream tasks like node classification and clustering. While graph representation learning has advanced significantly, many methods implicitly assume graph homophily (connections predominantly between nodes of the same class), struggling when faced with heterophily (connections predominantly between different classes). This paper introduces a novel method, Resilient Graph Learning for Heterophily (RGLH), designed to learn high-quality graph structures directly …

Halo: Heterophily-Aware Label-Free Ordering For Unsupervised Graph Fraud Detection

Defensive Publications Series

Graph fraud detection (GFD) has rapidly advanced in protecting online services by identifying malicious fraudsters. Recent supervised GFD research highlights that heterophilic connections between fraudsters and users can greatly impact detection performance, since fraudsters tend to camouflage themselves by building more connections to benign users. Despite the promising performance of supervised GFD methods, the reliance on labels limits their applications to unsupervised scenarios; Additionally, accurately capturing complex and diverse heterophily patterns without labels poses a further challenge. To fill the gap, we propose a Heterophily-guided Unsupervised Graph fraud dEtection approach (HUGE) for unsupervised GFD, which contains two essential components: a …

A Neuromorphic Spike-Based System For Real-Time Behavioral Anomaly Detection In Saas And Self-Hosted Environments, Marcel Witt

Defensive Publications Series

This publication introduces NeuraSentry™, a neuromorphic anomaly detection framework inspired by biological spike-based learning and synaptic plasticity. It combines competitive vector detectors, temporal spike traces, and retroactive learning signals to detect evolving behavioral fraud and insider threats in both SaaS and on-premise systems. Key innovations include sparse activation using cosine similarity, HDBSCAN-based meta-neuron clustering, and delayed reward modulation modeled after dopamine and serotonin systems. NeuraSentry delivers adaptive, explainable, and scalable behavioral threat detection, without relying on labeled training data or fixed rules. This work establishes prior art to prevent overly broad patents in the field of neuromorphic security analytics.

Functional Handle Taper And Nesting Mold For Dispenser - Compatible Cutlery, Northstar Maintenance Management Inc. Dba Compostitall.Com

Defensive Publications Series

This publication discloses the functional geometry of a tapered handle for compostable cutlery that enables consistent stacking and dispensing in tab-compatible or gravity-fed dispensers. This design has been in commercial use since approximately 2017 and is currently offered under the Forever Free brand. No claim of original invention is made; this disclosure is intended to serve as prior art.

Context-Aware Chatbot For Contextual Link Retrieval And Team Collaboration, Harish Murthy

Defensive Publications Series

Members of a team often individually bookmark or maintain knowledge assets, e.g., documents, meeting notes, etc. Sharing links to such assets and disseminating their contents within the team can become unwieldy, especially as the size of the team and the volume of assets grow. This disclosure describes techniques that enable individuals within a team to register certain mnemonics within a chat room or messaging app, such that when prompted with a mnemonic, a link to a contextually relevant asset is surfaced. In the case of chat rooms, context is derived from the chat room from which the query originated. In …

Key Management Server Driven Adaptive Client-Side Caching Of Cryptographic Keys, Himanshu Kishna Srivastava Mr.

Defensive Publications Series

In conventional cryptographic systems, encryption keys are typically retrieved from a centralized key management server (KMS). To reduce latency and improve performance, clients often cache these keys locally. While this caching mechanism enhances efficiency. However, this approach also introduces several security challenges e.g. cached keys may become outdated or exposed to unauthorized access, and clients generally lack a reliable mechanism to detect when a key has been updated or revoked by the KMS. Moreover, because each client may implement its own caching logic—governing how long keys are stored, when they are refreshed, or how they are secured—this results in inconsistent …

Space Socks: Quantum Entanglement Simulation Game, Suma Mallapragada, Celeste Gnits

Defensive Publications Series

Quantum teleportation and entanglement are foundational topics in quantum mechanics that are difficult to visualize and intuitively grasp. “Space Socks” is an interactive educational game that demystifies these quantum phenomena through playful simulation. Developed as part of the Celeste GNITS Space Research initiative, this game leverages quantum circuit logic and entangled metaphors to teach quantum teleportation, quantum internet, and entanglement swapping. This whitepaper outlines the scientific basis, game mechanics, implementation, and educational impact of the Space Socks project.

By publishing the work and its associated repository publicly, we are preventing the possibility of patent claims on the unique ideas …

Runtime Prediction Of Ai Model Operations Using A Gru-Based Neural Network, Suma Mallapragada

Defensive Publications Series

This work presents a defensive disclosure of a novel GRU-based pipeline to predict the execution runtime of AI model computational graphs on TPU hardware. Efficiently predicting the runtime of AI model operations is essential for optimizing deployment and hardware utilization. This paper proposes a novel pipeline that integrates opcode-based runtime estimation, graph edge de pendency embeddings, configurable node feature adjustments, and a Gated Recurrent Unit (GRU) neural network to predict operation runtimes on computational graphs. The approach is applied to the TPUGraphs dataset from the ”Google- Fast or Slow? Predict AI Model Runtime” Kaggle competition, which …

Systems And Methods For Training Frameworks To Optimize Model Based On Business Goals, Chinamy Narendra Sonar, Shuhan Zhang, Keyi Li, Lin Meng, Can Liu, Shubham Agrawal, Chiranjeet Chetia

Defensive Publications Series

[0001] The present disclosure provides a system and method for providing training frameworks to optimize model based on business goals. The processor of the training framework system introduces novel model evaluation metrics which are directly aligned with business goals and capture most business use cases for A2A fraud detection models. Metrics can be used in model training and evaluation. In the first approach, the processor of the training framework system sets validation criteria to a new metric but training the model with traditional loss. In the second approach, the processor of the training framework system trains and validates the model …

Context-Sensitive Self-Serve Platform For Cloud Computing Partners, Shashwat Anand, Rohan Chaudhury, Arvind Gangam

Defensive Publications Series

This disclosure describes a context-sensitive, self-serve platform that a cloud computing service can use to support its partners and stakeholders. The platform models business processes as policies and represents them as abstract syntax trees (AST), which can be parsed by an expression language to enhance partner insight and accessibility. The self-serve feature enables the platform to achieve a high degree of adaptability, e.g., the ability to rapidly respond to changes in the ecosystem, to quickly pinpoint and address areas of friction, to effectively adapt to the evolving needs of stakeholders, etc. Advantages include empowering non-technical users, bridging the gap to …

Video Discovery Api For Cross-Platform Content Recommendation, Salman Raza, Abhay Kumar Gupta, Hailin Wu, Weitong Liu, Trung Nguyen, Harshita Kasera, Venkata Kottapalli, Vivek Yadav, Jason Robertson, Vighnesh Raut, Vineet Kumar

Defensive Publications Series

Existing content discovery systems often require complex account linking processes or rely on device-specific data, creating fragmented user experiences across different platforms. Disclosed herein is a system and method for cross-platform content recommendation utilizing a Video Discovery API. This technology enables applications to securely transmit user activity data and content state information to a central service. The core operation involves inferring a relationship between third-party user identities and a platform’s internal user identity using device associations. This facilitates the seamless exchange and propagation of user data across various devices and platforms without explicit account linking. The primary purpose is to …

Generating Virtual Backgrounds Based On Event Streams, Andrew Mcglynn, Keith Griffin

Defensive Publications Series

Meetings/calling software often allows users to utilize virtual backgrounds within a call/meeting. Although such software allows users to be creative and express themselves through virtual backgrounds, this can lead to a lengthy search/generation of a suitable image to use as a virtual background and for the user to upload it to the calling/meetings software. This submission proposes a process for automating the generation of virtual backgrounds based on events that are relevant to a user/organization, such as birthdays, holidays, corporate events, sports events, etc. The proposed process can be integrated within the meetings/calling software, so that a user does not …

Adaptive Blockage Placement In Vlsi Design For Enhanced Utilization And Congestion Reduction, Pankaj Mudgil, Navin Dayani, Vipin Bargurjar, Arun Tyagi, Neil Malhotra

Defensive Publications Series

In VLSI chip design, congestion hotspots that emerge during place-and-route (PnR) pose a hurdle to routability. Poorly routed designs can lead to various issues such as excessive power consumption, signal integrity, timing violations, reliability, etc. This disclosure describes congestion-aware, adaptive techniques for the placement of blockages using gradient-based placement (GP) spirals and radial-diagonal scaling to enhance routability and wire utilization in VLSI design. Per the techniques, blockage placement is continually updated based on real-time routing feedback. The resulting iterative refinement enables a dynamic response to congestion shifts. Initial blockage placement is done using GP spirals and/or radial-diagonal utilization. Global routing …

Optimized Grouping For Packetized Atpg Scan Testing, Bhavika Ranjeet Kumar, Aalhad Deshpande, Rama Sireesha Arisetti, Maheedhar Jalasutram, Vasubabu Ravipati

Defensive Publications Series

Packetized scan testing utilizes a single scan network to efficiently test a large number of ATPG (automatic test-pattern generation) partitions and to enable concurrent testing across multiple partitions. However, concurrent testing can result in high operating current demand and local IR drops. This disclosure describes techniques of efficient ATPG partition grouping based on factors such as scan volume, physical proximity of the partitions, shared power rails, operating current requirements of the core, etc. The resulting groups have an optimal number of cores that avoid operating current overload and IR-drop without substantially impacting test time. A priority index algorithm minimizes power-rail …

Protecting Test Equipment From Current Surges Due To Non-Target Blocks, Vevekanenda Gonugunta, Ajaykumar Prajapati, Aditya Kota, Nischal Satyanarayan, Rajesh Gottumukkala

Defensive Publications Series

During automatic test pattern generation (ATPG) testing of integrated circuits, collateral clock propagation and high switching activity can occur in partitions not being tested, e.g., within the debug-and-system unit. This unintended activity can generate an excessive current draw that can cause excessive heat generation, potential damage to test probes and boards, reduced test safety, etc. This disclosure describes techniques to mitigate excessive current draw within the debug-and-system unit partition of a system-on-chip (SoC) being scan-tested using ATPG by augmenting the SoC with logic that gates and confines the propagation of the clock to blocks of the chip currently being tested. …