Digital Commons Network^™

Multi-Agent Ai For Automated Creation And Quality Assurance Of Cloud Incident Response Playbooks, Nitin Garg, Kunwar Vikas, Lakshmipriya Baskar, Ranjeet Ranjan, Anusha Bansal, Heer Mehta, Varun Bhardwaj, Savit Gupta

Defensive Publications Series

A trusted partner cloud (TPC) is a cloud deployment model that addresses the sovereignty requirements of governments by building isolated clouds operated by local partners and powered by a commercial cloud services provider. High-quality incident response playbooks are critical to successful TPCs. However, TPC partner operators may have limited skills and tools compared to the commercial cloud services provider. This disclosure describes multi-agent artificial intelligence (AI) techniques for the automated creation and quality assurance of TPC incident-response playbooks. A playbook-creation agent automatically converts playbooks internal to the commercial cloud-services provider to a high-quality, standardized format required for TPC operators. An …

System For Llm-Driven Orchestration And Live Migration Of Isolated Software Agents, Johannes Start, John Lunney

Defensive Publications Series

Systems for managing autonomous software agents in distributed environments can be limited by a lack of runtime mobility, rigid communication protocols, and poor observability. An integrated platform for executing software agents can combine isolated agent execution environments for fault containment with a live agent migration module for stateful mobility between host computing devices, such as physical servers or virtual machines. A central communication fabric, which can utilize a large language model (LLM), may mediate inter-agent communication. The LLM can analyze the semantic intent of a request to dynamically generate and execute a multi-step orchestration plan. This approach can support resilient …

Real-Time Preferred Agent Availability Predictions For Enhanced Contact Center Routing, Rajarathinam Chidambaram, Jibin George, Gururaj B T, Riyad Ottayil, Anil Kumar M L

Defensive Publications Series

A novel system and techniques are presented herein for intelligent contact center call routing that leverages a multi-factor, real-time predictive model using live transcript analysis, call complexity, agent historic performance, and scheduling data. When a customer is waiting and their preferred agent is predicted to become available imminently, the system dynamically calculates and applies an optimal wait extension to enable a more meaningful and efficient connection. This approach optimizes routing by combining agent availability forecasting with customer preference and contextual data, resulting in improved call resolution, reduced average handling time, and enhanced customer satisfaction.

Llm-Powered Agents For Context-Aware Validation Of Code And Configurations, Peng Chen

Defensive Publications Series

Enforcing complex, team-specific software development rules that are beyond the scope of some static analysis tools can present challenges, and manual review for such rules may be slow and error-prone. The described techniques may address these challenges by employing automated agents powered by large language models within a code review workflow. These agents can be configured with natural language prompts that codify domain-specific knowledge and complex validation logic. When triggered by a code change, an agent can analyze the modification and its relationship to other relevant files in the codebase. This method can provide automated, inline feedback on adherence to …

Time-Series Forecasting Using Dynamic Covariates From Text-Based Event Extraction, Leonid Kuligin, Thomas Zettl

Defensive Publications Series

Computer-implemented time-series forecasting systems, such as those executing on a server or distributed computing platform, may have difficulty anticipating demand volatility caused by real-world events due to a potential reliance on historical quantitative data. This disclosure relates to systems and methods for adjusting short-term forecasting by processing unstructured text data, for instance from news articles, to identify and quantify events that could influence future demand. A data processing pipeline can use a large language model to extract event descriptions, generate numerical embeddings for those events, and aggregate them into structured features. These features can then be integrated as dynamic covariates …

The Generative Asset Pipeline: "Live Code Assets" And Local Inference Orchestration, Matthew D. Judson

Defensive Publications Series

Disclosed is a Generative Asset Pipeline for NLE environments. The core innovation is the definition of "Live Code Assets"—executable code modules (e.g., React components, R3F scenes, .tsx files) treated as primary media primitives on the timeline. The NLE rendering engine includes a runtime for executing these modules. The system features "Dynamic Module Loading and Automatic Control Surfaces": it inspects the code asset (e.g., via Zod schema extraction) to identify parameters and automatically generates a visual control interface (Inspector Panel), allowing non-coders to manipulate the asset. The pipeline integrates local AI inference engines (e.g., ComfyUI) via the "Flow Workspace," orchestrated as …

Project Chimera: Real-Time Disaster Resilience Simulation Platform (Digital Twin), Matthew D. Judson

Defensive Publications Series

Disclosed is Project CHIMERA, a real-time simulation platform (Digital Twin) for testing the resilience of decentralized infrastructure (DePIN) during disaster scenarios.10 The integrated simulation engine combines high-fidelity geospatial data (DEM) with real-time environmental simulations (e.g., hurricanes) and infrastructure failures (power outages). The core innovation is the high-fidelity simulation of decentralized network dynamics (e.g., ChurchMesh/LoRa), including accurate RF propagation physics based on terrain, node failures, and dynamic message routing. The platform functions as a strategy evaluation tool, scoring infrastructure deployment strategies based on Coverage, Uptime, and Cost Efficiency within the simulated disaster.

Ktkn-Hub: Decentralized Git Hosting Via Ktkn-Os/Automerge And Isomorphic-Git, Matthew D. Judson

Defensive Publications Series

Disclosed is KTKN-HUB, a peer-to-peer (P2P) Git hosting platform utilizing decentralized storage (KTKN-OS) and CRDTs. The core innovation is the direct mapping of the Git object model into the KTKN-OS data fabric. Git objects (Commits, Trees, Blobs) are stored as immutable KTKN-OS documents. Crucially, Git References (Branches, Tags) are implemented as mutable CRDTs (Conflict-Free Replicated Data Types). This enables concurrent updates (pushes) without conflicts, guaranteeing mathematical convergence when peers reconnect. The system implements a novel filesystem adapter (KTKN-FS) that bridges the isomorphic-git library to the KTKN-OS backend, enabling Git operations without a traditional POSIX filesystem.

Deterministic Data Synthesis Pipeline For Multimodal Ai Training (The Trueos-Ktkn Data Engine), Matthew D. Judson

Defensive Publications Series

Disclosed is a system for generating "Golden Master" training datasets for Multimodal Vision-Language-Action (VLA) models. The core innovation is a "Deterministic Render-Capture Loop" that decouples simulation time (simTime) from wall-clock time. The system advances the simulation by a fixed delta ($\Delta t$), renders the state, and captures the frame using WebCodecs, timestamping it with simTime regardless of actual render duration. This guarantees a mathematically perfect framerate and perfect synchronization between video frames and telemetry logs, eliminating the frame drops and jitter inherent in real-time screen recording. The pipeline is orchestrated within the TrueOS/Electron shell, managing the simulation backend (e.g., Python …

The Porcupine Protocol: Automated Defensive Publication And Algorithmic Jurisprudence, Matthew D. Judson

Defensive Publications Series

Disclosed is the Porcupine Protocol, an automated system for generating and publishing industrial-grade defensive prior art using Local LLMs. The core innovation is the "Algorithmic Jurisprudence" engine (the "Recipe" algorithm), which automates the strategic decision to "File vs. Publish vs. Refactor." The algorithm calculates a Publication Score ($S_{Pub}$) based on weighted variables: Novelty, Detectability, Incumbent Risk, and Abstractness. It applies filters grounded in legal precedents: the "Alice" filter (Alice Corp. v. CLS Bank) assesses abstractness; the "Akamai" filter (Akamai v. Limelight) assesses divided infringement defense potential; and the "Porcupine" trigger initiates publication if $S_{Pub}$ exceeds a threshold. The Local LLM …

A Forensic Logging Framework For Llms, Giannis Tziakouris, Elio Biasiotto

Defensive Publications Series

Large language models (LLMs) are increasingly embedded in enterprise, operational, and security-sensitive workflows. These include automated coding assistants, knowledge management systems, and multi-agent orchestration platforms where LLMs act as intermediaries between users and external tools. While significant research has been directed toward preventive safeguards such as input filtering, refusal mechanisms, and adversarial robustness, far less attention has been paid to forensic readiness. When incidents occur, ranging from malicious prompt injections to unauthorized tool usage or exfiltration of sensitive data, existing LLM implementations fail to provide investigators with reliable records for reconstruction, containment, or attribution.

To address this deficiency, a comprehensive …

Scan To Email Summary, Hp Inc

Defensive Publications Series

This innovation focuses on the automated use of imaging technologies to simplify a well-known user experience that is frustrating to many customers today - the ability to scan to email, and then easily retrieve the scanned content. Providing an automated, simplified experience that results in significantly improved customer satisfaction, being able to search and retrieve emails that they have scanned on a multi-function device through their regular email client. The system leverages Large Language Models (LLMs) with engineered prompts to generate contextually relevant filenames, titles, and summaries for scanned documents.

Adaptive Fast Charging For Hearable Devices Based On State Of Charge, Yao Ding, Babak Hashemizadeh

Defensive Publications Series

Disclosed is an adaptive, conditional fast-charging method for hearable devices. In the method, a high charge rate (C-rate) is applied only when the battery state of charge (SOC) is below a low threshold (e.g., 10-30%) and desired conditions (e.g., temperature, age) are met. Further, charging immediately reverts to a standard rate once an SOC target or a temperature limit is reached. This selective, conditional approach provides a rapid initial charge that significantly enhances the user experience, reduces the risk of lithium plating, and avoids the cost and size constraints associated with batteries designed for sustained high C-rates.

The Master Engineering Compliance Atlas: A Unified Architecture For Automating Global Regulatory Governance, Ai Safety, And Cyber Risk, Ramkumar Bharathan

Defensive Publications Series

This technical disclosure presents the "Master Engineering Compliance Atlas," a comprehensive framework designed to bridge the gap between complex global regulations and modern software engineering practices. As regulatory landscapes fragment across jurisdictions, this document provides a unified methodology for harmonizing requirements from major frameworks, including Sarbanes-Oxley (SOX), GDPR, the EU Digital Markets Act (DMA), the Digital Services Act (DSA), the EU AI Act, NIST SP 800-53, and ISO 27001.

Unlike traditional legal summaries, this framework focuses on technical implementation, offering specific architectural patterns, "L7" control designs, and automation strategies to achieve "Compliance as Code." It introduces a "Mega-Matrix" that maps …

Generating A Personalized Artificial Intelligence Assistant For Domain-Specific Professional Work, Jan - Marten Ahrend

Defensive Publications Series

Professionals in specialized fields rely on their deep, domain-specific understanding that includes documented and undocumented knowledge such as project history, team dynamics, product goals, established communication styles, etc. Artificial intelligence (AI) assistants that support professionals may not capture the deep context specific to the domain, brand, and/or organization of the professional. This disclosure describes techniques for generating a personalized artificial intelligence (AI) assistant that is deeply contextualized to a specific professional’s unique work environment. Designed for domain-specific research and professional work, the AI assistant deploys a structured framework that enables the professional user to integrate their unique context and professional …

Real-Time Deepfake Detection During Video Conferences: Harnessing Bidirectional Audio-Video Features From Reflections And Echoes, Mahesh Godavarti, Ali Mouline, Kamil Wojcicki, Radhakrishna Achanta, Raul Casas

Defensive Publications Series

Liveliness detection is increasingly important in video conferencing due to the advent of deepfakes and their use in conducting online fraud. Currently, liveliness detection technologies lag behind deepfake creation technologies. This submission proposes the use of the bidirectional nature of video conferencing to counter deepfakes more reliably and more efficiently than the state-of-the-art.

Pet-Aware Navigation Using Aggregated And Verified Point-Of-Interest Data, Kshitij Suresh Gaikar, Pushkar Mittal

Defensive Publications Series

Digital mapping applications may lack integrated data on pet policies for points of interest (POIs), which can lead users to consult fragmented and potentially inaccurate external sources when planning travel with pets. A system and method are described for integrating pet-compatibility data into a navigation application. The system can utilize a multi-source data engine to aggregate and harmonize information from sources such as crowdsourced user reports, third-party data providers, and artificial intelligence analysis of public web data. A harmonization framework may use confidence scoring and a rules engine to resolve data conflicts and verify the pet policies of POIs, such …

Network Of Ai-Agents For Multi-Step Iterative Debugging, Akshay Jajoo, Timo Koehler, Josh Halley, Ramana Kompella, Myungjin Lee

Defensive Publications Series

Proposed herein is a system to facilitate fault detection and root cause analysis (RCA) in distributed systems using a network of artificial intelligence (AI) agents that can work together to perform distributed debugging. The proposed system employs a hybrid architecture of centralized and localized AI agents, each co-located with network devices, enabling near-real-time, context-aware analysis of logs, traces, ring-buffers and other telemetry at the point of generation.

Thematic Consolidation Of Multi-Turn Conversational Inputs Into A Single-Turn Prompt, Colby Hawker, Jennifer Otitigbe

Defensive Publications Series

In generative artificial intelligence application development, manually consolidating iterative, multi-turn conversational interactions into a single, static prompt for production use can be a time-consuming and error-prone process. Systems and methods are described that may automate this conversion. The technology can programmatically analyze a conversational history for thematic coherence. Based on the analysis, a consolidation module may apply conditional logic, for instance, by structuring coherent turns into a few-shot prompt or pruning less relevant turns to help isolate a user's intent. The resulting prompt can be scored for quality and optionally optimized by a language model. This process can transform an …

Automated Feature Engineering And Multi-Stage Selection For Time-Series Forecasting, Leonid Kuligin, Thomas Zettl

Defensive Publications Series

The time-series forecasting, for example in supply chain management, can be challenged by high-dimensional and noisy market data, which may lead to model overfitting and the inclusion of spurious correlations. A framework for automated feature engineering and multi-stage selection can help address these challenges. The process can programmatically generate a large set of potential features, such as time-lagged variables, from heterogeneous data sources. A multi-stage selection protocol can then be applied, which may first use correlation-based filtering to prune features, followed by an iterative, model-driven stage that uses backtesting to evaluate and promote features based on their predictive value. This …

A System For Brand-Sponsored Media Subscriptions With A Themed User Interface And Exclusive Ad Delivery, Caroline Pieracci

Defensive Publications Series

Digital media platforms may utilize ad-supported models that can present users with a high volume of advertisements, or subscription models that can include a price barrier. This disclosure describes a system for brand-sponsored platform subscriptions that can function as a hybrid monetization model. A method may allow a user to opt into a sponsorship from a single brand or a curated bundle of brands to receive access to platform features, such as those associated with a paid subscription, at a reduced or no monetary cost. In exchange, the user's platform experience can be modified for a defined period. This modification …

Cloud-Managed System For Distributing Audio Content To Point-Of-Sale Devices, Ramprasad S, Gulmohar Khan, Ankul Batra, Gajen Agrawal, Malay Singh, Saurabh Mimani, Ajay Prasad, Venu Annamraju, Sailendra Bathi

Defensive Publications Series

The utility for merchants of point-of-sale (POS) audio devices can be expanded beyond announcing transaction details. A cloud-managed system can facilitate the creation and distribution of custom audio content to these devices. For example, a user can input text through a client application, which can be sent to a cloud backend. The backend can use services, such as generative artificial intelligence for content refinement and a text-to-speech engine, to generate an audio file. The backend can then orchestrate the distribution of the audio file and an associated playback schedule to one or more endpoint audio devices. This system can expand …

Automatically Displaying Contextually Relevant Information In Response To Gesture Input On Displayed Content, Cliff Kuang

Defensive Publications Series

Obtaining information related to content viewed from a device currently requires users to engage in burdensome actions such as asking questions to an AI assistant or specifying a search query, both of which require the user to provide the context for their query. This disclosure describes a simple gesture-based user interface that helps users satisfy information needs from content they view. When the user performs a gesture, a prompt is automatically provided to a large language model (LLM) to generate relevant answers. With user permission, the LLM uses the user's current content and profile information to create personalized answers and/or …

A Comprehensive Sox And Icfr Architecture For Digital Asset Custodians, Exchanges, And Tokenization Platforms, Ramkumar Bharathan

Defensive Publications Series

This work presents a comprehensive, technology-agnostic SOX (Sarbanes-Oxley) and ICFR (Internal Controls over Financial Reporting) framework tailored for digital asset, blockchain, and Web3 enterprises. The rapid expansion of decentralized technologies and digital custody models introduces new risks that traditional financial control programs are not designed to address. This paper provides an end-to-end, multi-layer control architecture that integrates modern cybersecurity practices, blockchain-specific controls, and regulatory expectations across COSO, PCAOB, NIST, and evolving digital asset guidance.

The framework includes entity-level governance, digital asset custody controls, key-management procedures, multi-signature governance, smart-contract change management, stablecoin and token issuance safeguards, exchange integrity controls, and automated …

Mag-V: A Multi-Agent Framework For Synthetic Data Generation And Verification, Abhinav Mathur, Akshay Mallipeddi, Joseph Ross, Kristal Curtis, Liang Gou, Harsh Vashistha, Saptarshi Sengupta

Defensive Publications Series

Extending the capabilities of Large Language Models (LLMs) with functions or tools for environmental interaction has led to the emergence of the agent paradigm. In industry, training an LLM is not always feasible because of the scarcity of domain data, legal holds on proprietary customer data, rapidly changing business requirements, and the need to prototype new assistants. Artificial Intelligence (AI) agents can address such issues by relying on the zero-shot reasoning abilities of the underlying LLM and utilizing tools to explore and reason over customer data and respond to user requests. However, there are two concerns here: (I) acquiring large …

Distributed Security Orchestration With Adaptive Filtering And Data Minimization, Radhika Ravindranath, James Black, Jacques Mouton

Defensive Publications Series

Traditional security analytics, which relies on centralized data collection and processing, can raise concerns relating to privacy, data-transmission costs, threat-response latency, and jurisdictional data-handling requirements. This disclosure describes techniques to perform security analytics across a distributed network of edge computing nodes. By leveraging edge computing to collect and analyze data from endpoint devices, data collection/transmission is optimized and privacy is enhanced. Centralized processing via federated/distributed processing is limited. Adaptive filtering is used to dynamically adjust data processing and transmission based on detected events and risk levels. A tiered zoning approach processes high-risk events centrally, quarantines moderate-risk events for further analysis …

Generating A Granular Priority Score For Computational Workloads From Multiple Parameters, John Lunney, Johannes Start

Defensive Publications Series

Systems for managing computational workloads on shared hardware resources, such as graphics processing units or other accelerators, may utilize a limited number of coarse-grained priority levels. This approach may not sufficiently differentiate between a wide spectrum of competing tasks, potentially leading to sub-optimal resource allocation. A described system can generate a granular, numerical priority score for a workload using a computational component, for instance, a multi-dimensional calculator. This component can process multiple input parameters associated with a request, such as a user's service tier, the request's latency sensitivity, and its feature modality, to compute a synthesized priority index. This granular …

Advertising Targeting Based On Multi-Modal Detection Of Organic Brand Presence, Basel Hijazi, Nicolas Marchal

Defensive Publications Series

A potential difficulty in digital advertising can be measuring or capitalizing on the organic exposure of physical brand sponsorships within online media content, as some systems might not analyze visual and audio content for brand presence. A multi-modal analysis engine can be used to process media content. This engine can, for example, employ computer vision to identify visual brand indicators, such as logos and products, and use automatic speech recognition and natural language processing to detect spoken brand mentions in an audio track. Detected instances of brand presence can then be converted into actionable advertising targeting criteria. Such criteria can …

Advertising Constraint-Based Dynamic Generation Of Promotional Content User Interfaces., Lakshmi Kumar Dabbiru, Raghu Donamukkala

Defensive Publications Series

The use of static, pre-defined layouts for promotional content can limit adaptability to user requests for specific formats and may not maintain visual coherence with evolving page designs. This disclosure describes a system for the dynamic, advertising constraint-based generation of user interfaces for such content. The system can utilize a generative large language model to create front-end code for a promotional layout in real-time or near-real-time. The generation process can be guided by a consolidated set of constraints, such as formatting instructions parsed from a user's query, design system parameters from a surrounding page, and advertiser brand preferences. This approach …

Federated Agentic System For Adaptive Workflow Management In Enterprise Environments, Soumendra Kumar Mishra

Defensive Publications Series

The present disclosure relates to a federated agentic system for adaptive workflow management in enterprise environments. This system features a coordination hub for dynamically delegating tasks to multiple intelligent agents. The Resource Usage module forecasts resource consumption patterns based on historical execution data, while the Server Availability module monitors server health to ensure infrastructure readiness. Additionally, the Resource Availability module assesses the state of the Hadoop cluster for necessary computing and storage resources, and the Scheduling Optimization module identifies and resolves scheduling conflicts to enhance job execution efficiency. By integrating these components, the system ensures reliable data synchronization across distributed …