Digital Commons Network^™

Multi-Signal Fusion Architecture For Semantic Leak Detection In Large Language Model Inputs, Kaustav Saha Saha, Shwetank Raghuvanshi, Batool Haider, Kliton Andrea, Chenhui Hu, Subramanian Srinivasan, Ashwin Kesireddy, Zscaler, Inc.

Defensive Publications Series

A system detects semantic leakage of contextual information in a prompt to a large language model (LLM) by extracting, in parallel from a reference context and a user input, a cosine similarity signal from multilingual sentence embeddings, an entity overlap signal from extracted named entities, and a number match signal from extracted numerical patterns. Fusion weights are selected from a plurality of language-tier weight profiles based on a detected language of the user input, and a fused score is computed as a weighted combination of the signals. The reference context and the user input are routed onto a fast trigger …

Sensor Fusion Method For In-Cabin Zonal Solar Load Estimation And Hvac Mode Control, Anonymous

Defensive Publications Series

Vehicle HVAC (Heating, Ventilation, and Air Conditioning) systems rely on solar load sensing to bias cabin cooling and maintain occupant thermal comfort. Conventional implementations use a single dash-mounted sunload sensor to estimate an aggregate solar heat load across the entire cabin, applying a uniform or, at best, a fixed left/right cooling bias. This approach does not account for the vehicle's actual heading relative to the sun, transient shading events, or differences in glass geometry and tint across occupant zones. This publication discloses a sensor fusion method that combines GPS-derived solar position geometry, vehicle heading, the existing single sunload sensor, and …

Virtual Sensing And Ai Method For In-Cabin Airflow (Q) Estimation And Closed-Loop Hvac Control, Anonymous

Defensive Publications Series

Vehicle HVAC blower control is typically open-loop, using commanded voltage or PWM duty cycle and assuming a fixed relationship to delivered airflow. In reality, airflow reaching the occupant depends on duct losses, filter loading, blower/compressor efficiency, altitude, air density, and vehicle variation. Since comfort depends on actual vent airflow (i.e., not blower voltage),this method estimates delivered airflow (Q = A·v, m³/s) through physics-based and AI sensor fusion using existing vehicle signals, then closes the control loop on estimated Q rather than commanded voltage.

Artificial Intelligence Driven Insights And Predictions For A Television Application, Neetesh Mittal, Vivek Kushwaha, Vivek Yadav, Saravanan Santhamoorthy Theckyam

Defensive Publications Series

Traditional data analytics for television applications rely on manual, dashboard-centric models that track historical user activity. These traditional data analytics for television applications fail to explain underlying user behaviors or to correlate multiple simultaneously launched features with core business objectives. To address these limitations, the disclosed technology describes an autonomous analytical system that utilizes a distributed client-server architecture that hosts a hierarchical multi-agent backend built on an orchestration framework. A supervisor agent coordinates specialized domain engines to execute targeted statistical, performance, and viewership analysis algorithms over read-only television application databases. Communication under strict End-User Credential (EUC) zero-trust security and under …

Visual Representation Of Ai Agents Replacing Inactive Users In Chat Interfaces, Ramprasad Sedou, Punit Pandia

Defensive Publications Series

Chat-based communication platforms use visual identity markers such as profile pictures and avatars to identify message sources. When users become temporarily or permanently unavailable, AI (artificial intelligence) agents may generate responses on their behalf. However, displaying the original user's avatar for AI-generated messages may mislead recipients into believing they are communicating with the human user, while replacing avatars with generic AI icons may obscure user identification in group settings. Disclosed are systems and methods for visually representing AI agent activity within chat interfaces through modified avatar presentations that combine elements of the original user's visual identity with indicators of AI …

Caldwell Cmhre-Ω1.2 Open-Hardware Modular Thermal Stabilization, Direct-Liquid-Cooling, And Waste-Heat Utilization Research Platform, Michael Victor Caldwell Mr.

Defensive Publications Series

The Caldwell CMHRE-Ω1.2 is a modular open-hardware research platform for evaluating direct liquid cooling, thermal stabilization, phase-change buffering, low-grade thermoelectric energy recovery, gyroid-inspired flow distribution, localized vortex/impingement heat-transfer enhancement, and adaptive thermal control.

CMHRE-Ω1.2 uses a controlled 100–500 W electrically metered copper heater block as a chip analog. A primary, conventional liquid-cooling loop is always available to remove the full thermal load safely. Optional research modules operate in bypassable side streams so no experimental energy-recovery module becomes a mandatory thermal bottleneck.

The platform incorporates selected practical elements from the posted CIERS-TEG v3.0 disclosure: detailed metering and control architecture, copper-foam-enhanced PCM, …

Systemic Progress Regeneration Protocol (Sprp), Volodymyr Kotegov

Defensive Publications Series

This protocol defines the mandatory rules for the creation, evaluation, acceptance, and long-term stewardship of engineering work by parties that voluntarily adopt it.

The protocol exists to counteract a fundamental process: the spontaneous decay of civilizationally critical engineering knowledge. Its objective is not the production of superior individual artifacts. Its objective is the sustained growth of the survivable mass of engineering knowledge — the quantity of knowledge that remains comprehensible, reproducible, and extensible after complete institutional rupture, generational turnover, and loss of external infrastructure.

The protocol claims no universal authority. It applies only to nodes and Inheritance Lines that have …

System And Method For Swarm-Directed Autonomous Fleet Training And Assessment Across Vehicle Domains, Eric Zavesky

Defensive Publications Series

A system and method for coordinated autonomous vehicle fleet training and assessment across vehicle domains is disclosed. The system comprises a networked fleet of autonomous vehicles operating as self-directed agents with built-in self-preservation, each vehicle configured to receive semantic scenario constraints, interpret and fill gaps using autonomous driving intelligence, independently detect collision risk, and coordinate evasive maneuvers with other vehicles without blindly trusting central commands. A training orchestration engine transmits semantic constraints at multiple specificity levels without specifying complete trajectories. A distributed safety protocol layer implements a three-level detection hierarchy enabling any vehicle or system level to trigger coordinated safety …

Ciers‑Teg V3.0 Thermal Cascade Waste‑Heat Recovery System, Michael Victor Caldwell Mr.

Defensive Publications Series

CIERS‑TEG v3.0 is an open‑hardware, research‑grade thermal‑energy recovery architecture designed for high‑density compute environments such as AI GPU clusters, power‑electronics arrays, and industrial process lines. The system captures waste heat at the chip level, routes it through a controlled thermal cascade, and converts portions of that heat into electrical power, stored thermal energy, and useful low‑grade utilities. Unlike single‑stage thermoelectric approaches, CIERS‑TEG v3.0 maintains exergy across multiple descending temperature stages using microfluidic extraction, gyroid thermal spreading, phase‑change buffering, condensation‑based utility recovery, and parasitic pump‑energy harvesting.

The architecture integrates several previously independent open‑hardware subsystems: Arachne‑Chip ULTRA‑REAL for direct‑die microfluidic heat capture; …

Autonomous Flameless Thermochemical Heating Boiler With Gravity Feed Of Encapsulated Composite Metal Fuel Briquettes, Passive Bimetallic Ash Removal And In-Situ Catalytic Hydrogen Recombination, Volodymyr Kotegov

Defensive Publications Series

The invention relates to autonomous heating systems for residential and small commercial buildings, specifically to flameless thermochemical boilers utilizing controlled exothermic reactions between a solid composite metal-based fuel, water vapour and atmospheric oxygen. The boiler is classified under IPC classes F24H 1/00, F24V 30/00, C01B 3/08, F23C 13/00 and C10L 5/40. It represents a combined system for heat generation and a specialised solid composite fuel manufactured from processed industrial waste. Once initiated, the boiler operates passively in a self-sustained mode and requires only minimal electricity during a brief start-up phase.

Deterministic Predictive State (Dps) Architecture For High-Latency Interplanetary Command And Control, Volodymyr Kotegov

Defensive Publications Series

Round-trip light-time delays between Earth and Mars routinely reach 20–40 minutes. Under these conditions conventional transactional databases and consensus protocols either lock up or time out.

This paper describes a practical architecture that avoids the problem by running identical, bit-for-bit deterministic simulators on the ground and on the spacecraft. All calculations use software fixed-point arithmetic. Both sides share the same atomic-clock timeline (corrected for relativity) and the same sequence of pseudorandom seeds. When delayed telemetry arrives via DTN, the ground side rewinds its state to the timestamp of the data, merges the real measurements with CRDTs, and fast-forwards the simulation …

Path-Diverse Optical Mesh With Erasure Coding (Pdomec): A Non-Line-Of-Sight Tolerant Architecture For Free-Space Optical Inter-Satellite Links, Volodymyr Kotegov

Defensive Publications Series

Free-space optical links in satellite constellations deliver high bandwidth but are vulnerable to short-duration blockages caused by orbital debris, atmospheric turbulence, cloud cells and pointing jitter. Conventional link-level retransmission (ARQ) fails on long inter-satellite baselines because the propagation delay exhausts buffers and collapses throughput.

This paper describes a framing-layer architecture that tolerates such blockages without retransmissions or mechanical re-pointing. Incoming data are collected into fixed-length frames, expanded by erasure coding (Reed-Solomon or rateless fountain codes) with a configurable over-provisioning ratio, and the resulting shards are injected simultaneously across multiple spatially separated optical terminals on the same node. The receiving node …

Distributed Fault-Tolerant Ephemeris Synchronization (Dftes) For Autonomous Satellite Constellations, Volodymyr Kotegov

Defensive Publications Series

Large satellite constellations need continuous, high-precision knowledge of every node’s position and velocity to keep cross-links aligned and to perform collision-avoidance manoeuvres. Dependence on ground-based space situational awareness or GNSS creates a single point of failure that can be jammed, denied or simply out of contact.

This paper describes a fully decentralized software architecture that lets a constellation maintain a consistent ephemeris map using only inter-satellite links. Each satellite runs a local orbital propagator and keeps a state matrix for the whole constellation. When two nodes establish a link they exchange compact state vectors together with an explicit Epistemic Confidence …

Eclipse-Phase Passive Preprocessing With Graphene Thermal Structures (Epp-Gts) For Low-Earth-Orbit Sensor Satellites, Volodymyr Kotegov

Defensive Publications Series

High-resolution Earth-observation satellites in low Earth orbit generate large raw data volumes that require substantial onboard computing for feature extraction, compression and filtering. Performing this work during the sunlit portion of the orbit adds heat and power load at the same time that solar arrays, sensors and transmitters are already active.

This paper describes a practical architecture that moves the heavy preprocessing entirely into the eclipse phase. Raw sensor data are buffered in non-volatile memory while the satellite is in sunlight. Once the spacecraft enters Earth’s shadow the compute cluster is powered from a dedicated lithium-iron-phosphate battery bank sized for …

Chip-Based Physical Gift Card Redemption, Ravi Lakkaraju, Erik Duymelinck, Teruhisa Haruguchi

Defensive Publications Series

This disclosure describes a new approach to redeeming physical gift cards—also known as physical gift vouchers or gift certificates—using chip-based redemption technology to prevent common scams. Traditional gift cards rely on scratch-off private PINs, making them highly vulnerable to physical tampering and victim-assisted scams. To solve this, the proposed chip-based redemption system embeds a radio frequency identification (RFID) or near field communication (NFC) chip directly into the physical gift card. Instead of manually typing a code, a user simply taps the card against a tap-enabled computing device for a quick, touchless transaction. The embedded chip securely stores the card’s private …

Context Aware Wlan To Cellular Handover Via Sensor Fusion And Local Footprint Database, Xuqiang Zhang, Yingqun Yu, Saket Patkar, Roy Lin

Defensive Publications Series

Data stalls and dropped communication sessions can occur when a mobile device transitions from a Wireless Local Area Network (WLAN) to a cellular network in environments with abrupt signal degradation, such as elevators. The disclosed system resolves this limitation by configuring network handover parameters prior to signal loss. Based on environmental detection using hardware sensors or historical network stability metrics retrieved from a local footprint database, an Application Processor (AP) framework sets a shorter hysteresis timer and a lower WLAN signal strength threshold. When a rapid WLAN signal drop is subsequently detected, the user equipment quickly switches the data connection …

Enhanced Code Reviews Through Intelligent Code Diff Ordering And Tagging, Florent Garit, Sindhuja Manickam, Rohan Pandey, Adrian Henle, Ahrenn Sivananthan, Akila Rajesh, Gaurav Tungatkar, Laurentino Sbors Neto, Lulu Bai, Michael Tran, Prakhar Tiwari

Defensive Publications Series

As more code is generated, much of it using artificial intelligence (AI), human code reviews are increasingly critical. Traditional code-review tools provide a description, comments, and a flat code diff but little guidance about review order or attention. This proposal introduces a system that guides a code reviewer by organizing a pull request (PR) as a "story," making it easier to build mental context while keeping the reviewer in control.

Privacy-Preserving Standard Communication With Network Elements (Scone) Hint Tokens For Traffic Classification, Authorization, And Central Processing Unit (Cpu) Reduction, Dennis Lanov

Defensive Publications Series

A technique is proposed herein for privacy-preserving Standard Communication with Network Elements (SCONE) hint tokens that allow trusted policy systems to pre-authorize and classify encrypted Quick UDP Internet Connections (QUIC) traffic without exposing user identity or raw policy details. A token carries only coarse, non-identifying attributes such as traffic class, entitlement level, validity, and flow binding, allowing downstream SCONE-capable network elements to verify the token quickly and generate appropriate throughput advice. By replacing repeated deep classification and policy lookups with lightweight token verification, the technique reduces central processing unit (CPU) load and improves operational scalability across enterprise, wireless-access, mobile, and …

A System And Method For Outcome-Consequence Weighted Scoring In Packet-Switched Communication Networks, Shubham Shivhare

Defensive Publications Series

The present disclosure discloses a system and a method for per-packet Outcome-Consequence Weight Scoring (OCWS) in packet-switched communication networks. The system includes a processor and a memory storing instructions that, when executed by the processor, cause the system to: for each packet received at the system, compute a per-packet significance score as a weighted combination of five semantic factor values, the five semantic factor values including a meaning density value, a downstream consequence probability value, a synchronization criticality value, a time sensitivity value, and an irreplaceability value. The processor is configured to classify each packet into one of a plurality …

Cmefs-1.0 Caldwell Mobile Elevated Fire Suppression System Special-Purpose Mobile Water-Tanker Trailer For Transporting, Elevating, Stabilizing, And Operating A Remote Firefighting Robot, Michael Victor Caldwell Mr.

Defensive Publications Series

CMEFS-1.0 proposes a simple specialized trailer system built around an existing remotely controlled firefighting robot such as the Magirus AirCore TAF35-C.The CMEFS trailer architecture is not restricted to the TAF35-C. Other remotely operated firefighting robots or compact firefighting apparatus may be accommodated when their dimensions, mass, center of gravity, loading requirements, water requirements and operating reaction forces fall within the engineered capacity of the trailer.

The lateral stability member shall be installed/positioned on the side and at the extension required to produce a restoring moment opposing the measured operating overturning moment. The final side, extension and effective mass are engineering …

Multi-Component Automatic Context Window Optimization For Personalization In Large Language Models, Mugurel-Ionut Andreica, Rajiv Mathews

Defensive Publications Series

Personalization in chat assistants requires multiple interdependent components, where outputs from earlier components are consumed as inputs by downstream components. Manual optimization of these context windows remains highly complex. To address this issue, a method is disclosed wherein context windows are organized into a hierarchical schema of atomic blocks. A joint optimization algorithm is utilized to simultaneously refine top-level inputs, atomic block roles, and child block orderings. The optimization process is guided by past candidate performance, structured evaluation rubrics and processes, and internal model signals, including attention weights and introspection scores. Overall performance is maximized across all components without manual …

Recyclable Container, Anonymous

Defensive Publications Series

The invention relates to a recyclable, in  particular flexible, fibre-based container, in particular  5 a tube, a bottle or a flask, for packaging and dispensing  a product, in particular a liquid product, a pharma, dental or cosmetic product. The fibre-based container comprises a fibre-based sleeve (1) for holding the product,  a dispensing head (2) and a cap (3) attachable to the  10 dispensing head (2). The dispensing head (2) comprises a  base element (4) with a dispensing side (5), a bulk side  (6) and a lateral side (7). Furthermore, the dispensing  head (2) comprises an orifice (8) arranged on the …

Multi-Stage Artificial Intelligence-Based Quality Evaluation Of News Channels Using Aggregated Video Analysis, Rohit Patwa

Defensive Publications Series

A multi-stage machine-learning architecture assesses the quality of news channels based on substantive media content. A set of recent media items satisfying a news-content criterion is selected from a candidate channel. Multimodal features of the selected media items are processed using a first AI model to generate respective news quality assessments and qualitative information. The news quality assessments are aggregated using weighted metrics to generate aggregated news quality information. The aggregated news quality information and channel-level features are processed using a second AI model to generate a channel-level quality classification. The channel-level quality classification is used to determine whether the …

Internal Heating Of The Lfp Prismatic Cell For Improved Performance, Anonymous

Defensive Publications Series

LFP batteries are increasingly used in electric vehicles (EVs) and energy storage systems (ESS) due to their advantages in cycle life, safety, and cost. The optimal operating window for LFP batteries is typically around 15–35°C. Effective thermal management is critical for LFP batteries.

Today, most LFP cells—especially in EV applications—are packaged in prismatic form. Cold plates at the bottom of cells with circulating warm or cool fluid are commonly used to heat or cool the cells. To raise temperature more quickly, active heating is sometimes added, typically by placing a heating device outside the cell. The external heating elements must …

Methods For Concurrent Hot-Pressing For The Solid-State Battery Assembly, Anonymous

Defensive Publications Series

Solid-state batteries offer high energy density and safety. A step in their assembly is hot-pressing, which softens the binder to ensure adhesion and packs active materials densely to reduce impedance. Conventional hot-pressing presses the entire electrode stack (containing many layers) at once.

We propose a method to utilize concurrent hot-pressing of subdivided electrode stacks. Instead of pressing a single thick stack (e.g., 38 cathodes and 39 anodes), the assembly is divided into multiple thinner sub-stacks (e.g., 1 cathode and 2 anodes per stack). These sub-stacks are processed either concurrently or sequentially using a heated die set.

The Multi-Agent Ai System For Automated Media Ingestion And Metadata Enrichment, Shravan Nayak

Defensive Publications Series

The process of ingesting and cataloging large volumes of digital media can be inefficient and difficult to scale, often relying on manual workflows that may be slow and error-prone. A multi-agent artificial intelligence framework may be used for automated media content ingestion and metadata enrichment. Such a system can comprise a central orchestrator agent that analyzes incoming media and delegates processing tasks to multiple specialist agents that operate in parallel. A confidence-based routing mechanism can assign a score to generated metadata, where high-confidence results may be automatically committed and low-confidence results can be directed to a human-in-the-loop review process. This …

System For Document Generation Using A Decoupled Abstract Syntax Tree And Two-Pass Execution, Sparsh Dutta, Adithya R, Nikita Rungta, Dev Gupta, Saswata Bhattacharya, Shishir Suman

Defensive Publications Series

Programmatic document generation using large language models (LLMs) with cloud-based editor application programming interfaces can present challenges. Verbose native document formats may lead to high token consumption, and character-indexed updates can cause coordinate drift, potentially resulting in layout corruption. A disclosed technology can utilize a dual-representation architecture that decouples a document's semantic content from its coordinate metadata. A source document may be compiled into a simplified, token-efficient abstract syntax tree (AST) for content population by an LLM, while an isolated index dictionary can preserve the original coordinates and styling. A two-pass synchronization process may then reconstruct the final document by …

Ralph Loop-Driven Agentic Skill Refinement Through Attributed User Feedback, Robert Barton, Yathiraj B. Udupi, Maithri V M

Defensive Publications Series

An agentic platform may capture operator satisfaction signals at scale without connecting that feedback to the specific skill artifacts that drove the evaluated responses. A system is proposed herein that closes this gap by attributing Ralph Loop feedback and agent execution trajectories to specific skill artifacts, aggregating signals across the operator population to localize underperformance to an implicated skill section, and dynamically tailoring diagnostic feedback questions as confidence in the problem diagnosis grows. The resulting closed-loop skill quality management system transforms failed or suboptimal agent executions into targeted skill improvements and validates each improvement by monitoring whether the Ralph Loop …

Defensive Publication — Email Gateway With Message-Identifier Threading For Consolidated Inbound And Outbound Agent Communications On A Dedicated Subdomain, Yahor Karpovich

Defensive Publications Series

This disclosure describes an email gateway architecture for automated agents (including AI-based agents) that communicate with users by email on a dedicated subdomain, consolidating both inbound and outbound message handling in a single component to preserve correct conversational threading. The gateway assigns and retains a stable message identifier for each outbound message, mapped to its internal conversation, task, or record. For inbound messages, the gateway reconstructs the correct conversation by matching standard reference headers (such as In-Reply-To and References) against its stored identifiers, even where the inbound provider does not preserve application context. Where reference headers are absent, altered, or …

Defensive Publication — Automated Rebooking Savings Capture And Revenue-Share Settlement Through An Issuer Ledger, Yahor Karpovich

Defensive Publications Series

This disclosure describes a method for a corporate travel platform to capture, quantify, and settle monetary savings generated by automated re-shopping and rebooking of an existing travel reservation. The system computes realized savings as the deterministic difference between the original settled price and the new booked price of an equivalent reservation, net of change fees and non-equivalence adjustments. Savings are recognized only when the replacement reservation meets configured equivalence criteria (same property/route, comparable class, compatible dates and terms, and policy compliance). Recognized savings are apportioned between the client and platform operator according to a configured revenue-share arrangement, and the client's …