Section · 04
German-protected innovation in continuous biometric and behavioral identity verification — DE 20 2026 100 678 U1.
German Registered Utility Model
DE 20 2026 100 678 U1 · A privacy-focused security architecture for continuously verifying a user's identity during device operation.
Sagar Gupta is the applicant for German Registered Utility Model DE 20 2026 100 678 U1, covering a device architecture for continuously evaluating a user's identity through multiple biometric and behavioral characteristics. The system repeatedly determines an identity state during ongoing device use and applies local, graduated access controls without requiring repeated explicit authentication or external transmission of the captured identity signals.
Patent Protection at a Glance
Traditional authentication systems generally verify a user only at a specific moment — when a device is unlocked, an application is opened, or a password is entered. Once authentication succeeds, the system commonly assumes the same authorized person continues using the device. This creates an important security gap: a device may remain accessible when
The invention disclosed in DE 20 2026 100 678 U1 addresses this problem through continuous identity verification. Instead of treating authentication as a single event, the device repeatedly determines an identity state throughout the user's session. The architecture combines multiple biometric signals with behavioral usage signals, evaluates changes over time, and adjusts device access according to the resulting level of identity confidence — and the disclosed security response occurs locally, without requiring the captured signals to be transmitted to an external system.
The device continuously collects a combination of biometric signals (physical identity characteristics) and behavioral signals (interaction and usage patterns) throughout an active session — not only at login.
Rather than treating authentication as a one-time event, the system repeatedly determines an identity state, so the confidence that the current operator is the authorized user is re-established over time.
Changes across the biometric and behavioral inputs are evaluated over time. Deviations from the established identity profile drive a change in identity confidence.
As identity confidence changes, device access is adjusted in graduated steps rather than through a single accept/deny gate — enabling proportional responses instead of abrupt lockouts.
The disclosed security response occurs locally on the device. The captured biometric and behavioral signals are not required to be transmitted to an external system, keeping identity data on-device.
Summarized in plain English — three claims are published in the utility-model specification.
A security device that continuously determines an identity state during device operation using a combination of biometric characteristics and behavioral characteristics of the user, and adapts device access based on the determined identity state.
The device evaluates changes in the captured biometric and behavioral characteristics over time and updates the identity state accordingly, enabling ongoing rather than one-time verification.
The identity-driven access response is applied locally on the device in a graduated manner, without requiring transmission of the captured biometric or behavioral signals to an external system.
Patent 02 · Pending Indian Patent Application
Indian Patent Application No. 202611110434 · Filed September 15, 2026
Artificial intelligence is only as reliable as the data behind it. As organizations connect ERP platforms, CRM systems, cloud applications, APIs, data lakes, mobile applications, logs and operational databases, critical business data is often incomplete, inconsistent, outdated, duplicated or difficult to validate.
This pending application presents an adaptive distributed data-processing architecture that evaluates data quality across multiple sources, determines the reliability of individual data features, selects remediation actions based on confidence, and produces validated, auditable, model-ready datasets for AI, analytics and decision support.
Traditional data preparation relies on fixed cleaning rules — removing duplicates, filling missing values, standardizing formats — without determining whether a specific field is trustworthy enough for a particular analytical or AI-driven decision.
A revenue, inventory, customer, project or operational field may appear complete yet be unreliable due to outdated values, inconsistent definitions, outliers, data drift or differences between source systems. When such data enters an AI model, even a sophisticated model can produce inaccurate forecasts and poorly supported decisions.
Rather than treating a dataset as simply "clean" or "unclean," the system evaluates whether each individual feature is dependable enough for downstream use. It is designed to:
Not every data problem should be automatically corrected — filling a missing value or changing an outlier can create artificial patterns. When confidence is high, the system corrects and validates; when low, it flags for review, quarantines the feature, or keeps unreliable records out of the model-ready dataset, preserving genuine uncertainty rather than concealing it.
Machine-readable provenance records show where data originated, which quality conditions were detected, which remediations were selected, why a feature was retained, transformed, quarantined or excluded, which validations completed, and which transformation version produced the final dataset.
As AI becomes embedded in enterprise operations, organizations need more than powerful models — they need reliable, explainable and continuously governed data foundations where reliability, validation, provenance and adaptive governance are built into the processing lifecycle.
Patent notice: this describes a filed patent application. It is not a statement that a patent has been granted. Scope and status remain subject to applicable patent examination procedures and law.
Reference this work
Journalists, editors, and researchers may reference this record directly. Please use the wording below rather than "granted patent".
Gupta, S. Continuous Identity Verification Using Multiple Biometric and Behavioral Characteristics. German Registered Utility Model DE 20 2026 100 678 U1. Available at: https://sagargupta.us/patent
Related peer-reviewed research on behavioral modelling, fraud detection, and AI in enterprise systems is listed under Scholarly Articles, and the practitioner argument for continuous verification appears in my Forbes Technology Council article, indexed under Published Material.
Note on German Utility-Model Protection
A German utility model (Gebrauchsmuster) is an intellectual-property right registered by the German Patent and Trade Mark Office (DPMA). Utility models are registered after a formal examination and are not substantively examined for novelty, inventive step, or industrial applicability before registration — those requirements may be examined later if the right is challenged or enforced. Utility-model protection may continue for up to ten years, subject to applicable requirements and maintenance fees.
The accurate public description of this right is therefore "Holder of German Registered Utility Model DE 20 2026 100 678 U1" or "Developer and Applicant of a German-Protected Invention for Continuous Identity Verification." The phrase "granted German patent" is not used unless a separate examined patent is granted. Current procedural and legal status can be confirmed through DPMAregister, the official source for German patent and utility-model status.
This page is provided for informational and portfolio-presentation purposes and is not a patentability, validity, infringement, freedom-to-operate, valuation, or legal opinion.