DCT

8:23-cv-02031

Streit v. Private Identity LLC

Log In
Key Events
Complaint

I. Executive Summary and Procedural Information

  • Parties & Counsel:
  • Case Identification: 8:23-cv-02031, D. Md., 07/27/2023
  • Venue Allegations: Venue is alleged to be proper in the District of Maryland as Defendant's principal place of business is in Potomac, Maryland, and its two members reside in Maryland.
  • Core Dispute: Plaintiff alleges he is an unnamed joint inventor of fifteen U.S. patents assigned to the Defendant and seeks correction of inventorship under 35 U.S.C. § 256.
  • Technical Context: The patents relate to systems and methods for privacy-enabled biometric authentication, a field critical for securing personal identity on mobile and networked devices without exposing raw biometric data.
  • Key Procedural History: The complaint alleges that Plaintiff, Dr. Brian Streit, collaborated with the named inventor, Scott Streit, from at least January 2017 through March 2018. Plaintiff alleges his inventive contributions from this period are documented in emails, texts, meeting notes, programming code, and a co-authored paper titled "Privacy-Enabled Biometric Search," which is cited in the patents-in-suit. This alleged collaboration and documentation will be central to establishing Plaintiff's claim of joint inventorship.

Case Timeline

Date Event
2017-01-01 Alleged collaboration period between Dr. Streit and Scott Streit begins (approx.)
2018-03-07 Priority date for ’221, ’070, ’333, ’802, ’168, ’831, ’552, ’452, and ’559 Patents
2018-03-07 Four key patent applications filed, which matured into the '221, '070, '333, and '802 patents
2018-06-28 Priority date for ’084 Patent
2018-12-12 Priority date for ’375 Patent
2019-09-17 ’221 Patent Issued
2020-07-21 ’070 Patent Issued
2020-08-14 Priority date for ’852 and ’078 Patents
2021-03-02 ’852 Patent Issued
2021-09-14 ’078 Patent Issued
2021-10-05 ’333 Patent Issued
2021-11-09 ’084 Patent Issued
2021-12-28 ’375 Patent Issued
2022-03-01 ’168 Patent Issued
2022-06-14 ’831 Patent Issued
2022-07-19 ’802 and ’552 Patents Issued
2022-11-01 Priority date for ’866 Patent
2022-11-01 ’866 Patent Issued
2022-11-15 Priority date for ’841 Patent
2022-11-15 ’841 Patent Issued
2023-05-02 ’452 Patent Issued
2023-06-13 ’559 Patent Issued
2023-07-27 Complaint Filing Date

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 10,419,221 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: September 17, 2019)

The Invention Explained

  • Problem Addressed: The patent addresses the security and privacy vulnerabilities inherent in conventional biometric systems, which often require searching or matching unencrypted biometric data, creating risks of data compromise. (Compl. ¶3; ’221 Patent, col. 1:16-30). Conventional approaches are also noted to be limited in their ability to perform "one to many" searches efficiently on encrypted data. (’221 Patent, col. 1:21-25).
  • The Patented Solution: The invention uses a deep neural network (DNN) to convert raw biometric data into a one-way homomorphic encryption, creating a "feature vector." This encrypted vector can be used for matching and searching without exposing the original biometric data, which is discarded after the vector is generated. The homomorphic nature of the encryption allows for computations (comparisons) to be performed directly on the encrypted data, enhancing security and enabling scalable one-to-many searches. (’221 Patent, Abstract; col. 1:56-65).
  • Technical Importance: This approach enables biometric authentication on less secure platforms, like mobile devices, by ensuring the original, sensitive biometric data is never retained or transmitted in a vulnerable state. (Compl. ¶3).

Key Claims at a Glance

  • The complaint identifies Claim 1 as representative. (Compl. ¶23).
  • Essential elements of independent Claim 1 include:
    • A classification component with a deep neural network (DNN).
    • The DNN is configured to classify "Euclidean measurable encrypted feature vector and label inputs" during a training phase.
    • The DNN is also configured to accept at least one "unclassified encrypted feature vector" as an input during a prediction phase to determine a label or an unknown result.
    • The DNN generates an array of values in response to the input, and determines the label or result by analyzing the position and value of entries within that array.

U.S. Patent No. 10,721,070 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: July 21, 2020)

The Invention Explained

  • Problem Addressed: The '070 Patent addresses the challenge of dynamically adding new users to a biometric system. Conventional systems may require a full retraining of the neural network model to accommodate new people, which is computationally expensive and inefficient. (’070 Patent, col. 7:10-15).
  • The Patented Solution: The invention discloses a classification network architecture with a plurality of layers, where at least one layer has an initial number of "identification nodes" with a subset of those nodes being "unassigned." When a new user provides biometric information, the system triggers an "incremental training operation" that integrates the new user's information into one of the unassigned nodes, making it available for subsequent matching without a full system retraining. (’070 Patent, col. 7:15-20; col. 8:45-49).
  • Technical Importance: This incremental training capability allows a biometric security system to scale efficiently by adding new users in a computationally lightweight manner. (Compl. ¶14).

Key Claims at a Glance

  • The complaint references a "‘979 patent" as representative, which appears to be a typographical error for the '070 patent, as the quoted claim language matches Claim 1 of the '070 patent. (Compl. ¶22).
  • Essential elements of independent Claim 1 include:
    • A classification component with a deep neural network (DNN) for classifying feature vectors and labels during training.
    • The classification network has an architecture with a plurality of layers, at least one of which comprises nodes associated with feature vectors.
    • This layer has an "initial number of identification nodes and a subset of the identification nodes that are unassigned."
    • The system is configured to trigger an "incremental training operation" to integrate new biometric information into an unallocated identification node for subsequent matching.

U.S. Patent No. 10,938,852 - "Systems and Methods for Private Authentication With Helper Networks" (Issued: March 2, 2021)

Technology Synopsis

This patent describes an "authentication data gateway" that uses "helper networks" to pre-process and validate biometric information before it is used for enrollment or identification. A "geometry helper network" filters identifying characteristics, and a "validation helper network" validates or rejects the input, ensuring only high-quality data is used by subsequent neural networks. (Compl. ¶28).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶28).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,122,078 - "Systems and Methods for Private Authentication With Helper Networks" (Issued: September 14, 2021)

Technology Synopsis

This patent, related to the '852 Patent, also describes an authentication data gateway. It specifically claims a "pre-trained validation helper network" for voice identification information, which evaluates an unknown voice sample against pre-trained criteria that are independent of the subject seeking authentication. (Compl. ¶30).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶30).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,138,333 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: October 5, 2021)

Technology Synopsis

This patent describes a privacy-enabled biometric system with an "enrollment interface." The interface accepts unencrypted biometric information, provides the resulting encrypted feature vectors to a classification component for training, and then deletes the original unencrypted information to protect privacy. (Compl. ¶21). Plaintiff specifically alleges jointly inventing the features of accepting encrypted feature vectors and returning a label or unknown result. (Compl. ¶13).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶21).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,170,084 - "Biometric Authentication" (Issued: November 9, 2021)

Technology Synopsis

This patent claims a method for authorizing access by calculating a "user identity probability" based on a plurality of sources. These sources include passive user behavior information, active authentication input, and a "liveness evaluation" to validate that the information is from a live user. (Compl. ¶26).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶26).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,210,375 - "Systems and Methods for Biometric Processing With Liveness" (Issued: December 28, 2021)

Technology Synopsis

This patent describes an authentication system that validates the "contemporaneous input of biometrics." It analyzes a "liveness threshold" by processing a candidate set of biometric instances to determine if they match a random set of instances, thereby confirming the user is live during the authentication request. (Compl. ¶32).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶32).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,265,168 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: March 1, 2022)

Technology Synopsis

This patent discloses a system that determines an "authentication mode" and triggers one or two machine learning (ML) processes. A first ML process authenticates using already-encrypted feature vectors, while a second ML process generates new encrypted vectors from plain text biometric inputs. (Compl. ¶34).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶34).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,362,831 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: June 14, 2022)

Technology Synopsis

This patent describes a system where a deep neural network (DNN) is configured to accept "distance measurable encrypted feature vectors" as input. The DNN is trained on these vectors and corresponding labels to predict a match or return an unknown result for a new input vector. (Compl. ¶36).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶36).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,392,802 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: July 19, 2022)

Technology Synopsis

This patent describes a system using a linked pair of neural networks: a pre-trained convolutional neural network (CNN) generates encrypted feature vectors from unencrypted biometric input, and a fully connected neural network (FCNN) uses those vectors for classification. Plaintiff specifically alleges jointly inventing this feature. (Compl. ¶16).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶24).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,394,552 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: July 19, 2022)

Technology Synopsis

This patent discloses an authentication system where a deep neural network (DNN) outputs an array of values reflecting a probability of a match. The classification component is further configured to retrieve encrypted credentials from memory based on the match identified in the array, determine a distance, and return a "distance match" if a threshold is met. (Compl. ¶38).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶38).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,489,866 - "Systems and Methods for Private Authentication With Helper Networks" (Issued: November 1, 2022)

Technology Synopsis

This patent describes an identification data gateway using multiple "validation helper networks," including a first voice helper network and a first image helper network, to validate different types of biometric information independently of the subject seeking authentication. (Compl. ¶40).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶40).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,502,841 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: November 15, 2022)

Technology Synopsis

This patent claims a system that executes first and second machine learning (ML) processes based on an authentication mode. The system validates that the identification results are captured from a "live submission," including operations to determine liveness in multiple dimensions. (Compl. ¶42).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶42).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,640,452 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: May 2, 2023)

Technology Synopsis

This patent describes a system with an enrollment interface that accepts plaintext biometric information, provides encrypted feature vectors for classification, and then deletes the plaintext information. The deep neural network is configured to predict an outcome based on a trained model to match a result label or unknown. (Compl. ¶44).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶44).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

U.S. Patent No. 11,677,559 - "Systems and Methods for Privacy-Enabled Biometric Processing" (Issued: June 13, 2023)

Technology Synopsis

This patent describes a system where a deep neural network (DNN) is trained on "distance measurable homomorphic encrypted feature vector and respective label inputs." The DNN accepts such vectors as input and predicts a match to a label or returns an unknown result. (Compl. ¶46).

Asserted Claims

Claim 1 is asserted as representative. (Compl. ¶46).

Accused Features

No infringement is alleged; this is an inventorship action. (Compl. ¶1).

No probative visual evidence provided in complaint.

III. Analyst’s Conclusion: Key Questions for the Case

As this is not an infringement action, the dispute will not turn on claim construction or technical comparisons to an accused product. Instead, the central questions are factual and legal inquiries into the nature of inventorship.

  • A core issue will be one of contribution and conception: Did Dr. Brian Streit's alleged contributions—such as developing features related to accepting encrypted vectors, allocating unassigned class slots for new users, or linking CNNs with FCNNs—rise to the level of joint conception of the claimed inventions, or were they merely the work of a skilled artisan implementing the ideas of the named inventor? (Compl. ¶¶13-16, 47).
  • A key evidentiary question will be one of corroboration: What specific documentary evidence (e.g., emails, programming code, meeting notes) and testimony exists to corroborate the timing, substance, and significance of Dr. Streit's alleged inventive contributions to the subject matter of the claims in each of the fifteen patents? (Compl. ¶¶5, 18).
  • A central legal question will be one of ownership: If Dr. Streit is added as a joint inventor, what rights does he possess as a co-owner of the patents, given the complaint's allegation that he was never an employee of the Defendant and had no obligation to assign his invention rights? (Compl. ¶¶19, 51).