FAST: 1 Second Auth
ACCURATE: 99.96%, Face + Voice
PRIVATE: 100% GDPR Compliant
FAST: 1 Second Auth
ACCURATE: 99.96%, Face + Voice
PRIVATE: 100% GDPR Compliant

Private One-to-Many (1:N) Identification and Authentication + 2FA

The Private Biometrics Cloud™ uses private biometrics to identify individuals in a datastore (1:N identify) with 99,63% accuracy in onesecond. This allows Private Biometrics to replace "Username" and become the primary key

Use two Private Biometrics™ (Face + Voice + Liveness) for 2FA

Full Privacy. Full Human Rights. Full Compliance with GDPR.

Private Biometrics comply with privacy regulations and honor human rights worldwide by operating only in the encrypted space. We do not transmit, store, two-way encrypt or operate on plaintext biometrics.

Face + Voice + Liveness

pb-card
macbook-reimagine
Private One-to-Many (1:N) identity + 2FA
FACE + VOICE + LIVENESS
Private Biometrics BOPS RESTful Web Service™
EXPLORE DEMO

The Private Biometrics Web Service™ uses private biometrics to identify individuals in a datastore (1:N identify) with 99.63% accuracy in one second. This allows Private Biometrics to replace “username” and become the primary key. Our neural networks are so accurate they distinguish between identical twins and accommodate boundary conditions such as poor positioning and poor lighting.

The Private Biometrics Web Service is a full implementation of IEEE 2410 BOPS III and is available using any thin or thick client (phone app, tablet, laptop or browser) with a mic or camera >1MP.

One-to-Many (1:N) Voice Identification & 1-Factor Authentication (1FA) in Any Language

Voice Identification with Liveness

phone-pb-card
EXPLORE DEMO

The Private Biometrics Web Service™ uses private voice biometrics to identify individuals in a datastore (1:N identify) independent of spoken language with high accuracy in one second. Specifically, our neural networks measure the signals inside of the voice biometric. This allows Private Biometrics to replace “username” and become the primary key.

The Private Biometrics Web Service™ is a full implementation of IEEE 2410 BOPS III and is available using any thin or thick client (phone app, tablet, laptop or browser) with a mic or camera >1MP.

Advanced Anti-Spoofing and Liveness.

Liveness detection

phone-pb-card
EXPLORE

The Private Biometrics Web Service™ assesses Liveness by asking the user to read a few random words. We then concurrently process the voice biometric using two algorithms and return a result in less than one second.

The first algorithm uses text to speech to compare the pronounced text to the requested text. The second uses the Private Biometrics Cloud™ to perform a one-to-many (1:N) identification on the private voice biometric to ensure that the voice matches the user.

women-facecircle
Global-Scale Identity + Authentication Fully Compliant with IEEE 2410-2018 Biometric Open Protocol Standard (BOPS III)

Private biometrics. The Next Revolution in Digital Identity & Authentication

For centuries, private data has been encrypted for privacy and then decrypted for use. Alleviating the need to decrypt by performing mathematical operations on encrypted data (homomorphic encryption) has stood for thirty years as an unattainable goal – the “Holy Grail” of cryptography.

girl-cirle

We have solved homomorphic encryption, encrypted search, encrypted match, and several other key problems in biometrics and cryptography. These breakthroughs allow our privacy-enabled AI software to compute identity in the encrypted space in polynomial time using multiple DNNs with the same accuracyas plaintext and comply with data privacy laws worldwide.

The Private Biometrics BOPS RESTful Web Service™ is biometric agonistic and processes face, voice and liveness concurrently. Now, there are no remaining public policy or technical obstacles to using face as the primary key for identity.

Private Biometrics BOPS RESTful Web Service™

Supported by twenty years of published cryptographic research in private biometrics and four recently filed United States patent applications, the Private Biometrics BOPS RESTful Web Service enables fast and effortless 2-Factor Authentication (2FA, face+voice+liveness) for thin and thick clients with a camera (>1MP) and Voice Authentication with Liveness for clients with a microphone.

The Private Biometrics BOPS RESTful Web Service is industry’s first implementation of the (to be published 12/2018) IEEE 2410-2018 Biometric Open Protocol (BOPS III) Standard. This Standard ensures full compliance with global privacy regulations by mandating all biometric processing occur in the encrypted space and specifies three RESTful API calls: Enroll, Predict and Liveness.

girl

The Private Biometrics Developer Wiki provides sample code, apps and libraries for Web browsers, iOS and Android apps.

GLOBAL-SCALE IDENTITY BASED ON OPEN STANDARDS

BOPS represents a breakthrough in financial transactions. For the first time, financial brokers and customers are offered unique, repeatable assurance that every transaction can be tied to a person without question. The timing could not be better, as banks and financials are moving away from passwords and PINs, as we seek better vehicles to safeguard our data. This level of assurance stands head-and-shoulders above traditional authentication frameworks that suffer hacks on a daily basis.”

Kevin McNamara, Founder and CTO, HiveIO and
formally VP of R&D, JPMorgan Chase (Source)

IEEE 2410-2018 Biometric Open Protocol Standard (BOPS III) is a global open standard for identity and authentication that describes a system for identity assertion that provides legal non-repudiation for physical and logical transactions.

Specifically, the BOPS III specification, to be released 12/2018, serves as a foundation for biometric authentication of digital transactions, physical access, Active Directory®, and other instances where authentication is required. BOPS III is the only biometric standard that provides for high levels for security, privacy, compliance and user convenience while eliminating the risk of losing biometrics.

unlock-cirle

The Private Biometrics Web Service™ is industry’s first implementation of BOPS III.

Open standards, including BOPS III, help foster innovation and disruptive technologies by allowing a broad cohort of vendors to develop compatible products that interrelate.

explore-2FA
EXPLORE IDENTITY AUTHENTICATION + 2FA FACE + VOICE + LIVENESS

ONE-TO-MANY (1:N) IDENTIFY (FACE + VOICE + LIVENESS) WITH 2FA

The Private Biometrics Web Service™ uses private biometrics to identify individuals in a datastore (1:N identify) with 99.63% accuracy in one second. This allows Private Biometrics to replace “username” and become the primary key...

“The Private Biometrics Web Service maintains full accuracy during boundary cases, overcoming conditions such as poor lighting or positioning.”

Our neural networks are so accurate they distinguish between identical twins and accommodate boundary conditions such as poor positioning and poor lighting.

The Private Biometrics Web Service is a full implementation of IEEE 2410 BOPS III and is available using any thin or thick client (phone app, tablet, laptop or browser) with a mic or camera >1MP.

To predict (authenticate a user), the client-side device or Private Biometrics Web Service:

  • Concurrently acquires face, voice and liveness.
    The Service acquires and pre-processes facial and voice biometrics concurrently. At least two selfies and five seconds of voice are required to Predict or Enroll.
  • Creates Private Biometrics.
    The client-side device or Web Service will use a pre-trained CNN to create small (4kB for face, 2kB for voice) one-way encrypted, Euclidean measurable homomorphic feature vectors (Private Biometrics™) that can be stored and operated on locally or in the Cloud
  • Discard plaintext biometrics.
    Every plaintext biometric is deleted after preprocessing. Private Biometrics are one-way encryptions (they have no decrypt key) and all processing (encrypted search, encrypted match) occurs in the encrypted space. There is no method to reverse engineer a private biometric to return to the original biometric, as the encryption is 1/10,000th the size of the reference biometric.
  • Store private biometrics locally (optional).
    If the customer chooses to implement a 1:1 Verify function (like Apple® FaceID) where the private biometric will be used to verify a single user and the device may be online or offline, then the private biometrics are stored locally. The device performs encrypted match using linear math. Specifically, we use the Frobenius 2 distance function to determine the closeness of two private biometrics. Results are returned in milliseconds with >99% accuracy. There is no requirement to train a neural network on the device.
  • Transmit the private biometrics to the Private Biometrics BOPS RESTful Web Service.
    The private biometric is small (4kB or less) and one-way encrypted. The payload can be transmitted securely to the Web Service over TLS without risk.
  • Receive Results.
    The Web Service will return the GUID or other information offered at enroll time that has been associated with the person’s identity in less than 1 second or will identify the person as unknown.
SOMETIMES, COMPUTER VISION IS UNAVAILABLE.
EXPLORE VOICE IDENTIFICATION + 1FA
Voice + Liveness
blindfold

One-to-Many (1:N) Voice Identification in Any Language + 1FA

Now, any thick or thin client with a mic can perform frictionless biometric verification (1:1), one-to-many identification (1:N) and 1-factor authentication (1FA) in one second with high accuracy.

The Private Biometrics Web Service™ uses private voice biometrics to identify individuals in a datastore (1:N identify) independent of spoken language with high accuracy in one second. Our neural networks measure the signals inside of a voice sample with high accuracy and thus allow Private Biometrics to replace “username” and become the primary key.

We use face as the first biometric and voice as the second for 2-Factor Authentication (2FA) because the voice biometric is easily and intuitively sampled at the same time the face biometric is collected.

Typically, the individual wishing to authenticate is asked to read a few words while looking into a camera and we collect the face biometric and voice biometric while the user is speaking. The resulting audio is used to check liveness and to ensure the identity of the user’s voice matches the identity of the face.

Private voice biometrics can easily used for identity applications (“who is on the phone?”) and 1-Factor Authentication (1FA) by call centers, smart speakers (“Alexa”), phone, watch and TV apps, physical security devices (door locks), and other situations where a camera is unavailable.

wiki

Learn more about
Private Biometrics
on Wikipedia

We follow the same privacy-preserving encryption procedure for both face and voice biometrics. Immediately after collecting a voice biometric, we create a private biometric and discard the original unencrypted biometric template.

dialog

Within the next five years, more than 30% of US consumers – 78 million people – could be using voice payments according to a survey by Business Insider. Source

The private biometric allows a person to be identified or authenticated while still guaranteeing individual privacy and fundamental human rights by only operating on biometric data in the encrypted space.

To transform the unencrypted voice biometric into a private biometric, we first pre-process the voice signal and reduce it to its smallest form without any loss. The Nyquist sampling rate for this step is two times the frequency of the signal. We then sample the resulting data and use this sample as input to a Fourier transform.

The resulting frequencies are then used as input to a pre-trained voice neural network capable of returning a set of embeddings. These embeddings, each 64 floating point numbers, provide us with private biometrics which then serve as input to a second neural network for classification.

ADVANCED ANTI-SPOOFING

ADVANCED LIVENESS DETECTION

heart

Advanced Anti-Spoofing & Liveness.

The Private Biometrics Web Service™ assesses liveness by asking the user to read a few random words. It then concurrently processes the received voice signal through two algorithms and returns a result in less than one second.
The first algorithm uses speech-to-text to verify the words were read correctly and the second uses the Web Service’s Predict API to perform a one-to-many (1:N) identification on the private voice biometric to ensure that it correctly identifies the user.

Combining the Private Biometrics’ 2-Factor Authentication (2FA, face+voice) with Liveness offers a unique opportunity to improve security and convenience. The first factor, Face, is used to establish identity. The second factor, voice, is used to confirm identity, establish authentication and determine if the source presenting the features is live.

hand-with-phone micro

Private Biometrics performs liveness in a very simple and accurate form. We produce random text that takes roughly 5 seconds to speak (in whatever language the user prefers). The user reads the text and the Private Biometrics Cloud then performs a text to speech and compares the pronounced text to the requested text. This allows Private Biometrics to assert the liveness of the requestor.

For the final feature of the 2FA, Private Biometrics compares the liveness random text voice input and performs an identity assertion to ensure the voice that spoke the random words matches the user’s identity. The input audio is now used for liveness and identity.

meeting

About Us

For the past twenty years, cryptographic and biometric researchers have pursued four approaches to solving private biometrics: cancelable biometrics, BioHashing, Biometric Cryptosystems,and two-way partial homomorphic encryption. Each of these methods introduced computational and communications overheads which were reasonably inexpensive for 1:1 verification but proved infeasible for large one-to-many (1:N) identification requirements.
In late 2017, Scott Streit and Mike Pollard (seated together on the left side of the table in the picture above) co-founded Private Biometrics with the goal of solving homomorphic encryption. At the time, the problem was being actively worked on by IBM Research and Microsoft Research and was considered one of industry’s most significant unsolved computer science problems. The societal good achieved by solving one-way homomorphic encryption is full privacy.

We spent the first few weeks building the laboratory’s computing resources and recruiting a small team of top computer scientists and software developers from around the world with considerable experience in machine learning .

We first created a list of what not to try. This list detailed all prior approaches to the problem and helped narrow our focus. The team leveraged their knowledge and experience in machine learning, cloud computing and biometrics and converged on a solution. We then submitted patent applications and started development of production software.

In May 2018, we shared our work on homomorphic encryption with IEEE’s Biometric Open Protocol Standard (BOPS II) committee and IEEE updated the BOPS II standard (now called BOPS III) to reflect this advance. The most significant change in the specification was IEEE’s elimination of all key management requirements. This greatly reduced complexity and allowed the number of APIs in the specification to drop from 31 to 3. The Private Biometrics BOPS RESTful Web Service is a fully compliant BOPS III implementation.

In September 2018, we deployed the Private Biometrics Web Service on Amazon Web Services. Our write once, run anywhere software is written in Python and runs on AWS, a private cloud or locally.

Full documentation, demo apps and developer libraries are now available for Web browsers, Android and iPhone on the Private Biometrics Developer Wiki.

We have tested Private Biometrics against all known face datasets and we consistently maintain the same accuracy as plaintext (99.63% or greater). The Web Service concurrently processes multiple biometrics (face, voice), tests for liveness and returns results in less than 1 second.

We use Google’s Facenet pretrained CNN for face recognition. The second release of Facenet was a great step forward and allowed us to achieve much higher accuracy across large datasets of faces.

Interestingly, we could not locate a similar pre-trained CNN for Voice Identification so we built our own. We acquired voice samples, trained the network, and published our first pre-trained CNN voice model in September 2018. The secondary DNN behaves similarly to the second DNN for face, allowing a voice solution that is also 100% private.

“If I have seen further it is only by standing on the shoulders of Giants.” - Sir Isaac Newton, 1676

MEET OUR TEAM

interview

Our underlying technology was incubated with private research funding that supported a group of top computer scientists and software developers from around the world with considerable experience in machine learning.

Key Vendors

Software: TensorFlow, FaceNet Google
Public Cloud: AWS
IP: Matthew Grady Wolf, Greenfield & Sacks, P.C.
Corporate: Stephen Huttler Pillsbury Winthrop Shaw Pittman LLP
Audit: William Leffler RSM US LLP

MANAGEMENT

MIKE POLLARD CEO

Mike is an entrepreneur experienced in high-growth technology ventures, big data analytics and cyber security. Prior to co-founding Private Biometrics, Mike co-founded Discovery Logic (acquired by Thomson Reuters), founded and served as CEO and Chair of thinkXML (acquired by Zion Bancorp), and founded and served as CEO of Science Management Corp. (acquired by Northrop Grumman).

mike@privatebiometrics.com

SCOTT STREIT CTO

Scott is an experienced data scientist working in the fields of machine learning and cyber security. Scott also currently serves as Chair of Biometric Security for IEEE and leads the IEEE 2410 Biometric Open Protocol Standard (BOPS). Scott has authored several key patents and papers in machine learning, biometrics and authentication.

scott@privatebiometrics.com

Contact Us

Website

privatebiometrics.com
openinfer.com
openinference.com
scott@privatebiometrics.com
mike@privatebiometrics.com
+1-301-938-6300
13331 Signal Tree Ln, Potomac, MD 20854