Home FAQ's Blog Testimonials Videos
Pay TaxMe - Click Image Below

Facial Recognition

Last revised: December 2023
  1. Purpose Statement.
  2. Consent.
  3. Use of Face Recognition Information.
  4. Transparency Disclosure and Retention.
  5. Data Security.
  6. Integrity and Access.
  7. Accountability.
  8. Relationship with Amazon Web Services.
  9. Best Practices for Input Images.
  10. Understanding TaxMe's Face Recognition Utility (FRU).

1. Purpose Statement.

TaxMe uses facial detection technology and biometrics for authentication purposes. Customers who choose this technology, voluntarily upload an image which is stored as a vector file on Amazon Web Services (AWS). TaxMe’s proprietary, state-of-the-art verification technology, allows customers to authenticate themselves securely without the need of usernames/passwords or 2-step verification codes.

This policy delineates the manner in which requests for face recognition are received, processed, catalogued, and responded to.
Back to Top

2. Consent.

Use of TaxMe’s facial recognition technology is completely voluntary. By uploading an image, the user expressly agrees to TaxMe’s Terms of Use, Privacy Policy and Face Recognition Policy.
Back to Top

3. Use of Face Recognition Information.

TaxMe’s customers should expect a simpler more efficient and secure web experience.

TaxMe collects and uses the facial recognition data only for authentication purposes. The data will not be used for any other purpose and will not be sold to third parties.

TaxMe does not, at any time during the image capturing phase, obtain any additional information about the image (i.e., background information).

If TaxMe later decides that it would like to use the images in a materially different manner, before proceeding, TaxMe first will obtain the client’s express consent.
Back to Top

4. Transparency Disclosure and Retention.

When a customer uploads an image, the facial features of the image are captured and stored to facilitate a more efficient customer experience. TaxMe does not store the actual image, i.e. an image in .jpg or .png format. Instead, using advanced detection algorithms, facial features are collected from the image into a vector file which is stored in a database.

TaxMe does not store images for purposes other than those described in this document.

No images will ever be sold/shared with third parties for any purpose, except unless a client expressly consents or pursuant to a court order/law-enforcement directive.

Images will be stored for a period of 3 months, except in the following situations:
  1. TaxMe will remove a client’s image upon receiving a written request.
  2. Model versioning. See Model Versioning for additional details.
Back to Top

5. Data Security.

TaxMe maintains a data security program designed to protect facial recognition data linked to private and personal information through the use of administrative, technological and physical safeguards by:
  1. Not storing the actual image. Rather, storing an image vector which captures facial features from the image.
  2. TaxMe uses customized authentication techniques, together with AWS face matching algorithms, to ensure web users are authorized to access their data.
  3. Uploading an image, in itself, does not create privacy concerns. However, privacy concerns may arise when such image is matched against other data used for authentication process. To address these concerns, TaxMe takes the following steps:
    1. Access to the stored image vector files is strictly limited to TaxMe’s authentication programs. No human access is allowed. Only machine learning language is permitted.
    2. TaxMe’s authentication process combines face recognition with inherent security codes. Inherent security codes can only be interpreted, understood and combined by the owner of the information.
    3. TaxMe utilizes rigorous data security protections for all stored images to prevent unauthorized access.
    4. Access to this technology is available only to returning customers. First time users cannot use this technology.
Back to Top

6. Integrity and Access.

TaxMe has implemented accuracy measures to maintain the authenticity of facial recognition data by setting up prediction levels for images with 99% confidence scores for highly accurate facial matches.

Customers have full control over their image information. Deleting image information can be done by a written request to info@tax-me.com.
Back to Top

7. Accountability.

TaxMe adopts and follows procedures and practices by which it can ensure and evaluate the compliance of users with the face recognition system requirements and with the provisions of this policy. This will include logging access to face recognition information and will entail periodic reviews of systems so as not to establish a discernable pattern that may influence users’ actions.

Customers shall report errors, malfunctions, or deficiencies experienced during the face recognition process and suspected or confirmed violations of TaxMe’s face recognition policy by email at info@tax-me.com.

From time to time, TaxMe will review and update the provisions contained in this face recognition policy and will take appropriate actions in response to changes in applicable law, technology, and/or the purpose and use of the face recognition system; the review process; and customer expectations.
Back to Top

8. Relationship with Amazon Web Services.

TaxMe uses Machine Learning’s Amazon Rekognition services from Amazon Web Services (“AWS”), the world’s most comprehensive cloud platform broadly adopted by millions of customers, including fastest-growing startups, largest enterprises, and leading government agencies.

Amazon Rekognition makes it easy to use images with our applications using proven, highly scalable, deep learning technology. With Amazon Rekognition, TaxMe can identify an individual’s image and provide highly accurate facial detection, analysis, and comparison for user authentication purposes.

Amazon Rekognition provides an application programming interface (“API”) called Amazon Rekognition Image for analyzing images. Below is an overview of Amazon Rekognition Image features:

  1. Face Search: Amazon Rekognition can search for faces. Facial information is indexed into a container known as a collection. Face information in the collection can then be matched with faces detected in images.
  2. Amazon Rekognition Image Operations: Amazon Rekognition Image operations are synchronous and analyze an input image that is in .jpg or .png format.
  3. Storage-Based Operations: Amazon Rekognition Image detects faces in an image and persists information about facial features detected in an Amazon Rekognition collection.
Facial feature information is collected and stored in AWS through the TaxMe portal. TaxMe uses advanced authentication techniques together with face matching to ensure web users are authorized to access their information.

Note: Amazon’s service does not persist actual image bytes. Instead, advanced detection algorithms first detect a face in the input image, then extracts facial features into a feature vector for the face, and finally stores it in a database. Amazon Rekognition uses these feature vectors when performing face matches.

Model Versioning: Amazon Rekognition uses deep learning models to perform face detection and to search for faces in collections. It continues to improve the accuracy of its models based on customer feedback and advances in deep learning research. These improvements are shipped as model updates.

A collection is associated with the most recent version of the model at the time is created. To improve accuracy, the model is occasionally updated.

When a new version of the model is released, the following happens:
  1. New collections are associated with the latest model. Faces added to new collections are detected using the latest model.
  2. Existing collections continue to use the version of the model that they were created with. The face vectors stored in these collections aren't automatically updated to the latest version of the model.
  3. New faces that are added to an existing collection are detected by using the model that's already associated with that collection.
Different versions of the model aren't compatible with each other. Specifically, if an image is indexed into multiples collections that use different versions of the model, the face identifiers for the same detected faces are different. If an image is indexed into multiple collections that are associated with the same model, the face identifiers are the same. This can generate compatibility issues if collections don’t account for updates to the model.

Existing face vectors in a collection cannot be updated to a later version of the model because Amazon Rekognition doesn't store source images. Thus, later versions cannot automatically re-index uploaded facial images.

At all times, TaxMe maintains a single collection. Therefore, when an updated version of the model is available, TaxMe may update its collection. This action will remove all currently stored images. If and when this occurs, users must re-register their image. If a user does not register a new image, no access will be granted. Users can re-register their image at any time, giving them access to their data even when new versions of the model become available.

Error Handling: TaxMe’s uses computer logic to catch and respond to errors using advanced programming languages. This process is separate from the error checking techniques inherent in AWS Servers.

If a user encounters an error while using the system and the error cannot be resolved, the user is encouraged to provide feedback to TaxMe’s support team for further review and troubleshoot. The use of facial recognition technology is optional.
Back to Top

9. Best Practices for Input Images.

Amazon Rekognition Image Operation Latency: TaxMe strives for the lowest possible latency when retrieving images using Amazon Rekognition Image operations, by:
  1. Matching the locations that contains the image vectors with the location where the programing operations occur.
  2. Speeding up the uploading of the image by using image bytes rather than uploading the image itself. As a result, images are processed in near real-time.
Recommendations for Facial Comparison Input Images: TaxMe recommends using the following guidelines when choosing an image:
  1. Use an image with a face that is within the recommended range of angles. The pitch (face up/down) should be less than 30 degrees face down and less than 45 degrees face up. The yaw (face left/right) should be less than 45 degrees in either direction. There is no restriction on the roll (face incline).
  2. Use an image of a face with both eyes open and visible.
  3. Use an image of a face that is not obscured or tightly cropped. The image should contain the full head and shoulders of the person. It should not be cropped to the face bounding box.
  4. Avoid items that block the face, such as headbands and masks.
  5. Use an image of a face that occupies a large proportion of the image for greater accuracy.
  6. Ensure that images are sufficiently large. We recommend a distance of 8 to 12 inches from the camera. This provides a more accurate set of facial comparison results.
  7. Use color images.
  8. Use images with flat lighting on the face, avoid shadows.
  9. Use images that have sufficient contrast with the background.
  10. Use images of faces with neutral facial expressions with mouth closed and little to no smile for higher precision.
  11. Use images that are bright and sharp. Avoid moving cameras or dirty lenses.
Back to Top

10. Understanding TaxMe's Face Recognition Utility (FRU).

The use of TaxMe’s FRU is optional. We developed this technology with the purpose of eliminating the standard use of online accounts, usernames, passwords and other security access technologies. Using a combination of face images and unique inherent identifiers, customers no longer need to struggle with their online access credentials.

New customers cannot use this technology.

TaxMe’s web forms can be submitted without setting up an online account, so no usernames or passwords are needed. When a form is submitted, however, TaxMe collects that data.

Registration: To use the AutoFill features or to get access to forms, authorized users are required to register their face image using our FRU. We not only collect the facial features of a customer's image but also request additional information from existing submissions. If this additional information matches a submission, then our FRU creates a vector profile matching that image with that specific submission. The registration process allows only one image to access a specific submission profile. A submission profile is specific to a web form. Therefore, if a customer attempts multiple registrations under the same submission profile, the previous registration will be overwritten. However, the same individual may register their image multiple times only if that idividual submitted different employment forms, i.e. 941 and 940.

Registration takes seconds to complete and does not require memorizing any information. The information required for registration must be inherently known by the customer. If a customer cannot provide the information requested it would mean he/she is not an authorized individual to the original submission and will not be able to complete registration. As a result, no vector profile will be created.

Customers can use this technology to auto-populate forms or to access copies of tax forms.

AutoFill: Registered customers can use TaxMe’s AutoFill buttons located across all web forms and avoid repetitive tasks such as retyping their information every time they prepare a new tax form.

Access to Forms: Registered customers can download, view and print their submitted employment tax returns (94x family of forms only) by clicking the “Get Your Form” button located on our home page.

Security: TaxMe has implemented accuracy measures to maintain the authenticity of facial recognition data by setting up prediction levels for images with 99% confidence scores for highly accurate facial matches. The combination of unique inherent identifiers and highly accurate facial match levels creates a robust security infrastructure that is easy to use.
Back to Top