U.S. Government Accountability Office Urges Facial Recognition Regulation

By Zach Segal, Published Aug 27, 2020, 10:22am EDT

The US Government Accountability Office (GAO) is urging facial recognition regulation in a 65-page report. IPVM has reviewed this report, offers analysis on it, and the 3 main steps we recommend they take to achieve this.

IPVM Image

Executive *******

****.*. ********** ************** ****** (GAO)****** ********** **** *********** should ** **** *** transparent, ******* ** ******** uses, ********** ** **** quality ********* ********* ****, built **** **** *** most ******** *** ***** biased **********, *** ******** for **** *** **** protection. **** ****** ********* ******'*************** **** ******** ********** consumer ********** **** ** address *** **********. **** first **** **** ************** 7 ***** *** ** no ****** *** **** recognition *** ******* **** come ***** **** ********, so *********** ********** ** more ******.

What ** ***

****.*. ********** ************** ******** * *********** **** of *** *.*. ******** that ***** ******** ******* issues. *** ** ******** to ** **************** ********. **** *** ***** tasked **** ******** ******** resources *** ***** ***********.

Face *********** *** *****

*** ***** ********** **** recognition *** *** ******** since ****, ****** ** cheaper *** ****** ********** with **** *********. ** ********, *** cases **** ********. *** noted **** ****** ** 2015, **** *********** ** being **** ** ******** for ******* **********, ********** tracking, *** ***** ******* tracing.

IPVM Image

**** ******* **** ********* is ****** ** ************ in ********** *** ********* costs.

*******,**** *** ***** **** Face *********** *** ** limited** *** ************ ******** due ** **** ******, lack ** ******, ******** over *******, *** ******** over **********, **** ** what *** ** ************ here.

GAO **** **** ****** ******** *****

***** *** ********** *** market *** *******, **** relied ****** ****** ******** *****, ** **** ******* notes:

**did *** ************* ****** the global facial recognition market revenues or forecasted revenue estimates. However, we did review information from several market research firms—Allied ****** ********, ******** *** *******, *****************, ******* ********, **********, ****** ******** ******, *** ******* ****** ********— and found that the estimates fell within consistent ranges, with only one outlier. [emphasis added]

**** ** * ******* problem ** **** **** in ***** ************ *** now *** ** **********. This *** **** ***** hurt *** ********** ** making ****** *********.

Existing ******* ***********

******** ** ******* **** recognition ******** *********** ******,******* * ** *** Federal ***** ********** (***) Act*** ***** ** **** recognition, **** ***. *** act********** ******** **** ***** ****** against ********* **** *** unfair ** ********* ********* regarding ********. *** ******** the *** *** *** this ** ** ***** companies *** ******** ** improper ********* ********* **** recognition.

IPVM Image

*** ***** **** *** FTC ******** ****** ******* ************ *** ***************** ** users' ******* ********* ****** recognition. *** *********:

IPVM Image

*** *** *** ****** calls *** ***** *** ****** *********** best ********* ******. ** **, *** FTC ******:

*. ******* ** ******: Companies ****** ***** ** privacy ** ***** ***** of ******* ***********.

*. ********** ******** ******: For ********* **** *** not ********** **** *** context ** * *********** or * ********’* ************ with * ********, ********* should ******* ********* **** choices ** * ******** time *** *******.

*. ************: ********* ****** make *********** ********** *** use ********* ***********.

************, ************ *********** ******* before ********** ****** ***********:

*****, **** ****** ****** a ********’* *********** ******* consent ****** ***** * consumer’s ***** ** *** biometric **** ******* **** that ***** ** * materially ********* ****** **** they *********** **** **** collected *** ****. ******, companies ****** *** *** facial *********** ** ******** anonymous ****** ** * consumer ** ******* *** could *** ********* ******** him ** ***, ******* obtaining *** ********’* *********** express *******

*******, ** ** ***** surveillance ****** ***********, **** rarely *******.

*** *******, *** *** recommended **** *********** ********* to ******* ***** ****** before ********* *** *******:

*** *******, ********* ***** digital ***** ******* ** demographic ********* – ***** often **** ** ********* than ******* ***** **** do *** ******* ******* ****** ******* ***** notice ** ********* **** the ************ *** ** use, ****** ********* **** into ******* **** *** signs.

** *** ***** ****, they **********, **** *** only * **************, *** law:

*******, ** *** ****** the *********** **** ********* go ****** ******** ***** requirements, **** *** *** intended ** ***** ** a ******** *** *** enforcement ******* ** *********** under **** ********* ******** by *** ***

State ****

*** *** ******* * states **** **** ********** to **** ***********; **********, Illinois, **********, *** *****. GAO **** ********* **** it *** ** ****** to ****** **** ******** sets ** *********, ** companies *** ****** ** comply **** *** **** stringent ********* **********.

**** *** **** ****** recognition ********* ***-*** ** providing **** ******** ** Illinois, ** ** *******.

Existing **** *** ******

*******, *** *** *** believe ***** ******** **** adequate ** ******* ********* as *** **** ** developed, *******:

*** ********** *** ******** suggestion **** * **** report (***-**-***) **** ******** ******** strengthening *** ******** ******* framework ** ******* ******* in ********** *** *** marketplace.

IPVM Image

Principles *** **** ***********

*** ****** **** *********** should ** **** *** transparent, ******* ** ******** uses, ********** ** **** quality ********* ********* ****, built **** **** *** most ******** *** ***** biased **********, *** ******** for **** *** **** protection. *** ********** **** data ********** ****** ** limited, **** ******* ********* and *******. **** ****-******* relevant **** ****** ** kept. **** **** ****** be **** *** * specific ******* ****** *******. And *** ***** ****** be *******, *** *** expanded ******* *******. **** there ****** ** ********** for **** ********** *** use ********. ****, **** algorithms **** *** ******** across ************ ****** ** used.

****

*** ******** ******* **** *********** ***** Vendor ******** * ** *********** differences *** ***** **** exists ** **** *********** but ***** ** ** consensus *** *** ***** of ****. **** ***** that ***** **** ********** were ******** *** ********* similarly ******* *** ****** others ********* **** *********** on ********* ******. *** explained **** ***** ** no ********* *** *** cause ** **** ****. GAO ********* **** ******** data ** ***** ****, and ***** ***** **** in **********. **** **** that ***** ******* ** be ** “*****-**** ******” where ********** ********* ***** on ****** *** **** demographically ********* **** *** creators. *** **** ***** that ***** ******* ***** have ******* ******* ** accuracy ********* ** * subject's ****.

Recommendations *** ********** ****

*** *********** **** ******* training **** *** ****, algorithms *** ******** ** avoid ****, ****-******* ****** are ****, *** ********** are *********. **** ***** that ***** ******* ****, or ******** ** * way **** ******* ******** sets *** *******, ***** help******** ****. *** ********* that ********** *** ** modified ** ********* ****. Instead ** ************ ****** error, ********* *** **** algorithms ****** ******* ***** across ********* ************.

*** ******:

********** ***** ****** ********** to ******* ***** ***** rates ******* *********** ******

*** **** *********** **** image ********* **** ******* to. **** ***** **** algorithms ****** ** **** transparent *** *********. **** believe ***-**** ******** ***** help ******** ****. *** dismissed ********* ************ ******* they ******** *** ******** concern ********** *** ******** of *********** ****.

***, ***** *************** *** mitigating **** *** ** challenging. ** ***** ** hard ** *** * diverse ******** *** ******* violating *******. ********** *** be ****** ** ****** between ********* **** ** data **** ** *** diverse. ****, ******** ** mitigate **** ******* ******** can ***** ****-*******. ** algorithm ***** *** ** doing ****** ** ***** men (** ** *******) who *** *** ** the ******** ***, ******* it ** ********* *** the ********** ***** *** in *** ******** ****. Next, ******* ***** ***** rates *** *** **** problems ** ********* ************. It **** **** ********** less ******** ****-***, ***** GAO ******** ** * security *******.

IPVM ***************

***** *** **********'* ********** are ****, *** ********** will **** ** ******* in * ****** ** areas.

(*) ******* ****-******* ******** algorithms *** ****.****'* **** *********** ******** ********* *** ** not ******* ********* ********** are ********** ********. **** means **** **** **** recognition *********, ******* **** Google *** ******** ** most ********** ***** ************ providers, *** ******* **** the *****. *** **** the ***** ** *** necessarily ********* *** ********** a ******* *****. ****,*********** **** *********** ******* are ****** *********,****** ******** ******* *** regulation ****. *** ********** would **** ** ******* some **** ** ******* or ****** ** ****** quality, ** **** ** for **** ***** ******** ranging **** ******* ******** to *********** ** ****, etc.

(*) ***-******* ****** *** a ****** **** ** surveillance. ****** ********* *** to *** ***** ****** is *********, *** **** to ******* *** ***********. The ********** ***** ****** test ******* ***-******* ****** to ********* *** ********* improved *********** ** ****** real-world ***** ************ *** cases.

(*) ********** ** *********** abuse ******** **** **** of ********* *** ******. Facial *********** ********, ******, cannot ************* ********* ** a **** ** ***** it *** ********* ** illegal ********. *** ********** could ******* ****** *********** users ** ******** **** actions **** **** **** for ** ** *** public ** ***** ******* some **** ** ******** upon ********** *******. **** does *** ******* **** the ******** ***** ** able ** ****-******** **** as ******* **** *** much ********** ** **** products *** *** **** to **** ** ******** customer ***.

Comments (3)

*** ****, **** *******. Have *** **** ****** Swiftlane ***? ***** ** use *****'* **** ** for ****** ***********. ***** door ******* ** ********* an ******. * **** a **** *** ******* myself, **** ******* ** you **** ***** ******** notable. ******.

Agree
Disagree
Informative
Unhelpful
Funny

** *** * **** with *********. **** ***** face *********** ** ***** own *** ** *** Apple ********. * ******* they ***** ***** **** reader ** ***** ** Android ********, ****. *'** double ***** *** ****** back.

** *****'* ****** *********, but *** ** *** more *********** ******** ** the **** ******* ****** doubles ** * ***** intercom.

Agree
Disagree
Informative
Unhelpful
Funny

* ***** **** ** see * ****** ** Blue **** ************. * understand ** *** ** the **** ****** *********** software ******* **** * U.S. ******, *** ***** licensed ******** *** ** EMEA.

**** **** ********** – State ** *** *** facial *********** **********

Agree
Disagree
Informative
Unhelpful
Funny
Read this IPVM report for free.

This article is part of IPVM's 7,250 reports and 966 tests and is only available to subscribers. To get a one-time preview of our work, enter your work email to access the full article.

Already a subscriber? Login here | Join now
Loading Related Reports