Facit Cloak Identity GDPR Redaction Software Tested

RS
Robert Shih
Published Nov 28, 2018 13:53 PM

*** **** ** **** ***** *** releasing ******** *********** *** ********* ******** in ***** ********* ********.

*** ** ***** ******** ** *** UK's ***** **** ******'* ******** *****.

** **** ****, ** ******* *** performance ** *** ******** *** ******** it *********'* ************* ********* *******.

Performance *******

******** ***** *** ******* ***** *** redacting *********** *** ******** **** ******** false ********* (*********** **** *** ****** objects ** *****) *** **** ***** negatives (** ** ******** **** ****** as ******). ****** ********* *** ****** yet *******. ******* ********* ********* *********** was ******** ** ******* *********.

*******, ******** ***** ** ********* ** a ******* ****** ********** ** *,*** GBP *** **** **** **** ***** redacted ******* * ******* ** **.** GBP (~$*** ***).

General ********

*** ********* ** ******** ***** ******* a ******-***** *********** ********* **** *** "Import" ********* *** ******** ** "*** & *****", "****", "*******", **** "******." 

****** ****** *** **** ** .***, .mp4, .***, *** .*** *******. *** resulting ******** ***** ** ** .*** format.

*** ********* ****** **** ***** ************ the ******* ** *** ********* ******** within *** "*******" ********, ***** ****** of ******** ** *** *********, *** gives * ****** ** *** ********* as *********:

Trial **** ********* ********

** ******** ******** ** ********* ** their ********** *-*** ***** ****: "** is ******* ** *** *** ******** in **** ****":

trial mode watermark

** **** **** ***** ****** ** clarify **** ** ** *** '*******', simply ** ******** ***** ********* **********.

Detection ** **** **********, ****** *** ******

***** ***** ***** **** * ***** perspective, ****** *** ******, ******** ***** was **** ** ****** ******** *****. Interestingly ** **** ******** *** ******* privacy ******* ** * ******* *******'* license *****:

detection-on-high-resolution -narrow-fov-strong

***** ***** **** * ******* *** camera, **** **** *********** ********, ******** Cloak *** **** ** ****** ***** and ****** ****** ******** ****** *** FoV:

detection-on-high-resolution -narrow-fov-strong2

Full **** ****** ******** ****

******** ***** ****** * ***** ** blurring, ******** *** **** ****. ** testing, ******** **** ******* *****, *** would **** ************** **** ***** **** and ****, *** *** **** ** feet. **** **** ******** ********* ******** in ******** *** ****** ******* **** blurring.

******** ********:

standard-blurring

**** **** ******** ********* ******** ** covering *** ****** ******* **** ********, as ************ *****:

full-body-blurring

False ********* ** ****** ********** *******

***** **** ******** ********** ****** ***** analysis **** ******** ***** ******* ****** signs, ********, *******, *** **** ****** OSD:

false-positives-on-moving-background-objects

Frames ***** ********* ** ****

***** ******** ***** *** *** **** any ******** ******** ** *** *****, there **** ********* ***** ******* ****** and ******** ********** ***** ***** ********* to ****:

frames-where-detection-is-lost

Fallback ** ****** *********

******** ***** ****** * ****** ********* function ** *** ***********. ****** *** Genetic ***** *** *** **** **** a *** ** ****** * ******, Facit ******** ******** ****** ** ***** a ****** ** ***** *****:

manual-redaction

Local ** *********** ****

**** ******** *****, ***** ** ** cloud/hosted ******. *** *********** ****** ** local ******* *********. *** ********* ** the *** *** *** ******** ** load ****** ********:

cpu usage

Sold ****** ** ******* *******

******** ***** ** ********* ******** ** through ********** *********.

********* *** ******** ******* ****** *** the ******* ** *** ********* ********:

******** '*********' ***** ** ********* *** month *** *******.

Compared ** ******* ********* **** *********

******** *****'* **** ********* ********* **** consistently **** *********, ** ****** ****** of *********, ***** **** *** ****** faces ** **** **** **** **** the ****** ** ****** *** ***** of ****.

******* ****** ***** *******, ******* *** sharing ** *****, ***** ******** ***** is * ***** ***********.

******* ** ******** ************ ********** $*** *** ***** (*** ** new ***** * ***** *** *********** unlimited ******), ***** ******** ***** *** yearly ******* ******* **** ** ******* toward *** ***** ******** *******.

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

*** ********** ********** ******* ***** ******** Cloak *** ******* ********* ** **** Facit ****** **** **** *** **** body ********* ***** ******* ** **** only.

**** *********** *** *** ******** ***** what **** ** ***** ** ***** redaction ** ********. ** ***** ***** the "*************" ** "****************", ********* ***** of *** "**** ********** ** ****** and ** *******" ******** ** ****'* ******* **. *******, *** **** ***** ******* ************ **** *************/**************** ********** should ** ****. ** ****** **** that **** *********** "********* *********** ********* and ************** ********" **** ** ****************, which ** * ********* **** "*** reduce *** ***** ** *** **** subjects *********" (******* **). ** **** ** ** '**** law' *** **** ********** ******* **** vs **** ********* ****** ********* **** is *** **** ************ ********** *********** part ** ***'* ******* **** ** cameras.

GDPR *************

***** ********* ** **** **** *** product's *** *** ************* ** *** price ** *** **** ** ***** GDPR *****:

************* *** ******** *** **** ********* to *********** *** ******** **** ***********, which **** **** ***** ** *** 2018. **** *** **** *** ***** being **** ** *** ** ** either * €**,***,*** ** * ******* *% of ********* ******* ** * ******* fine ** **** ************ ** ** breach (********* ** *** *******). ** everyone ***** ** **** ****** ** some **** ** * ********* **** but ** *** ***** *** **** times **** **** ******* **.

***** *** **** *****, ******* **** the *********** ** *** ***** ******** to ***** **** *** ** *** two **** ******** *** ****** £*,*** or £*,*** *** ****. **** ** like ** ********* ******* **** ******* the ************ *** * ***** ********* software ****** *** *** *********** ******* nullifying *** **** ** * **** Breach ** **** **** ********* ********** the **** ** ** *** *****.

**** **** ******* ** **** ** you ** ******** **** ** *** the ***********, *** £** *** ****** (for *** £*,*** *******) ** £** per ****** (*** *** £,*** *******) is ******** **** **** ****** ***, this *** ******** ** ****** ***** your ******* ******* £*. ** ***** point *** **** *** ****** ** topping ** **** ******* ** ********.

*** ******** ** *** ***** ******** have ***** **** ***** ** ***** is ************ ******** ** *** ** their ********** ** ** *** **** redact ***** ******* *** ****** ** be ******** *** ** *** ******* knowledge ** *** *** *** ******* throughout *** ******** *** **** *** more ******* ****** ******** (***) *** made ** ************* **** **** **** to ******* ***** ******** ******.

****** **** **** ** ** ***** an ********** ******* ** £*,*** * month, ***** ****** **** ********* ****** to *** ******, ***** *** ** attractive ** **** ********** **** **** gain ********** **** ** *** *** many ***** *** ****** ** ***** utilized.

** ** ******* *** ********* *** point ** *** ** * ******* the *****, ** ** * **** cheap ********* ****** ********** *** **** a **** ***** **** ** *** any ****** ** ********** ***** **** your ************ ** **** ****** ** any ***.

***** **** ***-** ***** *** ******** to ***** **** ********, ** ***** countries *********** ********* ** *** **** similar ********, ***** *** ******** ** other *******, ******** ***** *** ***** competitive ***********.

**** ************* **** ********* ** ****** rival ********* ********* *** ** **** be *********** ** *** *** ****, what ******* *** **** ******* ******* in ****.

Comments (7)
U
Undisclosed #1
Nov 29, 2018
IPVMU Certified

Call me a stickler, but for $5,000 a year they can’t spell “Identity” correctly on the main screen?

(2)
(4)
RS
Robert Shih
Nov 29, 2018
IPVM

That's actually a me typo... Not them. First video I ever made...and now I have to fix it. Burned the midnight oil, worked hard on the script and didn't notice that the title was wrong. Sorry guys, I'll do better.

(1)
(1)
WH
Waqas Hassan
Dec 05, 2018

As an updated I would like to advise that Facit has released a new iteration of the software in the past week. We are pleased that the application now offers, under the Tools icon a screen grab function.

We are conscious that many large estates have a number of legacy systems within them and aligning all these codec’s is difficult and frustrating for operators, the tools we offered for this alignment process in earlier versions remain in place, though now complimented with the addition of the screen grab, for those systems which prove more challenging to align.

We continue to listen to feedback and work to make Identity Cloak as user friendly as possible.

 

Chris Selwood

Facit Data Systems

(1)
RS
Robert Shih
Dec 06, 2018
IPVM

Thank you for this update. We hope to stay informed as privacy concerns in recorded video gain more attention.

Avatar
Skip Cusack
Jan 08, 2020

IPVM, can you comment on technical performance in any more detail please? For example, what is the probability that a face will be blurred given a particular false alarm rate? How often did you see a face erroneously not blurred? Or not completely blurred? Any numbers here would be great, thanks.

JH
John Honovich
Jan 08, 2020
IPVM

what is the probability that a face will be blurred given a particular false alarm rate?

We avoid giving numeric figures on analytic performance, you'll notice this across our tests. The reason is that it gives a false sense of precision. We focus on calling out specific phenomenons that cause issues and let readers than assess the likelihood of those phenomenon being applicable to their use case.

Avatar
Skip Cusack
Jan 08, 2020

Okay thanks John. I'm trying to bound the limitations of this technology, or at least identify them. Would it be safe to say that the false positives and negatives are a function of how many subjects are surveilled at once, and what the ambient conditions happen to be? It seems like to blur the faces of a busy sidewalk you'd need to throw a lot of processing at it. And would you expect that a busy, confused scene with occlusions and challenging tracks on dozens of people would likely lead to more false alarms?