脸部识别及生物特征提取.pptx

《脸部识别及生物特征提取.pptx》由会员分享，可在线阅读，更多相关《脸部识别及生物特征提取.pptx（11页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、Facial Recognition,What is it and how does it work? 1. LEAs collect photo mugshots of arrestees and ask other government agencies (like the DMV or the State Dept.) that also collect photos of faces to share their info,Facial Recognition,What is it and how does it work? 2. The digital images are then

2、 converted into a mathematical representation of pre-designated measurements,Facial Recognition,What is it and how does it work? 3. The mathematical templates are uploaded into a common database,Facial Recognition,What is it and how does it work? 4. Cops take a “query” photo they want to ID, convert

3、 the image into a mathematical representation that can be compared with the known photos in the database(s), using facial recognition algorithms that rely on unique physical markers on your face to find the closest mathematical matches.,Facial Recognition,Where do cops get a query photo? In the fiel

4、d: Social media: CCTV: Smart City cameras:,How do I know if cops used FR in my case ? Look for the following factors which may indicate use of FR: An officer or witness took your clients photo before booking. Your client was arrested in an unusual location or without notice. You see certain buzzword

5、s in the police report.,Facial Recognition,Buzzwords to look for in reports: “photos from a police database” Next Generation Identification-Interstate Photo System (“NGI-IPS”) FBI database of nearly 73 million criminal records and over 53 million civilian records Facial Analysis, Comparison and Eval

6、uation (“FACE”) Services over 411 million non-criminal photos from DMV & passports Universal Control Number (“UCN”) ID # assigned to photo Repository for Individuals of Special Concern (“RISC”) “Candidate List” list of potential matches,Facial Recognition,How do I challenge FR evidence? Subpoena con

7、tract between the LEA, DOJ, and FR prgm developers to review limits in programming capabilities. Object to error rates and false positives: NGI purports to provide the “true candidate” in the top 50 profiles only 85% of the time and thats presumably only if the “true candidate” is even contained wit

8、hin the database leaving the system prone to false positives.,Facial Recognition,How do I challenge FR evidence? Consider filing a motion to compel the FR algorithm source code in order to search for flaws that may affect search results If in IL and a stock photo is taken from a private actor, take

9、advantage of the IL Biometric Information Privacy Act (740 ILCS 14/15, 15(b) that requires notice and consent before use of FR tech,Facial Recognition,How do I learn more? Read Govt Accountability Office report on FR: eff.org/FRGAO2016 EFFs 2017 Senate Testimony on FR: eff.org/FR2017 Georgetowns Law

10、s Report on FR: https:/www.perpetuallineup.org/ Ex. of law enforcements FR training: eff.org/FRLEAtraining Review of FR program flaws: eff.org/FRflaws NGI/RISC Privacy Impact Assessment: eff.org/FRPIA,Facial Recognition,Stephanie Lacambra Criminal Defense Staff Attorney 415-436-9333 x130 stephanieeff.org,