Ea Early Warning Zones Pr Prof. Raul Sanchez-Re Reillo (o (on - - PowerPoint PPT Presentation

Ea Early Warning Zones

Pr

• Prof. Raul Sanchez-Re

Reillo (o (on behalf alf of f Dr. Peter Wa Waggett)

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

1

22/11/2019

Gr Group O up Obj bjec ectiv ives es

To Define Challenges Faced by Implementers of Early Warning Zone Systems To Define the ‘State of the Art’ for Contributing Technologies To Highlight Areas of Work Needed to Make EWZ’s Practical

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

2

22/11/2019

Ch Challenges

Concept of Operation and Scenario’s for EWZ’s Capability of Biometric Technologies to contribute to EWZ’s Capabilities for AI to contribute to EWZ’s State of Standardization for EWZ technologies Data Protection Implications Implementation Check List for implementers of EWZ

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

3

22/11/2019

Co Concept of Operations s and Scenari rios

With overt biometric sensors, counter methods can be 100% effective (e.g. running past a camera).

• For truly high-risk scenarios, timely detection of failure-to-capture is essential.

Data describing the operational performance of systems is sparse and it’s difficult to make comparisons between different sources of information. Even with optimum performance settings, false alarm rates may be high.

• If it is used in a supporting role to a human-in-the-loop, it can be useful with acceptably

low hindrance to the well-willing audience.

• However, the definition of what is acceptable hindrance from surveillance systems is very

subjective, and is subject to active societal debate.

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

4

22/11/2019

Co Concept of Operations s and Scenari rios

Concepts of operations with cooperating subjects and overt biometric sensors are more realistic.

• Uncooperating (i.e. neutral and hostile) subjects should be processed without the use of

biometrics on the move.

Supporting technologies (such as algorithms that determine relevant image quality metrics) are available, but these are not standardized.

• CI operators and their LEA’s typically have to custom build biometric black-list

surveillance systems, which may lead to unexpected risks.

The ERNCIP TG on EWZ believes that there will be a need to revisit black-list biometric surveillance in any future work in this space.

• This report may serve as inspiration for the development of relevant supporting

technologies and standards.

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

5

22/11/2019

Bi Biometri ric Ca Capabilities

Work covered:

• Analysis of target applications
• Overview of Biometric System Components and Biometric Modalities
• Summary of Presentation Attack Detection and Evaluation of Biometrics

With this work the following partial conclusions have been obtained:

• For long distance recognition ( > 50m):
• Most of soft biometrics
• Physiological (e.g. height) and behavioral (e.g. movement tracking)
• Some adhered soft biometrics (e.g. clothing)
• Gait on video
• 2D Face
• Voice
• If considering cooperative environment: Gait on accelerometer or ECG

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

6

22/11/2019

Bi Biometri ric Ca Capabilities

• For recognition at vicinity (5m – 50m):
• In addition to long distance:
• 3D Face
• Adhere soft biometrics (e.g. visible tattoos)
• For Proximity Recognition ( < 5m):
• In addition to all others:
• Contactless fingerprint
• Iris on the move
• Vascular on the move

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

7

22/11/2019

Video Video and C and Cognit nitiv ive S e Sur urveillanc eillance

AI-powered automated video (e.g. CCTV) surveillance plays important role in EWZs

• Machine learning (e.g. DNNs), data analytics and artificial cognition and perception

ERNCIP studied main surveillance topics/challenges relevant to different EWZ use cases including

• Video analytics – e.g. re-identification, situational awareness and behaviour

analysis/anomaly/threat detection

• Multisensor, multimodal fixed and aerial platforms
• Automatic deception detection, video archive search, 4D visualization

as well as privacy considerations; benchmarking and standards.

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

8

Video Video and C and Cognit nitiv ive S e Sur urveillanc eillance

ENRCIP further identified significant academic, commercial and national and EU projects (e.g. DRS, CIP, FCT) addressing these topics

• However, remains significant gap between what is claimed by R&D community and what is

actually endorsed by authoritative users w.r.t. performance, especially false alarm rates

ERNCIP identified limitations and further R&D needed to attain level of maturity required for EWZ deployment

• Future research needed includes: advanced (unsupervised) machine learning and reasoning capability, multi

sensor (identity-aware) tracking; new threat recognition approaches; advanced video data mining methodology; multimodal deception detection; new visualization (e.g. 4D) approaches; sensing; aerial surveillance platforms; etc.

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

9

St Stan andar ardiz ization ion C Challe allenge

Standards are essential for interoperability, uniform quality in provision of services, reduction in costs, future-proofing and in enabling the EU security industry to be more competitive globally Multiple relevant standards bodies and working groups, including ISO, IEC, CEN and industry-led groups such as AFNOR and ONVIF.

• Scope of standards overlaps and are not always aligned

Categories of standards considered are:

• alarm systems
• multimedia
• forensics
• video surveillance
• risk management
• biometrics

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

10

St Stan andar ardiz ization ion C Challe allenge

Key standards currently in development:

• ISO IEC 30137 series: Use of biometrics with Video Surveillance Systems
• EN 62676 series: Video surveillance systems for use in security applications

Still huge challenges to be addressed:

• Performance - A particular concern relates to the measurement and reporting of the performance of

technologies in EWZ systems (both separately and in combination) and especially the lack of any EU wide certification scheme for surveillance systems

• Spoofing – Growing recognition of the problem, and in particular for EWZ applications, vulnerabilities with

biometric recognition at range

• Sample quality – Known to have a big impact on performance but what is meant by quality in EWZ

applications and how to measure it?

• Pace of change – Rapidly changing technology landscape means technical standards risk being obsolete

before they are published

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

11

Da Data Prot

• tection
• n Implication
• ns

Operators implementing EWZ approach need to take deep attention on the balance between the “need-for-privacy” and the “need-for-public-security”, according to Charter

• f Fundamental Rights of the EU, the General Data Protection Regulation (Reg. UE no.

2016/679) and the Law Enforcement Directive (Dir. UE no. 2016/680). Necessity and proportionality principles have to be assessed in concrete scenarios with a “privacy by design” approach. Policy makers can follow three different approaches to promote the implementation of EWZ approach:

• Fostering private autonomy on the basis of effective need of the Operators;
• Promoting code of conducts with a common approach and a governance assigned to

Data Protection Authorities within Member States;

• Making EWZ approach mandatory, with a clear risk based approach.

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

12

Da Data Prot

• tection
• n Implication
• ns

Operators shall perform a Security and Privacy Impact Assessment (SPIA) for governing and balancing the risk exposure of EWZs and privacy risk of data subject with a dynamic approach leveraging on live flows of information from different sources with the following benefits:

• enhance the overall security of critical infrastructures;
• Increase standardization of information structures, as they will need to be

shared among many different actors;

• Improve the protection of personal data by Operators, as they will be able

to modulate their security measure according to a dynamic balancing between security and privacy in any given moment.

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

13

Impl Implemen ementation Issues n Issues

Implementers of EWZ’s Need to be Aware of a Number of Specific Questions That They Will Need to Answer in Building a System. In Particular:

• Does Their System Strike the Correct Balance Between Legitimate Privacy and Safety and

Security Needs

• Are Implementers Treating Sensitive Personal Information Gathered from Biometric

Elements of the System Correctly

We Have Provided Questions and a Checklist to be Used by Implementers for Info and Audit Purposes

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

14

Pr Proposed Next Steps

To fully develop the conceptual model to approach CIP with compliance with data protection directives, while using adequate levels of security Technology is mature enough to leverage fully on the development of dynamic risk analysis

• The future promises to make it even more effective through technologies such as big data mining and AI

This approach to risk analysis can also respond to the legal requirements from GDPR

• This kind of analysis provides a continual modulation of the security measures implemented
• Primarily, the principles of necessity and proportionality could be addressed in nearly real-time with such

dynamic approach

Next Steps:

• Investigate further on the relationship between security and privacy, which holds the key to ensure that

public security is maintained at adequate level without forfeiting the values of a democratic society.

• Work with groups involved with producing ConOps, to outline practical areas for trials and testing
• Revisit black-list biometric surveillance to achieve further improvements with viable application

22/11/2019 EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

15

Ea Early Warning Zones

Pr

• Prof. Raul Sanchez-Re

Reillo (o (on behalf alf of f Dr. Peter Wa Waggett)

EARLY WARNING ZONES (3RD ERNCIP WORKSHOP)

16

22/11/2019