Safety and Security Pieter van Gelder Professor of Safety Science and TU Safety and Security Institute KIVI Jaarccongres 30 November 2016 1/50
Outline The setting Innovations in monitoring of, and dealing with safety and security issues How to model safety and security? The Safety and Security research agenda for the coming decade 2/50
The setting A world of new technology; robotics, internet of things, 3D printing, wireless data via wearables, social media and citizen participation for massive data collection, autonomous vehicles, new inspection techniques with remote sensing, new sniff machines, artificial intelligence, powerful microscopy, and nanotechnology, etc. At the same time we are being threatened by natural hazards, climate change, changing land use, security threats, cascading effects, anarchist cookbooks, ageing, etc. 3/50
Trends in society with an impact on safety and security 4/50
Real time satellite imagery in managing natural and man made hazards in society vimeo.com/92251790 5/50
http://richiecarmichael.github.io/sat/index.html# 6/50
Cloudy with a Chance of Big Data A constellation of tiny satellites may take the guesswork out of predicting tomorrow's weather. 7/50
Captured by satellite (Sept. 8 th 2015) 8/50
9/50
Better prediction of floods via Twitter 10/50
Drones as ongevalsregisseur 11/50
Big data analytics, artificial intelligence, machine learning 12/50
Dangers of risk profiling Based on ethnic profiling? Some ethnic groups are over represented in crime numbers. Does ethnic profiling reduce crime? 13/50
We have 5,120 investigators at our disposal (via Facebook) 14/50
Facial recognition 15/50
Issues What are the benefits of this new technology? (Improved safety, fewer incidents, Better Service?) How to deal with false positives (make algorithms more robust, with loss of performance, What level of false/positives is acceptable?) Differences between the use of this technology in the public and private space (Only by police authorities or also by private parties?) Legal issues (options for people outside the legal system to punish?) Perception of privacy degradation in relation to gender / age differences? Communication issues (facial recognition takes place for your and our security?) Possibility of whitelisting (use the technology only for desired persons?) 16/50
Robotisation 17/50
Statistics on fatalities with autonomous vehicles Tesla Model S, June 2016, the first known fatality in just over 200 million km where Autopilot was activated. Among all vehicles in the NL, there is a fatality for every 150 million km. Worldwide, there is a fatality approximately every 100 million km. 18/50
Proving the case for autonomous cars scientifically More data is needed (hundreds of billions of vehicle-km) to clearly demonstrate their safety In combination with Accelerated life testing (ALT) 19/50
Influence of robotisation on safety in society Innovations may cause new failure mechanisms, although other failure mechanisms may vanish. Example: the many car accidents caused by drunken drivers will disappear when all cars become robotized (or installing an alcohol tester) At the same time, a new failure mechanism (pattern recognition errors / software vulnerability / hacking) is introduced. If hackers can hack one Tesla, they may be able to hack all Tesla s and cause many accidents. 20/50
Problem of the many components and actors The manufacturer of the robot, who delivered an unsafe / not functioning system? The sub-contractor, who delivered an unsafe component to the manufacture of the robot? The operator of the robot, who didn t follow the instruction manual? The inspectorate, who didn t inspect the robot regularly? 21/50
Robotisation in warfare Leading to military personnel facing legal, ethical, and political dilemmas presented by modern drone warfare against those using terrorist tactics, and civilians who are endangered by it. 22/50
23/50
Models for safety and security Which models are being used? Which frameworks are there to judge the safety and security levels? How to weigh safety and security to other values in society? 24/50
A safety model in its most simple form Safety is the complement of risk, being probability times consequences. 25/50
Van Gelder, P.H.A.J.M., 1999. Statistical Methods for the Risk- Based Design of Civil Structures, Communications on Hydraulic and Geotechnical Engineering, ISSN:0169-6548 00-1 26/50
An Alpine case study (EU RAIN project) A transportation system consisting of 38 bridges in two parallel roads: 27/50
The system can be modelled by the following RBD (reliability block diagram) 28/50
And the following FT (Fault Tree) with 150 MCS (Minimum Cut Sets) 29/50
And the following BN (Bayesian Net) 30/50
Posterior failure probabilities of the bridges given sytem failure 31/50
Comparison of cost s CDF when considering all bridges (case 1) and without the top 5 most critical bridges (case 2). 32/50
Bayesian networks Can accommodate technical, human and organisational causes Can accommodate probabilistic influences (not only logical deterministic relations) Probabilistic influences: The occurrence of event X increases the likelihood that event Y also occurs X Y 33/50
BBN s for chemical plants 2000+ pieces of equipment: valves; vessels; pipes; detectors; actuators; etc. November 30, 2016 34 34/50
CATS Causal Model for Air Transport Safety 1400 nodes 5000 arcs air traffic control 35/50
Which frameworks are there to judge the safety and security levels? 36/50
Balance of safety and costs Dantzig, D. van, 1960, The economic decision problem concerning the safety of the Netherlands against storm surges; Report of the Delta Commission, Contribution II.2, p 59-110. 37/50
Overall probability of dying Probability of dying 1 0 20 40 60 80 100 120 140 0.1 (CBS data of 22 August 2013) 0.01 0.001 0.0001 38/50
How to weigh safety and security to other values in society? 39/50
A child that is loved goes by many names Multi Criteria Analysis Multi Criteria Evaluation (MCE) Multi Criteria Preference Analysis Multi Criteria Decision Making Multi Objective Evaluation These methods are essentially one and the same! MCE = Multi-criteria evaluation is primarily concerned with how to combine the information from several criteria to form a single index of evaluation 40/50
Concluding remarks Potential of technological innovations to increase the safety and security in society is enormous, but also think of legal, ethical implications Analysis based on Bayesian Belief Networks, incorporating multiple / combined failure mechanisms, and multiple actors is needed Security seems overlooked / will be a challenge in many domains where only safety was an issue until now 41/50
Thank you 42/50