Usability and Security –
Why we need to look at the big picture
M. Angela Sasse
Professor of Human-Centred Technology
Department of Computer Science
University College London, UK
[email protected]
www.ucl.cs.ac.uk/staff/A.Sasse
Acknowledgements
• My Doctoral Students
– Anne Adams
– Sacha Brostoff
•
•
•
•
- Dirk Weirich
- Ivan Flechais
Marek Rejmann-Greene (BT Exact)
BIOVISION (EU Roadmap Project)
UK e-Science Programme
German Federal Office for
Information Security
2
Overview
• Yes, security needs usability
• Beyond UI design: changing user
behaviour
• Organisational factors
– Designing and maintaining security culture
– Taking responsibility
• Changing the development process
– Involving all stakeholders
– Changing the design/development process
3
4
Human Memory
• Limited capacity
• Decays over time (items cannot be
recalled at all or not 100% correct)
• Frequent recall improves memorability
• Unaided recall is harder than recognition
• Non-meaningful items much harder to
recall than meaningful ones
• Similar items are easily confused
• Items linger - cannot “forget on demand”
5
Password Systems
•
•
•
•
Require unaided recall
Entry must be 100% correct
Not meaningful
Many similar items compete
– Frequently change
– Proliferation passwords and PINs
(banking, phones, websites)
6
So users write them down …
7
Consequences of unusable
password systems
• Cost of secure re-setting (help
desks) is high
• security undermined by
– cheap reset techniques (reminders)
– user workarounds
• Organisations where everybody has
password problems are vulnerable to
social engineering attacks
8
Research challenges for
authentication
• Different mechanisms for frequent and
infrequently used passwords
• Make mechanisms more forgiving
– Replace “all or nothing” with more forgiving
mechanism
– Provide feedback and instructions
– Cued rather than unaided recall
– Keep password changes to a minimum
• while making it sufficiently secure
(including spouse-proof), AND
• providing universal access
9
Is biometrics the answer?
• Potential to reduce mental workload
• But
– No single biometric provides universal
access
– False rejection rates still high
– Current equipment has raft of usability
issues
10
Fingerprint readers
11
Which finger was it again?
12
Height adjustment
13
Iris – difficulty focussing
14
Where do I stand?
15
Where am I supposed to look?
16
Mental Models/Metaphors
• “Why Johnny Can’t Encrypt” [Whitten &
Tygar, Procs USENIX 1999]
• UI problems
• User Tasks not represented
• Misleading labels
• Lack of feedback
• Problem lies deeper:
• “key” cues the wrong mental model, usage of
“public” “private” does not match everyday
use of language
17
Research Challenges:
simple concepts or metaphors
• Move from “shorthand” metaphors for
security community to metaphors that
work for wider user base
• Adopt/adapt conceptual design approach
to make security concepts & tools more
accessible
– Identify suitable metaphors
– Engineer system and discourse to
communicate these
18
User Knowledge
• “Users are not the Enemy” [Adams &
Sasse, 1999]: employees’ and managers’
knowledge about security is sketchy
• Replicated by Fitzgibbons et al. [Univ. of
Colorado, 2003]
• “Namedropping” of security concepts and
unwarranted assumptions among
software developers [Flechais & Sasse,
2003]
• User education about security needed
19
User knowledge of security
“You know, if you think about, who’s
actually going to go through all that
struggle to hack a departmental computer
science account of some academic at
xxxx college. It’s not like NASA or
anything, nothing of interest.
“What would make it more likely?” Answer:
“Maybe if I was more famous, or…”
[laughs]
Weirich & Sasse, Procs NSPW 2003
20
“Adams & Sasse propose that educating
users in security is a solution for the
problem of chosing weak passwords.
They claim that if users receive specific
security training and understand security
models, they will select secure passwords
and refrain from insecure behaviour. In
our study, however, we discovered that
the level of security training did not
prevent users from choosing trivial
passwords and refrain from engaging in
insecure behaviour.”
[Dhamija & Perrig, 2001]
21
Education vs. Training
• Education not one-way information transfer
• Aim of Training: change behaviour
• Form good habits and change bad ones
– Checked to establish correctness, and provide
feedback
– Repeat and reinforce sufficiently often to form
habit
– Checked again after certain time to ensure
desired behaviours have been established, and
have desired effect
22
Changing User Behaviour
• More than increasing user knowledge:
• Changing user behaviour beyond their
interaction with the system
– Motivation
– Persuasion
– Social norms
23
Perceptions of, and attitudes
to security
• Weirich & Sasse, 2001
• “How would you describe a person
who cares about security?”
24
“People who would want to be more secure. I
don’t know. That’s really a question for
psychologists. What sort of people keep
their desks tidy. What sort of people comb
their hair in the morning.”
“People therefore who are obedient. People
who follow the crowd.”
25
“So, you could probably be changing
your password every week, for no
obvious reason apart from your
paranoia, whereas I am not terribly
paranoid about this sort of thing.”
26
Research Challenges:
Perceptions & Attitudes
• Ways of persuading and motivating users
to be secure
– Appeal to self-interest:
• Link security to goals that matter to people
– Economic impact
– Part of professional and ethical conduct
– Make threats believable, appear real
• Changing the image of security
– Social marketing
• Role models, “it’s cool to be secure”
• Persuasive technology (based on Fogg 2003)
– Can we make security fun?
27
Task Factors
• For most people, most of the time,
security is enabling task to one or more
production tasks
• Enabling tasks perceived as “hurdles” if
relevance to production task not clear
• Human nature to take short-cuts,
especially when workload is felt to be high
28
“We in engineering like to leave things
fairly unprotected so we can go and
access other people’s directories, so
if he people I’m working with are
changing files, I can work with their
latest revisions.”
From Fitzgibbons et al., Univ. of Colorado, 2003
29
Example: Passfaces
• Good recall rates even
after long periods of
non-use (90+ % after 3
months
• But: in field trial,
Passfaces users only
had 30% login
frequency of
password users
Brostoff & Sasse, Procs HCI 2000
Design for production tasks
• If competing with production tasks, it will
be eliminated/circumvented whenever
• Security behaviour most fit production
tasks
– No competing demands for user resources
(physical/mental workload)
– Cost of keeping out legitimate users: viable
contingencies
– Performance criteria: Speed, errors
31
Research Challenges:
Task design
• Develop security mechanisms that
can be configured to match
requirements of production tasks
• Support individuals and
organisations to identify and deal
with competing goals
• Resolve locally vs. globally optimal
solutions
32
Physical Context
• Outdoors usage is different
– Lighting, pollution, temperature, noise, etc.
– Limit performance especially of novel complex
mechanisms (such as  biometrics)
• Mobile and handheld systems
– Physical ease of use
• Nomadic use of devices and networks
– creates new threats
Research Challenges:
Physical Context
• Develop usable & secure mechanisms for
interaction with pervasive/ubiquitous
systems
– Specific?
– General?
– Multiple?
• Consider implications for
– physical & mental workload
– economic viability
34
Organisational Context
• Security culture
– “Do as I say, not as I do”
– Being able to violate “petty” security
regulations is badge of seniority
• Link security into business goals
• Design of specific (goal- and risk-based)
security policies that are enforced, and are
seen to be enforced
35
Research Challenges:
Organisations & Security
• Integrate security into organisation &
business model
– Socio-technical design approach
– Adapt safety-critical design approaches
[e.g. Reason 1990]
– Apply risk analysis and economic
principles to decision-making about
security
36
Cultural Context: Trust
• Social norms, key: trust
– People want to trust, and be trusted, but
this can create risks [e.g. Mitnick, 2002]
– But
• low-trust systems are expensive to run
[Handy, 1985]
• do not allow building of social capital
• May be counterproductive since people
are only defense against novel attacks.
37
Research Challenges:
Social Norms
• Identify social norms that may
interfere with desired security
behaviour
• Trust is a key norm
• Create clear conceptual basis for role
of trust in security systems
38
Conclusions
• Usable and effective security needs a
systemic approach
• Security technology, and improving user
interfaces to that technology, is by itself
not the answer.
39
Any Questions?
contact: [email protected]
www.ucl.cs.ac.uk/staff/A.Sasse
40
Descargar

Some Password Results