Third Annual

Workshop on Economics and Information Security (WEIS04)

Digital Technology Center, University of Minnesota

May 13-14, 2004

[About these Notes]

Thursday

Eric Riscorla - Is finding Security Holes a Good Idea?

Hao Xu -  Optimal Policy for Software Vulnerability Disclosure

Invited Talk: Dan Greer - The Economics of Shared Risk at the National Scale

Hal Varian -  Who Signed up for the Do-not-call List?

Alessandro Acquisti - Privacy and Rationality

Ashish Arora - Impact of Vulnerability Disclosure and Patch Availability

Rahul Telang – An Economic Analysis of Market for Software Vulnerabilities

George Danezis – Economics of Censorship Resistance

Roger Adkins – An Insurance Style Model for Determining the Appropriate Investment Level against Maximum Loss arising from an Information Security Breach

Friday

Andrei Serjantov – On Dealing with Adversaries Fairly

Michal Feldman – Free-riding and Whitewashing on P2P Systems

Stephen Lewis – Sufficiently secure P2P Networks

Joan Feigenbaum – Towards and Economic Analysis of Trusted Systems

Stuart Schecter – Towards Econometric Models of Software Security Risks From Remote Attacks

Richard Clayton - “Proof-of-work” proves not to work

Andy Ozment – Bug Auctions: Vulnerability Markets reconsidered

Nicholas Weaver - A worst Case Worm

 

Welcome by Andrew Odlyzko

            Welcome to the Digital Technology Center

            40 submissions, 17 papers accepted

 

Thursday, May 13, 2004

Eric Riscorla, Is finding Security Holes a Good Idea?

 

Background theory

Security holes: routine activity of finding and revealing flaws: bugtraq, etc

State of the world: bugs, and security bugs exist

Solution: find bugs, fix them

                        Find them before the bad guys, fix them

                        Also: corporate incentives, academic incentives, openness is good

Costs of vulnerability research: give attackers info about bugs

                        Vendors, sysadmins forced to implement patches

Standard of medicine care: first do no harm – show that treatment works

Driving research question: Does research for vulnerabilities payoff?

Simplified vulnerability discovery model

            White hat – bug found by good guy, patched & published

            Black hat – exploited, patched

Graph: # vulnerable machines & exploitation over time

            Black hat and white hat similar, but black hat has small window of private exploitation

WH better than blackhat BUT:

                        If bug is never rediscovered by BH, then better not to publish

Disclosure pays off if Pr(rediscover) * (costs) > (costs of public)

A model to find Pr(rediscover)

            Assumption in model: all bugs are equally likely to be found

            F bugs out of N found: probability of finding a bug is F/N

Data: data on the rate of bug discovery

            NIST ICAT metebase from multiple DBs: CVA, affected program, bug release time

            Data issues: noisy, only know about discovered bugs, heavily censored

Approach 1: Program level should expect reduced disclosure over time: less bugs per program

            No downward trend easily observed

            Laplace factors only shows a little in the last few quarters BUT censoring issues

Approach 2: Bug level – when was it put in SW, how long until discovered

            Should expect fewer disclosures each year

            Don’t see much, except for 1999

Approach 3: ignore censoring and consider recent developments – trend exists

            We do see a trend, but it’s very slow – approx 2.5 years

            Discount rate comes into play in considering actions in the future

                        3% à BH has to be 13% worse than WH discovery

                        Bigger differential for less

Conclusions

            Appear to be faced with an infinite stream of bugs, no dent being made in it

                        BUT: clearly not an infinite number, b/c that means 0 prob of a given bug

            If depletion is slow, is it cost effective?

            Automatic patching makes disclosure more attractive

            Faster malware makes disclosure less attractive

            No data-driven evidence that bug finding is good

                        It’s expensive to do, any effects seem small at best

                        Need better record keeping

            What about vendor incentives? 

Q: IBM did a study on old code in 80s – defect rate for fixes worse than new code

            Maintenance shouldn’t be preventative

A: Yes, assuming that patches are perferct, this makes it less attractive

 

Q: Difference b/n WH and BH bugs in data?

A: databases don’t say.  No evidence on how long to get from BH to WH data

           

Q: Analogy with oil: increased tech reduces bugs before release, so we have few bugs?

            Tech at finding bugs in post-release SW is getting better?

A: Probably not, also more effort

 

Q: Assumption in paper is that all bugs are equally bad.  Not true.

            Maybe even insert innocuous bugs as honey pots

A: Same results in data for “high severity”

 

Q: Over time, find more bugs, but less severity

            Risk thermostat – project failure is constant, for different scope projects

            -Less incentives for good code by not punishing vendor to make bug-free code

                        vendor pays for patches

A: Can use “buying up loose nukes” or spot auditing to enforce code w/out vulnerabilities

 

[top]

Hao Xu: Optimal Policy for Software Vulnerability Disclosure

With Ashish Arora and Rahul Telang

State of the world of disclosure policies

            Every party follows own interests

            Secrecy, Full disclosure, CERT (short term secret)

Goal: develop theoretic model for optimal policy for social utility

            Social planner perspective and interactions with vendors

Game-theoretic approach

            Vendor and social planner interact w/ disclosure time

            Vendor can lose customers, pass on patching costs to customers

            Vendor’s costs: patch development time +

            Loss to customers: depends on when social planner discloses

                        Patch before or after attacks

            Vendor’s decision: when vendor internalizes more customer loss, vendor delivers patch earlier

            If social planner reduces the disclosure window, vendor delivers patch earlier

                        Should enlarge the time window for the vendors – reduce patching costs

Comparing with a numeric example

            Social costs vs. internalization ration

            Optimal is better than secrecy and instant

                        Most of the time, secrecy is better than instance

            All models converge to low social costs as more is internalized

Extentions w model

            Diffusion of patching, patching quality

Conclusions

            Vendor will release later than socially optimal time

            Loss from exploitation trade off w/ delay in release of patch

 

Q: How is patching incentivized from the user perspective

            DRM: suspicious motivation

Q: Vendor in a position to know the impact of the patch?  Customers use it differently?

 

Q: Vendor liability?  Has it been tested?

A: No.

 

Q: EULA prohibits product liability

 

[top]

           

Invited Talk: Dan Greer, The Economics of Shared Risk at the National Scale

“The essence of security is really risk management”

            This field is the moment of maximum hybrid vigor

Hard to sell security – only willing customers are those who were attacked or will be audited

Ask the right questions (get the problem right!)

            What can attack a national infrastructure?

            What can be done about it?

State of the world

            Interdependence, location irrelevance (no safe neighborhoods on the net)

            Tech advances faster than public comprehension (avg clue is dropping)

            Info assets are in motion (where we should be looking anyway)

            No one owns the risk

            Computing is dynamic, Moore’s law- need to build this in

Economic state of the world:

Most efficient to attack the applications

                        Apps are federated: multiple security domains, more moving parts, jurisdictions

Perimeter defenses are diseconomic

Data on spendingVolume doubling every 30 months, getting much cheaper

The public cares about: spam, viruses, theft (identity, cycles)

            Privacy regulation has grown over the past 20 years (EPIC)

                        The public notices

Critical issues of Public Interest – (everything else is less important)

            Inherently unique asset: GPS, FAA’s broadcast ability, DNS

            Cascade failure: victims become attackers at a higher rate

Unique assets

            Concentrated data or communication

            Attack: Targeted attack of high power

            Counter: defense in depth of unit, replication of functionality

                        This is expensive, need to spend money

            This just requires will and good leadership inside insitutions

Cascade Failure

            Precondition: always-on monoculture

            Ignition: any exploitable vulnerability

            Counter: risk diversification

            Requires: resolve to create heterogeneous

Why ‘sploits matter

            Monoculture is a force multiplier

            Amateurs provide smokescreen for pros

            Are unknown  exploits held in reserve?

Is the absence of a serious event an evidence of zero threat, or just a failure to detect

Field repairs

            Patching won’t cure everything

            Due care vs. force majeur

            Attractive nuisance vs. unwitting accomplice

            Automatic update is brittling, not toughening

Predictions

            Private sector will treat traffic analysis as they did crypto

            Security and privacy will hit each other

            Meritocracy yields to governance

NSF assessment saw: low skill should be secure, info risk > fin risk

Metrics – adapt rather than create

            Public health, insurance, portfolio mgmt, physics, accelerated failure time testing

            Catastrophe bonds, value at risk

            Immunity is expensive

            Models can be found anywhere

            Process-based vs. goal based

                        Ordinal ordering may be good enough

Security is necessary but not sufficient for reliability: security is a subset

Security spend – how much of growth/profit can we spend on security

Valuing info is really hard, but it’s very valuable

Security requires preemption, which requires surveillance

            IT: freedom (default permit) yields to safety (default deny)

Numbers:

            Hosts * vulnerabilities: huge curve

                        BUT: greater than number of incidents?

            Are we doing a good job, or are the attackers satiated?

            Complexity is proportional to square of code,

            Epidemic modeling

                        Malcom Gladwell – tipping point

                        Small change in initial conditions,

                        Worst case disease:

                        Virulence is an adaptation against defense – spread if immune system is good

            If the epidemic model gets worse, may have to make patching mandatory

Policy solutions

            Closed code = single source of fixes

            User-level lock in

Q: Is it true that patching increases security?  Do we really have a market failure?

A: If we can’t say it isn’t, you’ll get bad regulation

            S-O is a security issue

 

Q: Patching makes the code base less diverse – we’re all using the same version now

A: Most attacked system is the one that’s one rev off current

            Make a program install more like Build, less like Copy

            Randomize instructions, implement microdiversity

 

Q: What about social pressure?  “Nice people don’t connect weak computers to the internet”

A: We don’t have a meritocracy.  How do you quantify performance standards?

            i.e. emissions standards

            Big fear: embedded systems that are networks but not updatable

                        à replacement levels (WMA requires updates)

 

[top]

 

Hal Varian -  Who Signed up for the Do-not-call List?

With Fredrik Wallenberg and Glenn Woroch

Interested in the demand for privacy

            Multi-dimensional

            Explicitly about “solitude”

Do not call (DNC) list – free signup, high penalties

            Big peak at the beginning, the last day à publicity

            Register by phone, web,

Research goals – what is correlated with DNC?

Household decision as a flow chart

            Aware that it exists

            Random utility model of household registration

Data – FOIA from FTC

            Area code + exchange of all numbers in census for all registrations

            Map to county-level numbers, got census data, info about telco

Analsysis - basic

            Frequency of signup for each county vs. demographic

            Assume constant fraction of each demographic signed up across counties

            Racial composition is important: more black = less sign up

            Income: very low and very high are sig. different, little variation for middle

            Age:  positive correlation with age until 50-60, then high rate

            High signup in central city, highest at farms

            Small differences in the middle of continuous areas

Analysis – grouped logit models

            Kitchen sink model: .75 r-squared

            Parsimonious: Income, kids, educ & state-merge à .7 r squared

Demand for do-not spam (Pew study)

            The same people who thought that telemarketing was a big deal were annoyed by spam

            Means we don’t have to rest too heavily on use of one or the other

Value of DNC registry

            $.10 annoyance per call x 104 million calls/day = 3.6 billion per year

            WTP – costs of

 

[top]

 

Alessandro Acquisti - Privacy and Rationality

With Jens Grossklags

Open questions in economics of privacy

            Do individuals care?  Can they protect themselves?  Should they?

Claim: framing the debate in terms of rational trade-off is misleading

Two different markets: market for personal information vs. market for privacy

Privacy trade-offs - Protection has immediate costs, uncertain benefits

            Incomplete info

            Bounded rationality

            Hyperbolic discounting

            Availability

Data – 100 questions

            Privacy attitudes, privacy behavior, market characteristics and psych distortions

            Knowledge of privacy risks, privacy protections

            Risk aversion

            Buy/sell value

            Hyperbolic discounting

Results

            General privacy, has it increased à high (as predicted)

            Concerns: more for marketing that price discrimination

            Limited awareness of gov’t monitoring

            40% of people think that CC doesn’t know about CC transaction

                        Actually, 36% answered “nobody” to “who else knows about transaction?”

                        à open-ended question

                        Methods: it’s part of the decision process

            56% were overconfident

            54% could not quote or describe a privacy act

            51% would not know what to do to browse anonymously

            74% of respondents took some action to protect their privacy

                        BUT – few actually use any specific one

            Test the infosec chocolate/password tradeoff

                        98% said no!  (as opposed to get through 71%)

            Buy lower than sell

                        Sell > expected loss 70%

                        Buy > expected loss 35%

 

[top]

 

Ashish Arora - Impact of Vulnerability Disclosure and Patch Availability

With Krishnan, Nandkumar, Telang and Yang

Motivation – understanding the optimal policy requires measuring

            Likelihood/frequency of exploit attempts, loss from exploits,

Patch diffusion rates

            Cost of patching

Data: attack data from honeypots

            Number of attacks/day for each of 308 vulnerabilities

            2772 obs. Over 9 weeks over course of year

            Vulnerabilities as either secret, published or patches

Results – effects of patching and publishing

            Difference-in-difference – publication & patching increase attacks by .02 attacks/day

            OLS – Disclosure increases attacks by .26, patching decreases by .5

                        Dummies of vulnerability attributes reduce coefs

            Tobit specification for the average of the real data

                        Vulnerabilities have big spike in attacks

                        Patching has immediate plummet before patch is released!

                                    Data artifact – we have to rely on DB”s patch release date

Majority of vulnerabilities are patched on disclosure, long tail

            CERT will informally delay publication if requested à publish after patch

Open source vendors are more likely to patch, and patch more quickly

            Bruce: could be a marketing hook

Parametric Model specs:

            Do they patch?

                        Instantaneous disclosure, large vendors, severity

 

Q: what about publishing a false vulnerability?

Q: Why does patching have this effect? (tobit?

           

Q: Did you observe any vulnerabilities that demonstrated this behavior? (tobit)

 

[top]

 

Rahul Telang – An Economic Analysis of Market for Software Vulnerabilities

With Karthik Kannan

Motivation – users voluntarily report vulnerability organization

            BUT – what if there was a market for vulnerability information

                        i.e. iDefence

Business model:

            Buy info from identifiers

            Help protect clients, defense from exploit codes

Assumptions

            Analyze market and CERT separately

Model parameters

            Infomediary pays Pb for each vulnerability, CERT doesn’t pay anything

            Infomediary charges Ps for each vulnerability, CERT doesn’t charge

Infomediary gives identifier incentive to find vulnerability first

            If identifier finds vulnerability, non-subscribers are hit

            If attacker finds vulnerability first, all are hit

Results of the model

            Benign identifier exerts negative externality on hackers

            If default prob. Of finding a vulnerability is low, then social benefit

            Need to define compensation as greater than the reputation capital

            Infomediary has incentive to release vulnerabilities to put non-subscribers at risk

                        Create incentive to subscribe à hurts others

                        Could be worse than no market at all

Conclusions – markets can work if we handle info properly

Q: What about public good effects?

Q: CERT could be seen as market in reputation, even though others may not want vulnerabilities published

Q: Security academics (like private schools) decrease social utility by increasing the supply of published bugs

 

[top]

 

George Danezis – Economics of Censorship Resistance

With Ross Anderson

Censorhsip issues – Scientology, copyright issues, libel laws

            First wave of attacks hit the centralization

            P2P model can resist this attack

Model:

            Nodes with heterogeneous prefs, same capability T

                        Some pref split between red and blue (i.e. liberal and conservative)

                        Nodes are happier to distribute resources in proportion r and b

            Discretionary Model – each node can choose what to distribute

            Random model – node does not have ability to choose

                        Policy set from system, reflect total system distribution

                        Each node has an incentive to shift towards their own preference

                                    Bad things, w/out censorship attempts

                                    Need a fair distrib: elections, ecash, rep,

Censorship – from an external censor

            Censor targets specific nodes

            Imposes particular distribution on each node

Resistance – a node can choose to resist

            Nodes have a defense budget t, subtracted from the ability to distribute resources

            Tradeoff b/n censorship resistance and distributing resources

Defense budget in discretionary model will cause more reaction than in the random model

Caveat:

Utility of each node is local, rather than global

Don’t model implementation costs or the censor

Discretionary models provide greater utility and higher local and global defense budget

 

Q: What about availability

A: In the discretionary model, you are vulnerable to this—could be good or bad

 

Q: Utility from being able to conceal preferences that’s in the discretionary model

 

Q: Censor to generate fake information  (i.e. seed bad music files)

A: No way to model this

 

Q: Conditions where censor would have to work less in random than discretionary?

A: Sometimes, you might not want to resist censorship

 

[top]

 

Roger Adkins – An Insurance Style Model for Determining the Appropriate Investment Level against Maximum Loss arising from an Information Security Breach

Traditional capital budgeting – select investment to maximize NPV

            BUT – change the level of risk, and thus the discount

Need an investment mode that has:

            Inclusion of time, distribution of losses, possible multiple breaches, discount rate

Conceptual model as a Binomial Option Pricing Model

            Either a net savings, or not

There is a breach, which can result in a loss or a net savings from protection

            Find investment with protection such that loss isn’t greater than a certain point

Assume a twin security S perfectly correlated with the probability of loss

            Construct a portfolio of S and a riskless security

Find a certainty equivalent for this portfolio

Can extend this mechanism for multiple vulnerabilities à Black-Scholes option pricing formula

What kind of twinning security exists from a security breach?

When you have high variability in distribution of losses, then investment will be expensive

As planning horizon gets longer, investement level gets more linear

            BUT – generally dealing with a short term vehicle

Rationality assumption

            Underinsurance if you haven’t had an incident

            Overinsurance if you have invested

Q: Internet as the “Land of Heavy Tails” – very rare, very bad incidents

A: Model doesn’t cut off tail, but tails are very long

Q: info a company would need to know would be about the distribution of losses

A: Yes—need to understand the distribution, not just the expected value

 

Q: How do you measure those losses?

A: That’s the problem, isn’t it?

 

Q: Can’t you use this option pricing to deterimine whether firms are spending as much as they should/ too much?

A: Similar to how we use financial assets, to reverse engineer things to review decisions

 

Q: What if variance of losses are a power-law style distribution with a heavy tail?

A: Some theoretical distributions do predict this

 

[top]

 

Andrei Serjantov – On Dealing with Adversaries Fairly

With Ross Anderson

Problem – a group of people need to make a decision from a group of preferences

            Generalized voting

            Preference aggregation: a social welfare function turns many prefs into one

                        Should be transitive, reflexive, etc

Criteria for good and bad

            Non-dictatorship

            Unrestricted domain

            Pareto (unanimity)

            Independence of irrelevant alternatives

            BUT – Arrow says that we can’t have all of these

                        Democracy doesn’t always work

Look at reputation systems w/ social choice theory

Aberer & Despotovic

            Reputation is a product of your complaints against and how many complaints

            BUT – if no one complains about you, or you never complain: perfect reputation

                        Solution: add 1 to the components before multiplying

            Not pareto – everyone thinking something doesn’t lead that as the outcome

Delloracas (2003)

            Users rate each other on 1-100, Outliers are detected and removed

            Ignores some votes!

            Involves inter-personal outcomes

            Why not just compute the median?

Kemeny-Young – allows fuzzy prefs to find optimal outcome

            Good results, reflects conderset preferences

Manipulability

            Gibbard-Satterthwaite: all schemes dictatorial or manipulable

            Manipulation might be computationally hard

Conclusions

            Economists have already looked at a lot of these aggregating issues

            Lots of things are impossible

            Some tools are directly borrowable

 

Q: Randomness moves you away from pure preference aggregation

 

Q: What about socio-cognitive trust?  It has been mathematically modeled.

 

Q: Econ assumes ordinal, not cardinal prefs.  Reputation can be thought of as a counting, rather than voting.  What about paying for votes/expression

A: Fairness issues,

 

[top]

 

Michal Feldman – Free-riding and Whitewashing on P2P Systems

With Christos Papadimitriou, John Chuang, Ion Stoica

P2P require voluntary actions, that may not be optimal for individuals

            Adar (2000) and Saroiu (2002) found empirical free riding – rational behavior

            Need to incentivize cooperation, punish free-riders

            Punishment requires reputation

                        BUT – cheap pseudonyms: Whitewash attach

Model – rational agents with a type of generosity level – U(0, t)

            Decide whether to contribute of free ride

            As more people constribute, marginal cost of participating declines

            Rational – contribute if marginal cost is less than generosity level

Graphical representation

            Intersection of distribution and 1/t

            A stable and an unstable equilibrium

                        Works a max generosity level that is high enough

Performance is the total benefit less that total cost

Penalty mechanism p < 1

            P could be a service diff. coef

            P also could be a probability of catch and exclude

            Both reduce burden on contributors and introduce threat to free-riders

            System performance improves

                        BUT – reduced benefit to free-riders

            If p>(1/a) – there is no social loss

                        It’s a threat, but no reduced performance

White-washing attack – free identity enables easy free-riding

            Need to extend dynamic model with entrance & exit

                        Some portion of those who leave return

            Impossible to penalize known free-riders

            Have to penalize newcomers – including new contributors

            With very low turnover rate, free identity has little impact

            Social loss increases as turnover increases

                        But – higher levels of generosity is still better than market

Conclusions – model of free riding

            Quantify penalty mechs and identity

                       

Q: Assume the turnover is exogenous, some % of departures cycle

A: It would be interesting to build arrival and departure into model

 

Q: What exactly can be observed?  Why not have a pricing mechanism

A: Model assumes that we know who is a free rider.  P can be thought of as accuracy of catching free-rider.  Based on contribution of each user.  Contribution and consumption might also be useful to meter.

 

Q; Incentives are negative?

A: Also works with positive incentives, same system.

 

Q: Actually, positive incentives can be targeted to

 

Q: Low-level participation that neither consumes or free-rides, but is just there—not a white-washer

A: need a non-binary decision

 

Q: Like MUDs – longer you play, the more you get, discourages free-riding

 

[top]

 

Stephen Lewis – Sufficiently secure P2P Networks

Most security models assume a very powerful attacker

            Try to model a cost-effective attack

Model – n nodes with d documents

            Attacker corrupts x percent of nodes

            Publisher wants to make sure at least one copy of document exists

            Attacker wants to get rid of all of them

            (perfect search, attack doesn’t affect document)

Attacker will either attack all of network of none of it if there is linear attack cost

            Simple utility comparison

            Publisher’s Best Response:

                        No attack – publish a single copy of the document

                        If there is an attack – no publish at all

Non-linear attack costs

            If d < exponent (few enough documents in the network)

                        Internal solution to maximize

                        Otherwise, all or nothing response

            No analytic best reponse function b/c of shape of function

            Can use a mixed strategy probabilistic solution for the attacker and whether to publish

Security of network depends on size and the payoff of censorship

            Linear cost – all or nothing

            Non-linear costs  - both expend some effort

 

Q: Cost of bringing down the system: take down first node, the rest are free.

A: This only captures that you compromise the node to stop publishing, as opposed to system itself.

 

[top]

 

Joan Feigenbaum – Towards and Economic Analysis of Trusted Systems

With Bergemann, Shenker, Smith

A research agenda, rather than presented results

How much are still interested, given market conditions

Love-hate relationship b/n content providers and the internet

            Low costs + high volume may not equal high profits

            Desire to keep the old prices, and the new costs

Trusted platforms

            Machines can prove to each other that they are running authorized software

            E.g. TCG, NGSCB

            Could enable remote control of data after it has been transferred to another machine

            Applications

                        Copyright enforcement / content business

                        ? Privacy Protection

                        ? Security-policy enforcement generally

            Permission-enforcement mechanisms

(Apparently, the industry is backing off things)

Multidisciplinary study

            Economics – would they be wildey adopted?

                        Are their economically better, feasible alternatives?

Two-sided markets – must be adopted by both content providers and content consumers

            Each market has externalities

            (only a little bit of recent literature about the externalities)

Timing and uncertainty

            Dependence on consumer valuations, early adopter driver

Information asymmetry – will vendors reveal security vulnerabilities?

Alternatives to trusted platforms

            (detection, rather than prevention)

            Customer gets specific permissions

                        Authorized use only is worse than

                        Flexible private use is worse than

                        Uncontrolled network use

            Suppose unauthorized action can be detected with probability q

                        Once caught, have to use trusted computing

            Participation constraint: utility less price greater for flexible

            Incentive constraint:

            Price difference will be positive if q is high enough

                        Intuitive: catch the really successful

Model may explain part of iTunes

            Allows flexible private use, but not full-scale network piracy

Will users prefer monitoring to copy prevention?

Technologically – how do you make q high enough?

Prevention vs. detecting mishaps and working around them

            This is a general security question

Open research agenda, may need to be re-examined in light of industry realities.

 

Q: Prob. Of catching vs. severity of punishment

A: focus on how you monitor things that well

            Need a cost-effective system to foist on cheaters

 

[top]

 

Stuart Schecter – Towards Econometric Models of Software Security Risks From Remote Attacks

Risk = (expected loss frequency) * (cost of loss)

            In security, most losses result from attacks

            Frequency of attack metrics

                        Number of individuals, incentives, personal risk

                        Costs of time, equipment, information

Measuring these today

            Frequency of attack as high, medium or low

            Strength is a sum of resource costs à find a vulnerability

History

            Computer Security Act of 1987

            Guidelines set by NIST

            Fault trees (attack trees) – calculate total risk as sum of specific ones

                        BUT – don’t really know leaf nodes

            Qualitative approach – low, medium & high

Desired metrics

            Total security losses – expected

                        Can determine return-on-investment

Lessons from the study of safety – forecasting uses historical assumptions

            Stationary model depends on natural rules and assumptions

            BUT Security doesn’t do this – attackers exploit us

Security regression models

            Indicators must be measurable and relevant

            Risk is positively correlated with adversary population and incentive to attack

            Risk is negatively correlated with personal risk and security strength

            Analog – Modelling burglary

                        Security strength not significant – it’s other risk factors

                        BUT – attacks to software are more diverse, bigger scale, time is important

Strength-based models

            Choice to attack as a fn of incentive to attack and strength

            Stationary model is unaffected as safeguards become obsolete

                        Data remains constantly protected as security changes

Certainty vs. applicability

Measuring strengths

            Costs of finding a vulnerability

Conclusion – risk management is essential, need measures of security strengths

            Can use markets for vulnerabilities

 

Q: What about malware?  How can you capture that?

A: They’re not terribly profitable

 

Q: There’s more than one coin.  Reputation, fun—you attacker could be earning non-monetizable actions

A: It’s so cheap to break in, and the hacker would rather be rich.

 

[top]

 

Richard Clayton - “Proof-of-work” proves not to work

With Ben Laurie

Spam as an economic problem – no charge for email

            $.01/email = $91 billion annually

            eCash has not happened

            513 million net users, 230 million hosts, 56% of emails are spam

                        50 emails/day avg, 60 spam/day

            32 billion spam/day

Proof of work: useless piece of work to show you care

            Dwork & Naur, Adam Back

            One idea: A hash with n leading 0s to show that they tried a lot of hashes

            Another: base it on memory speed, since there’s less variation

What about mailing lists?

            We don’t know how much

            Estimate: 40% of emails had same source, multiple destinations

            Adjust calculations: legit host must send 75 emails/day

Econ analysis

            Spammers charge b/n .001 and .03 cents to send

                        Spammers invest $50k, wants $30k/year

            Need to send 35,000 emails/day to break even

            Response rate to spam: .0023%

                        With 1/10 cent/email – ad costs $4.35/sale

                        Efficient for $50 mortgage lead, cellphone, pills

            Legit email: 0.7-1.6%

            Differentials b/n good guys and bad guys:

                        75 emails vs. 1750 emails/host

                        Possible “factor of 4” for work ability

                        Some room, but not a lot

Security analysis

            Lots of 0wned machines

            Currently easy to find a compromised machine spamming

                        BUT – suppose they were doing proof-of-work

            Trying to allow good guys to send 75 emails/day

            BUT bad guys can spam with 250 emails/host using 0wned machines

                        With a 1% of inbox as spam caveat

Real world email analysis

            People really do send a lot of email

Pure proof-of-work schemes don’t work

 

Q: what’s wrong with charging for email?

A: Changing the current system is expensive

            Spammers can still pay for that

 

Q: what about an escrow system where you get the money back if it’s not a spam

A: Too many insecurities.

 

Q: Why is white-listing hard?  What is inherently hard?

A: We don’t have the tools for doing it.  Can’t tell where the email actually comes from, and doing that is really bad for privacy—and hard to do. 

 

Q: Why not a simple hash-cash person-to-person?

A: compromised machines…

 

Q: Why not do it at the ISP list?

 

Q: Why wouldn’t Bruce’s subscribers pay for it?

A:

 

Q: Micropayments are too expensive to implement

 

[top]

 

Andy Ozment – Bug Auctions: Vulnerability Markets reconsidered

No good way to measure software security – market for lemons

Producer’s motivation for vulnerability markets

            Improved product quality

            Useful metric

Assumes vulnerabilities are ordered

Vulnerability Auctions – single buyer, many sellers

            Ascending, first price (reverse Dutch)

            Bidders are asymmetric – auction is not revenue equivalent

Conveys no information about the number of bidders

            Preferred by risk averse producers

            A reward is always offered – want to make sure that vulnerabities are purchased

Producers want to encourage testes to enter the auction

            High min bid

… Had to leave, and did not capture the rest of the paper.  Sorry.

 

[top]

 

About these notes

These notes were recorded on the fly by Allan Friedman, and any omissions or inaccuracies are purely his fault.  To learn more about the papers here, please see the conference website, or contact the authors directly.  Please contact me (allan_friedman at ksg.harvard.edu) for any corrections or clarifications.

 

I apologize for the horrendous formatting; I took notes in Microsoft Word and was lazy about dumping things into HTML, so plenty of nasty artifacts remain.

 

[top]