Let the Pirates Patch? An
Economic Analysis of Software
Security Patch Restrictions
Terrence August
*Joint work with Tunay I. Tunca
Internet Server Software Market
Motivation
Motivation
Code Red and the Problem
Code Red / Code Red II
Worm that attacks web servers running IIS
Installs back door and propagates 100 times over per infection
Distributed Denial of Service (DDoS) attack on www1.whitehouse.gov
Patch issued by Microsoft on June 18, 2001
Code Red worm strikes on July 19, 2001
$2.75 Billion in damages
Worm Date
Code Red 7.19.2001
Vulnerability
Notice
Motivation
Estimated Cost
($)
1 month 2.75 Billion
Slammer 1.25.2003
Blaster
Sasser
Zotob
8.11.2003
5.1.2004
8.13.2005
6 months
1 month
2 weeks
4 days
1.5 Billion
750 Million
14.8 Billion
$98K/company
(on average)
US-CERT Coordination Center
CERT Reported Incidents
160
140
120
100
80
60
40
20
0
1988 1993
Year
1998 2003
Motivation
Motivation
Microsoft (Windows Genuine Advantage)
Microsoft issues statement
Mike Nash (VP, Security Business and
Permit
Pirates
SP2
Restrict
Pirates counterfeit keys for obtain security patches
SP2 SP2
SP2 update
WGA
Permit
Pirates
Apr-04
May-04
Late
May-04
Jul-04 Sept-04
Feb-05
May-05
Motivation
Two Options
Make security patches available to all users
Network is more secure
Sasser worm: $14.8B
Slammer worm: $1.5B
Network effects
Restrict security patches only to legitimate users
Network is less secure
Curb piracy
Motivation
Motivation
Piracy in the Software Industry
Business Software Alliance (BSA) and International
Data Corporation (IDC)
Piracy rates
35% in 2004
Exceeds 75% in 24 countries
Economic Losses (globally)
$59B spent on packaged software
$90B+ installed
Motivation
Research Questions
Under high network security risk, should a software vendor make security patches readily available to all users?
Why might a vendor such as Microsoft allow pirates to patch security vulnerabilities?
Can piracy lead to less secure software products?
Are the arguments made by the security community that software vendors should “do the right thing” valid?
Literature Review
Economics of Info. Security and Piracy
Information Security Piracy
• Interdependent Security e.g., Kunreuther et al. (2002),
Kunreuther and Heal (2003, 2005),
Varian (2004), August and Tunca
(2006)
• Quantification of Losses e.g., Moore and Shannon (2002),
Cavusoglu (2004)
• Worm Spread Dynamics e.g., Weaver et al (2003) e.g., Peitz and
Waelbroeck (2003)
Key Observations
Software patching is costly
Losses from security breaches are positively correlated with valuations
Piracy tendencies vary across users
Model
Timeline
Vendor sets price and policy
Consumers make usage decisions
Vendor releases security patches /
Consumers make patching decisions
Worm attack realizes on network t = 0 t = 1 t = 2 t = 3
Model
Consumer Model
Consumer valuation space:
Consumer heterogeneity in regard to piracy:
Consumer action space:
Model
Costs and Losses
Effective cost of patching:
Loss from attack:
Expected cost of piracy:
Model
Consumer’s Problem
Consumer Market Structure
Consumer Market Structure
Equilibrium Characteristics
There is always a group of consumers who use but do not patch
There is always a population of users whose valuations are higher than the price but end up not purchasing the software
Users impose negative externalities on:
Other users
The software vendor
Consumer Market Structure
Pricing and Piracy
Pricing to deter piracy:
Two regions – August and Tunca (2006)
Region 1:
• Low price
1
Region 2:
• High price
1
0 0
vb
Consumer Market Structure
Threshold Characterization
Consumer Market Structure
Pricing and Piracy
Two policies which the firm can enforce:
Permissive policy:
“Let” the pirates patch
Restrictive policy:
Do “not let” the pirates patch
Let the Pirates Patch:
Unpatched population:
Consumer Market Structure
Consumer Market Structure
Let the Pirates Patch:
Four possible equilibrium market structures
Increasing security risk
Consumer Market Structure
Don’t Let the Pirates Patch:
Unpatched population:
Consumer Market Structure
Don’t Let the Pirates Patch:
Six possible equilibrium market structures
Increasing security risk
Vendor Profit Maximization
Profit Functions and the Vendor’s Problem:
Optimal Policy Decision for the Vendor
When to restrict security patches?
When to let pirates patch?
Results
Proposition 1: When to be restrictive
When the effective security risk is high, a software vendor can strictly increase his profit by restricting pirates from receiving security patches.
Common perception
Reduce the risk on the network
A more secure product benefits all users
Results
Don’t let them patch when…
Results
Let Do not Let
Proposition 2: When to be permissive
When the patching cost is not too high and the effective security risk is below a threshold value, a software vendor should permit pirates with access to security patches.
Contrast
Strong incentives to patch
Vendor wants to price high
Not willing to provide incentives for conversion
Increased usage due to reduction in negative network effects
Results
Let them patch when…
Results
Do not Let Let
Proposition 3
When the potential for piracy in a market is high, a software vendor should enforce a restrictive policy.
Candidates: Vietnam, Ukraine, China, …
Small size of low piracy tendency (Type L) population
When the potential for piracy in a market is high, a software vendor prefers a less secure product to a more secure product.
Results
Lack of Incentives for Secure Software
Results
Proposition 4
When the effective security risk is high and the patching cost is affordable to some users, the vendor’s optimal profit can decrease in the level of piracy enforcement.
Security Risk
Low High
Piracy
Enforcement
Low
High
Increasing
Increasing
Results
0.22
0.21
0.2
0.19
0.18
0.17
I
I'
0.16
0.15
0.14
0.13
0.12
0
II
0.1
Increasing Returns to Enforcement
III
0.4
0.2
d c d
0.3
0.5
Results
Proposition 4
When the effective security risk is high and the patching cost is affordable to some users, the vendor’s optimal profit can decrease in the level of piracy enforcement.
Security Risk
Low High
Decreasing
Piracy
Enforcement
Low
High
Increasing
Increasing Increasing
Results
Results
0.196
0.17
0.194
0.16
0.192
0.15
0.19
0.14
0.188
0.13
0.186
0.12
I
I'
II
II
III
III
0.15
0.14
0.13
0.12
0
0.22
0.21
0.2
0.19
0.18
0.17
I
I'
0.16
II
0.1
Increasing Returns to Enforcement
III
0.4
0.2
d c d
0.3
0.5
0.22
0.21
0.2
0.19
0.18
0.17
I
I'
0.16
0.15
0.14
0.13
0.12
0
II
0.1
Increasing Returns to Enforcement
III
0.4
0.2
d c d
0.3
0.5
Results
0.4
0.38
0.36
I
0.34
Impact of Piracy Enforcement on Social Welfare
II III
0.32
0.3
I'
0 0.1
0.2
d c d
0.3
0.4
0.5
Results
Decreasing Returns to Enforcement
0.206
0.204
0.202
0.2
0.198
0.196
0.194
0.192
0.19
0.188
0.186
0
II III
0.1
0.2
d c d
0.3
0.4
0.5
Results
0.33
0.32
0.31
0.3
0
0.36
0.35
0.34
Impact of Piracy Enforcement on Social Welfare
0.1
II
0.2
d c d
0.3
0.4
III
0.5
Results
Proposition 5
Security patch restrictions can be welfare superior to a permissive approach
When the patching cost and the effective security risk is low, social welfare can increase under a restrictive policy.
Results
Let the Pirates Patch?
Results
Concluding Remarks
Summary
Model of network software security with piracy
Role of incentives in setting security patch restriction policies
Explain patch restrictions under high security risk
Microsoft’s permissive policy
Security risk can be strategically used by vendors as a tool to convert pirates into legitimate users
Security patch restrictions do not necessarily reduce welfare