University of Virginia Linking the Economics of Cyber Security and Corporate Reputation Reverse Engineering of Rationale for Decisions Barry Horowitz University of Virginia January 19th, 2007 Center for Risk Management of Engineering Systems University of Virginia Outline • • • • • • Reverse Engineering Concept Breach Disclosure Laws Impetus for Research Methodology Results Conclusions Center for Risk Management of Engineering Systems 2 University of Virginia Reverse Engineering Actual Decisions Implied Values of the Decision Makers Multi-Objective Analytical Model for Decision Support Uses of Reverse Engineering Results Provide decision-makers an opportunity to reconsider Evaluate the values of others (competitors, adversaries, constituents) Center for Risk Management of Engineering Systems 3 University of Virginia Economics of Cyber Security • New Technologies = New Risks • Evolution of various cyber attacks – Short-term Disruptions: • Denial of Service Attacks • Viruses • Worms – Long-term Disruptions: • Loss of Reputation • Loss of Intellectual Property • Legal Liability • Substantial Internet Infrastructure Outages Center for Risk Management of Engineering Systems 4 University of Virginia Breach Disclosure Laws • Growth of e-commerce sector and companies’ growing dependence on the internet and digitized data has garnered attention to cyber security • A newspaper article publicizing a cyber security breach can: – Damage reputation – Damage consumer confidence – Damage supply chain relations – Lower revenues • Companies invest to minimize the probability of being highlighted in a news article by: – Increasing cyber investment – Keeping cyber breaches & corresponding impacts secret • Prior to 2003 - no laws enacted requiring security breach reporting Center for Risk Management of Engineering Systems 5 University of Virginia Breach Disclosure Laws • Recent events have led to a movement on the state and national level towards mandating companies to report on cyber breaches – California Security Breach Notification Law (July, 2003) – first state to enact legislation that requires any company operating within the state to report any compromise of private information to the affected parties – ChoicePoint Security Breach (February, 2005) – company announced that it had unwittingly sold the personal information of at least 145,000 Americans to identity thieves in 2004 Center for Risk Management of Engineering Systems 6 University of Virginia Federal Legislation • No direct mention of breach notification requirements, but gives authority to create them • Gramm-Leach-Bliley Act – Requires financial institutions to protect the security and confidentiality of their customers’ nonpublic personal information • Health Insurance Portability and Accountability Act (HIPAA) – Require health plans and health care providers to take appropriate safeguards to ensure the integrity and confidentiality of health information • Sarbanes-Oxley Act (SOX) – Authorizes the SEC to prescribe regulations requiring companies to report on the assessment of the security of information technology Center for Risk Management of Engineering Systems 7 University of Virginia State Legislation • • • 34 states currently have legislation enacted – California enacted legislation in 2003, other states follow by 2005 • 2003: 1 • 2004: 0 • 2005: 11 • 2006: 17 • 2007: 5 (1/07) Laws require responsible parties to report the breach to affected party and in some cases: – identify the likelihood of harm – offer assistance in limiting potential harm Out of the 34 states that have enacted legislation – 27 state laws apply to businesses within the state – 14 state laws apply to state agencies – 1 state law applies to insurers Center for Risk Management of Engineering Systems 8 University of Virginia • • • • • Breach Disclosure Laws Impetus for Research Methodology Results Conclusions Center for Risk Management of Engineering Systems 9 University of Virginia Bi-Products of Legislation • Bi-product of change in breach reporting - visibility to the press • Given that the press has interest in reporting cyber breaches, this gives visibility to the public • Thus, a company’s reputation now can be impacted in a manner that it hasn’t been in the past Center for Risk Management of Engineering Systems 10 University of Virginia Research Questions • Question Raised - How will companies invest in cyber security given its impact on their reputation and corresponding impacts on their revenues and profits? • We would like to understand: – How reporting laws could effect companies’ actions with regard to cyber security investments – The differences between various industries regarding how they relate cyber security investments and protecting their reputation: • Example: A bank would be more concerned with protecting its reputation and bolstering customer confidence through heightened cyber security than a manufacturing company. Center for Risk Management of Engineering Systems 11 University of Virginia • • • • • Breach Disclosure Laws Impetus for Research Methodology Results Conclusions Center for Risk Management of Engineering Systems 12 University of Virginia Methodology - Model Center for Risk Management of Engineering Systems 13 University of Virginia Methodology - Assumptions • β = current observed annual probability of a security breach being publicized, no differentiation among companies in the same sector • The added cyber security investment is made in the hope that the probability of a publicized cyber attack will be reduced to zero (α=0) • The value of K2 is the same from one company to another – Treat this in a manner similar to insurance • Rates are risk-based • Rates are the same from buyer to buyer when the risks are the same • Investment decisions are made on expected value analyses that compare costs with potential consequences of successful attacks Center for Risk Management of Engineering Systems 14 University of Virginia Methodology - Variables • β: # Companies (>5000 Employees) with Publicized Cyber Breach # Companies (>5000 Employees) in Industry – # companies with publicized cyber breach determined from online databases of published newspaper articles – # companies in industry determined from Census Bureau data • C: (% Revenue Spent on IT) * (% IT Spent on Cyber Security) – Percentages determined from Forrester Group reports • PM: – Financial data taken from Yahoo Finance and Morningstar.com Center for Risk Management of Engineering Systems 15 University of Virginia Methodology - Variables • K1: – Representation of how a company is concerned about its reputation with respect to its cyber security spending – K1 ratio quantitatively shows how much one industry believes cyber security has an impact on its reputation compared to another • K2: – Assume equal from company to company - K2 ratio = 1 • V: – Likely correlation with K1 ratio – If companies have different revenues at risk and one has a sense of it, it can be plugged into the equation Center for Risk Management of Engineering Systems 16 University of Virginia Methodology • Three industries compared: – Finance • Bank, Insurance, and Credit Sectors – Retail – Manufacturing • Three sets of results: – Reputation-based financial loss due to a news article: • Independent of the details of the breach • When breach impacts customers for the company’s products • When breach impacts company employees & supply chain partners • β’s calculated for period between October 1, 2005 and September 30, 2006 Center for Risk Management of Engineering Systems 17 University of Virginia • • • • • Breach Disclosure Laws Impetus for Research Methodology Results Conclusions Center for Risk Management of Engineering Systems 18 University of Virginia Results – β’s Center for Risk Management of Engineering Systems 19 University of Virginia Results – K1 Ratios Center for Risk Management of Engineering Systems 20 University of Virginia Results – V Ratio Ind Var K1 Ratios with V Ratio as Independent Variable 70 60 Unbiased - FvsR Unbiased - FvsM Unbiased - MvsR Customer - FvsR Customer - FvsRM SupplyC - FvsR SupplyC - FvsM SupplyC - MvsR K1 Ratio 50 40 30 20 10 0 0 1 2 3 4 5 V Ratio Center for Risk Management of Engineering Systems 21 University of Virginia Results - Interpretations • Unbiased Reader – β • Finance: .0648 • Retail: .0111 • Manufacturing: .0110 – K1 ratios • Finance allocates 6.72 and 3.37 times more than retail and manufacturing • Manufacturing industry allocates twice as much as retail Center for Risk Management of Engineering Systems 22 University of Virginia Results - Interpretations • Customers – No data for manufacturing – combined manufacturing and retail for analysis – β • Finance: .0605 • Retail: .0093 • Retail & Manufacturing: .0043 – K1 ratios • Finance allocates 7.52 times more than retail • Finance allocates 11.01 times more than retail and manufacturing combined – Financial institutions most concerned with reputation with customers – Retailers more with customer reputation than manufacturers • Retailers work more directly with customers, depend more on customer trust Center for Risk Management of Engineering Systems 23 University of Virginia Results - Interpretations • Supply Chain – β • Finance: .0086 • Retail: .0019 • Manufacturing: .0110 – K1 ratios • Manufacturing allocates 11.95 and 2 times more than retail and finance, respectively • Finance allocates 5.37 times more than retail – Manufacturers are willing to invest more to protect reputation with their partner companies and employees • Depend greatly on supply chain partners • Customers of manufacturers are often other companies Center for Risk Management of Engineering Systems 24 University of Virginia • • • • • Breach Disclosure Laws Impetus for Research Methodology Results Conclusions Center for Risk Management of Engineering Systems 25 University of Virginia Conclusion - Results • This is one analysis, but others could be conducted… – Example: different results likely from an analysis of reputation effects of policies concerning intellectual property protection • Results support the claims that: – A financial institution has greater concern about protecting against reputation-based financial loss due to publicized security breaches than a retailer or manufacturer – Closer to end customers → care more about negative publicity than suppliers to those companies • Policy makers should take into account the likelihood that different sectors will have different responses to certain policies Center for Risk Management of Engineering Systems 26 Future Work –Bringing in time as a Variable University of Virginia • Reputation-based financial effects seen as a function of time: – the actual attacks – the reporting of those attacks by law – the reporting of those attacks by the media • Policy makers must be wary of companies covering up security breaches Evaluating the alternatives of avoiding reporting and adding security • Assume companies cannot control the media • Can only reduce effects by: – Decreasing probability of an attack – Decreasing probability of an attack becoming visible to the public • Reducing visibility < reducing the probability of an attack? • Evaluating the behavior of the press as reported cases increase over time Center for Risk Management of Engineering Systems 27 University of Virginia Addressing Lack of Data • We try to understand decision-making even though we lack fundamental data: – Specific cyber security investments – Cyber attacks – Cyber attack financial effects • Using reverse engineering, we make inferences from limited available financial data, news articles, and prior research and data collection efforts • We hope our study encourages future research efforts related to reverse engineering of decisions, and that more innovative ideas emerge that can work around data limitations Center for Risk Management of Engineering Systems 28