TL;DR
This paper provides the first granular financial performance analysis of the cybersecurity industry by breaking it down into 19 distinct sub-sectors. Using a decade of Crunchbase data and machine learning to fill in missing valuation gaps, the researchers found that while the industry performs on par with broader private equity, individual sectors like Blockchain and AI show massive variations in returns, ranging from 9.7% to over 177% annually.
What Problem Does This Paper Address?
Cybercrime is projected to inflict trillions of dollars in damages globally, yet the "defense" side of the equation—cybersecurity startups—remains a financial black box. Because most cybersecurity firms are private, investors and policymakers lack the transparent data needed to understand which technologies are actually maturing and providing returns.
Existing research often treats "cybersecurity" as a monolith or focuses only on the few firms that have gone public. This paper addresses the lack of granular, sub-sector-level financial metrics (like risk-adjusted returns and systematic risk) for the private equity landscape, helping to identify which specific niches—from biometrics to cloud security—are actually driving the industry's growth.
Methods
The authors faced a classic "missing data" problem common in private equity: many startups report funding rounds but hide their actual valuations. To solve this, they employed a two-step approach:
- Missing Data Interpolation: They used a machine learning regression approach (AutoSklearn) to estimate missing Post-Money Valuations (PMVs). They trained the model on features like the amount raised, time between rounds, number of investors, and geographical location.
- Financial Modeling: They adopted Cochrane’s (2005) log return model. This is a standard in venture capital research because it accounts for the "power law" distribution of returns (where most startups fail but a few "home runs" provide massive gains). They calculated:
- Alpha ($\alpha$): The return in excess of the market.
- Beta ($\beta$): The systematic risk, or how much the sector moves in sync with the broader stock market (S&P 500).
- Expected Returns: Both arithmetic and log-based annualized returns.
Datasets
- Crunchbase (2010–2022): The primary source, covering funding rounds, valuations, and exits (IPOs/acquisitions) for firms tagged with 19 cybersecurity-related keywords.
- S&P 500 (Yahoo Finance): Used as the public market benchmark.
- 3-Month T-Bill Rate (FRED): Used as the risk-free rate for calculating excess returns.
Key Insights and Ideas
- Cybersecurity is Not a Monolith: The most striking takeaway is the extreme heterogeneity. Treating the industry as a single category masks the fact that a Privacy startup and a Blockchain security startup have fundamentally different financial DNA.
- The "Safety" of Cyber: Most cybersecurity sectors exhibit contra-cyclicality (a $\beta$ well below 1). This suggests that cybersecurity is often viewed as a "must-have" utility; even when the broader economy dips, security spending and investment remain relatively stable.
- AI is the Exception: Unlike most "pure" security sectors, Artificial Intelligence startups in this space are highly pro-cyclical ($\beta > 1.6$). This indicates that AI investment is more tied to general tech market enthusiasm than to the specific defensive needs of the security industry.
- Geographic Specialization: While the US dominates overall, specific niches have different hubs. China captures over 80% of investment in Facial Recognition, while India dominates 98% of the QR Code security sector.
Key Results
- Top Returns: The Blockchain sector was the clear outlier with a staggering 177.27% annualized arithmetic return, likely fueled by the crypto boom during the sample period. Artificial Intelligence followed at 67.25%.
- Lowest Returns: Privacy (9.72%) and Biometrics (23.22%) ranked at the bottom, suggesting these sectors may be more mature or face higher regulatory/adoption hurdles.
- Valuation Leaders: The Private Cloud sector dominates in terms of average firm valuation (USD 447.22 million) and average capital raised (USD 45.46 million).
- Time to Exit: On average, it takes nearly 10 years for an E-signature startup to reach an IPO, whereas Blockchain startups reached that milestone in just under 3.5 years.
Limitations and Future Work
The authors acknowledge several caveats:
- Selection Bias: Crunchbase is more likely to record successful "unicorns" than quiet failures, which might inflate the absolute return figures.
- Sample Size: Some niche sectors (like Spam Filtering or QR Codes) had too few observations to perform the full log-model regression.
- Taxonomy Mapping: The tags in Crunchbase don't always perfectly align with formal industry classifications (like ENISA).
Future Work: The researchers plan to use Natural Language Processing (NLP) to better categorize firms based on their actual business descriptions rather than just tags. They also aim to incorporate more granular risk factors, such as the number of patents or GitHub repositories, to better explain why certain sectors outperform others.
Summary updated