Interdependent Privacy Games: The Case of Genomics

Mathias Humbert, Erman Ayday, Jean‐Pierre Hubaux, Amalio TelentiInfoscience 2015

Abstract

Over the last few years, the vast progress in genome sequencing has highly increased the availability of genomic data. Individuals can now obtain their digital genomic sequences at reasonable prices and choose to store, reveal, or share them. However, genomic data is highly privacy-sensitive and correlated between relatives. This paper uses a game-theoretic approach to study the tension between family members with differing opinions on data protection and disclosure. We model the interplay between selfish and altruistic relatives, defining closed-form Nash equilibria and extending the model to N players using multi-agent influence diagrams (MAIDs). Our results demonstrate that altruism does not always lead to more efficient outcomes and that high discrepancies in perceived benefits can negatively impact familial utility.

The Problem of Interdependent Privacy

As the cost of genome sequencing drops below $100, more individuals are uploading their genetic data to public databases like OpenSNP for ancestry or health insights. However, DNA is not just personal; it is inherently familial. Because DNA is highly correlated between relatives, an individual's decision to disclose their genome creates "interdependent privacy risks."

Inheritance of SNPs (e.g., rs191486604 in the Parkin gene) illustrates how a parent's genetic information directly reveals information about their child's health risks.

If a father shares his Single Nucleotide Polymorphisms (SNPs), he inadvertently reveals significant information about his children's predispositions to diseases like Parkinson’s. This creates a strategic tension: one relative might value the medical benefits of sharing, while another might prioritize privacy to avoid genetic discrimination.

Methodology: Modeling the Genomic Game

The researchers model this tension using two types of Interdependent Privacy (IDP) games:

  1. The Storage-Security Game ($G_s$): Players decide how much to invest in securing their genomic data on personal devices.
  2. The Disclosure Game ($G_d$): Players decide whether to publicly share their SNPs.

To quantify privacy, the authors use the expected estimation error of SNP values, derived using Markov Chains to model Mendelian inheritance. For complex N-player scenarios, they utilize Multi-Agent Influence Diagrams (MAIDs). MAIDs extend Bayesian networks to allow efficient computation of Nash equilibria by exploiting conditional independence, reducing complexity from $O(|\Omega|3^{2N})$ to $O(|\Omega|2^N)$.

Key Insights and Results

The study shifts the view of genomic privacy from an individual right to a collective familial good. One person's "utility" from sharing creates a "negative externality" (privacy loss) for their kin.

The Altruism Paradox

A surprising finding is the Altruism Paradox. Altruism—defined as caring about a relative's payoff—does not always improve social welfare. In some scenarios, a "familial factor" can lead to a "non-disclose" equilibrium that is actually worse for the family's total utility than if everyone had been selfish and shared their data.

Nash Equilibria and Strategy Alignment

The authors derived closed-form Nash equilibria for two-player games. They found that relatives with similar costs and benefits usually align their strategies (both secure or both don't).

Nash Equilibrium regions for the disclosure game based on the cost of disclosure (c). Different regions represent where players both disclose (1,1), both hide (0,0), or have mixed strategies.

In N-player simulations using the CEPH/Utah Pedigree 1463 dataset (focusing on 82,000 SNPs from Chromosome 1), a "cascade" effect was observed. If parents share their genomes, children and grandparents have less incentive to protect their own data because their privacy is already effectively compromised.

Limitations and Future Outlook

While the model provides a robust framework, it has limitations. It is currently static, whereas genomic decisions are dynamic as new family members are born. Furthermore, the model assumes sequential decision-making with perfect information, which may not reflect real-world family dynamics. Future work aims to map privacy loss to concrete monetary costs (e.g., insurance premiums) and extend altruistic modeling to larger N-player groups.

Ultimately, the "Price of Kinship" highlights that moving from selfish to altruistic decision-making does not always yield a windfall for familial welfare, especially when incentives are misaligned.

Summary updated

Abstract

Over the last few years, the vast progress in genome sequencing has highly increased the availability of genomic data. Today, individuals can obtain their digital genomic sequences at reasonable prices from many online service providers. Individuals can store their data on personal devices, reveal it on public online databases, or share it with third parties. Yet, it has been shown that genomic data is very privacy-sensitive and highly correlated between relatives. Therefore, individuals' decisions about how to manage and secure their genomic data are crucial. People of the same family might have very different opinions about (i) how to protect and (ii) whether or not to reveal their genome. We study this tension by using a game-theoretic approach. First, we model the interplay between two purely-selfish family members. We also analyze how the game evolves when relatives behave altruistically. We define closed-form Nash equilibria in different settings. We then extend the game to N players by means of multi-agent influence diagrams that enable us to efficiently compute Nash equilibria. Our results notably demonstrate that altruism does not always lead to a more efficient outcome in genomic-privacy games. They also show that, if the discrepancy between the genome-sharing benefits that players perceive is too high, they will follow opposite sharing strategies, which has a negative impact on the familial utility.
BibTeX
@inproceedings{humbert2015interdependent,
  author = {Mathias Humbert and Erman Ayday and Jean‐Pierre Hubaux and Amalio Telenti},
  title = {{Interdependent Privacy Games: The Case of Genomics}},
  booktitle = {Infoscience},
  year = {2015},
}
Figure