(How) Do people change their passwords after a breach?

Sruti Bhagavatula, Lujo Bauer, Apu KapadiaarXiv 2020 · arXiv

Abstract

Security experts recommend immediate password changes following a data breach, yet empirical evidence on user compliance is scarce. This study utilizes longitudinal data from the Security Behavior Observatory (SBO) to track the actual behavior of 249 participants over two years. Analyzing responses to nine major breaches, we find that only 33% of affected users changed their passwords. Furthermore, those who did often created passwords with high similarity to the compromised ones or failed to update reused credentials on other sites. These findings highlight a significant gap between security recommendations and real-world user inertia.

Introduction

Password reuse is a systemic security flaw: a breach at a low-stakes site can grant attackers access to high-value accounts like email or banking. While companies issue notifications advising password resets after a breach, most research on user response has relied on self-reported surveys, which are prone to social desirability bias. This paper addresses the lack of empirical, longitudinal data by observing how users actually behave in the wild following a known security incident.

Methodology

The researchers leveraged data from the Security Behavior Observatory (SBO), a longitudinal study that instrumented the home computers of 249 participants between 2017 and 2018. To preserve privacy while analyzing behavior, the team used:

  • Browser Extensions: Custom tools for Chrome and Firefox captured salted hashes of passwords and estimated their strength.
  • Substring Hashing: To detect password similarity without plaintext access, the researchers hashed all three-character-or-longer substrings.
  • Heuristic Filtering: Algorithms were used to distinguish successful logins from failed attempts or secondary accounts.

Datasets and Breach Exposure

The study cross-referenced participant activity against nine major data breaches identified via Identity Force and Have I Been Pwned. As shown in the table below, Yahoo! was the most prominent breach affecting the cohort, accounting for the vast majority of affected participants.

Table 1: Distribution of participants across the nine analyzed breached domains.

Key Findings

The Response Gap and Procrastination

Out of 63 participants directly affected by a breach, only 33% (21 users) changed their password on the compromised domain. Even among this minority, action was slow; only 13% of affected users responded within three months of a public announcement. Most activity was driven by high-profile breaches like Yahoo!, suggesting that users only react when a brand is sufficiently recognizable to break through the noise.

The Illusion of Strength

While new passwords were, on average, 1.3× stronger on a log scale, many changes resulted in passwords of equal or lesser strength. Users frequently made minimal modifications—such as changing a single character—to satisfy system requirements rather than creating robust new secrets. As seen in Figure 1, breach-related changes often failed to exceed the average strength improvement seen in routine password updates.

Figure 1: Ratio of new to old password strength. Breach-related changes (red dots) often cluster near the baseline of 1, indicating minimal improvement.

Failure of the Ripple Effect

Perhaps most concerning is the lack of cross-site updates. Users who changed a password on a breached site typically had approximately 30 other accounts using similar passwords, yet they only updated an average of four of those accounts. This leaves the vast majority of their digital footprint vulnerable to credential stuffing attacks.

Conclusion and Future Work

The study reveals a profound inertia in user security behavior. 67% of users whose data was leaked took no action at all. The authors suggest that relying on advisory notifications is insufficient. Instead, they advocate for:

  1. Forced Password Resets: Mandatory resets by service providers following a confirmed breach.
  2. Browser Integration: Implementing password-reuse trackers directly into browsers to warn users when a compromised password remains in use on other domains.

Future research should expand the sample size and include mobile and macOS users to determine if these patterns hold across different demographics and platforms.

Summary updated

Abstract

To protect against misuse of passwords compromised in a breach, consumers should promptly change affected passwords and any similar passwords on other accounts. Ideally, affected companies should strongly encourage this behavior and have mechanisms in place to mitigate harm. In order to make recommendations to companies about how to help their users perform these and other security-enhancing actions after breaches, we must first have some understanding of the current effectiveness of companies' post-breach practices. To study the effectiveness of password-related breach notifications and practices enforced after a breach, we examine---based on real-world password data from 249 participants---whether and how constructively participants changed their passwords after a breach announcement. Of the 249 participants, 63 had accounts on breached domains; only 33% of the 63 changed their passwords and only 13% (of 63) did so within three months of the announcement. New passwords were on average 1.3x stronger than old passwords (when comparing log10-transformed strength), though most were weaker or of equal strength. Concerningly, new passwords were overall more similar to participants' other passwords, and participants rarely changed passwords on other sites even when these were the same or similar to their password on the breached domain. Our results highlight the need for more rigorous password-changing requirements following a breach and more effective breach notifications that deliver comprehensive advice.
BibTeX
@misc{bhagavatula2020how,
  author = {Sruti Bhagavatula and Lujo Bauer and Apu Kapadia},
  title = {{(How) Do people change their passwords after a breach?}},
  year = {2020},
  eprint = {2010.09853},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2010.09853},
}
Figure