← The Breach Files

2020 · Data breach

Gravatar

Breached 3 October 2020 · Published 5 December 2021 · gravatar.com

113,990,759Accounts exposed
58/100Serious severity
3Types of data
VerifiedBy HIBP

On 3 October 2020, Gravatar was breached. 113,990,759 accounts were exposed, including email addresses, names, usernames. MyRecon rates it 58/100 — serious. That is roughly everyone in Germany.

Checked against this breach and every other on record, against public breach data only. Your address is not sent to us as a form and is not stored — it is handed straight to the lookup tool in your own browser. See the privacy policy.

There is a design lesson buried in this record that outlives the incident. Gravatar identified users by the MD5 hash of their email address, published in the page source of every site that showed an avatar. The reasoning was reasonable at the time: a hash is not the address, so the address is not being disclosed.

It does not hold. MD5 is fast, email addresses follow predictable patterns, and the total space of plausible addresses is small enough to work through. Of 167 million scraped hashes, 114 million were reversed to the original address. Hashing an identifier that an attacker can guess is not anonymisation; it is a lookup table waiting to be built.

Nothing sensitive was stored and nothing was stolen in the conventional sense — this was scraping a service that was answering exactly as designed. The exposure is the mapping itself: a name, a username and the email address behind it, for a hundred million people who had only ever wanted a profile picture to follow them between websites.

What happened

In October 2020, a security researcher published a technique for scraping large volumes of data from Gravatar, the service for providing globally unique avatars . 167 million names, usernames and MD5 hashes of email addresses used to reference users' avatars were subsequently scraped and distributed within the hacking community. 114 million of the MD5 hashes were cracked and distributed alongside the source hash, thus disclosing the original email address and accompanying data. Following the impacted email addresses being searchable in HIBP, Gravatar release an FAQ detailing the incident.

Breach description from Have I Been Pwned, used under a CC BY 4.0 licence.

Who was behind it

No party has been publicly confirmed as responsible, and this page will not name one. Most breaches are never formally attributed: data surfaces on a forum or inside a combined dump long after the intrusion, and the trail back to a specific actor is rarely made public. Where a group has claimed responsibility it is usually named in the account above — that claim is theirs, not a finding of ours.

What was exposed, and why it matters

What to do if you were in it

  1. Expect better-aimed phishing. A message that already knows your name and where you have an account is the whole point of a breach like this.

Questions people ask about this breach

Was my email address in the Gravatar breach?

Enter it in the box at the top of this page. MyRecon checks it against this breach and every other one on record, and the address is never stored.

What data was leaked in the Gravatar breach?

email addresses, names, usernames. Each one is explained above, along with what it means for the person it belongs to.

When did it happen, and when did it become public?

The breach is dated 3 October 2020. It was published to Have I Been Pwned on 5 December 2021, a gap of 429 days during which the data was already out.

Is the Gravatar breach real?

Yes. Have I Been Pwned lists it as verified, meaning the data was checked against the source rather than taken on trust.

How many people were affected?

113,990,759 accounts. That is roughly everyone in Germany. That is accounts rather than people — one person often has several.

Read next

Also in the archive

Case Files: how the landmark breaches actually happened →

Three kinds of content appear on this page and they are kept apart deliberately. The breach description is quoted from Have I Been Pwned under its licence. The severity score, the field-by-field explanation and the advice are computed from the record itself — every sentence derives from a number or a flag in it, which is what stops the analysis drifting from the evidence. Anything under the "MyRecon's take" byline is editorial: our reading of this breach, presented as opinion rather than as a finding.

← All breaches