What Is Data Poisoning in Privacy?
In privacy work, data poisoning means introducing fake or altered personal details onto people-search sites and other databases so that the true details are hard to pick out. If an attacker finds ten different birthdays for you, they can’t tell which one is real. The method assumes you can’t get fully invisible, so it makes the visible record unreliable.
This is a different thing from the machine-learning use of the same words, where “data poisoning” means corrupting a model’s training data to degrade it. The two share a name and nothing else. There is also no standard definition of the privacy version. I use it the way my kidnapping article does.
The idea
An entirely invisible public profile is near impossible to maintain, and it looks suspicious to anyone investigating. A few believable false details do more work than nothing at all. The article’s example table shows the shape: an address that is close but wrong, a phone number with digits shuffled, a middle name changed, a birth date shifted. It is labeled illustrative because it is. None of it is a real person.
It also wastes time. Every wild goose chase is time an attacker is not spending on someone else.
Where it sits in the toolbox
Poisoning is the third option, after removal and burial. A privacy broker removes records where removal works. Where a record can’t be removed or keeps returning, they bury it with de-indexing. Where neither holds, they poison. It works best on records that are already public and will stay public.
Related idea: obfuscation
Researchers Finn Brunton and Helen Nissenbaum wrote a book on this family of tactics, Obfuscation: A User’s Guide for Privacy and Protest (MIT Press, 2015). They define obfuscation as the deliberate use of ambiguous, confusing or misleading information to interfere with surveillance and data collection. It is aimed at people who can’t opt out. Poisoning broker profiles is one case of it. See also counter-profiling.
Where it can backfire
- Terms of service and law. Submitting false information to some services can violate their terms or local rules. This is a decision for a person who understands the specific site, not a default.
- It can make real records harder to remove. If your removal request doesn’t match the listing, the site may not find the real one.
- It doesn’t replace removal. The true record is still out there unless it is removed.
- No guarantees. Nobody can promise it works every time, and I wouldn’t trust anyone who does.
What I leave out
I don’t publish how to do this on specific sites. The article says what and why. The how is a client engagement.