Security Bite is exclusively brought to you by Mosyle, Apple's only unified platform.Everything we do is about making Apple devices enterprise-ready and secure.Our unique integrated approach to management and security combines Apple-specific state-of-the-art security solutions for fully automated hardening and compliance, next-generation EDR, AI-powered Zero Trust and exclusive privilege management with the most powerful and advanced Apple MDM on the market.The result is a fully automated unified Apple platform that is now trusted by more than 45,000 organizations to get millions of Apple devices up and running effortlessly and affordably.Request your EXTENDED TRIAL today and see why Mosyle is everything you need to work with Apple.
Amid a huge EU fine levied against X earlier this month, Elon Musk announced that the platform's entire recommendation algorithm would be open source, apparently to help cool the regulatory waters by providing more transparency into how the social media giant organizes users' work schedules.
Usually, IT professionals see the news of something open source, smile, and move on with their lives.But last week I came across an interesting thread from none other than X that explains how the movement can expose anonymous sub-accounts for better or worse through "behavioral fingerprinting."
OSINT enthusiast under the handle @Harrris0n at X recently posted about what he found digging through the platform's now open-source recommendation code.His findings are a little scary if you're concerned about privacy or if you manage a whole network of bot accounts.
There was a "user action sequence" buried in X's repository.
This is not just a document.It is a point of change that encodes your entire history of behavior on the platform.Track the specific milliseconds that pause to scroll, the type of money that makes a block, the specific taste of the content that you are, and the season that interacts with it.It represents thousands of human-written content from the moment you saw your first cat post.
Now, this is where it gets weird.X uses this sequence to predict what will happen (essentially give priority to your platform) while creating a more reliable interface.
Harrison found that if you run this encryption on a known account and then compare it to thousands of anonymous accounts with what the repository calls "Isolation Candidates," you'll get a match.Very unusual matches, and he even provided the specific recipe needed to build this de-identification tool, and the barrier to entry here is very low.
According to his thread, all you need is an action sequence encoder (which the X repo just sent), an embed similarity search, and a bit of luck (lol).The only part most people are missing is training data for verified alternative accounts, but Harrison notes that he already has that from years of threat monitoring.
In theory, you could compare the same behavior from a public X user to an anonymous one, or even cross-platform between Reddit and Discord accounts.It goes to show that it's easy to change your username, but it's much harder to change your habits.
So, is a burner account truly anonymous?I'll let you decide.
I wanted to share this topic here on Security Bite because it's a good reminder that these algorithms often know you better than you know yourself.And that digital version of you is still vulnerable.
Subscribe to the Security Byte Podcast for bi-weekly deep dives and interviews with leading Apple security researchers and experts:
- Apple Podcasts
- Badrang (coming soon)
- Pocket Cast (coming soon)
FTC: We use automatic affiliate links to generate income.More.
