
Threat Modeling for Journalists and Activists
Security advice usually starts with a tool: “use Signal,” “use Tor,” “use a VPN.” That’s backwards. The right tool depends entirely on what you’re protecting and from whom — skip that step and you’ll either burn enormous effort protecting against a threat that doesn’t apply to you, or leave a real one wide open.
Three questions, in order
1. What am I actually trying to protect? Be specific: a source’s identity, the contents of a document before publication, your location at a given time, the fact that two people have been in contact at all. Each of these calls for a different defense.
2. Who is the adversary, and what can they actually do? A local police department serving a subpoena to a US-based email provider is a very different adversary than a state intelligence service with network-level surveillance capability. Overestimating a weak adversary wastes effort; underestimating a strong one gets someone hurt.
3. What happens if I fail? For some stories, exposure means an awkward conversation. For others, it means a source loses their job, their freedom, or their life. The acceptable failure rate should shape how much friction you’re willing to accept in your own workflow.
Why this matters more than tool choice
Two journalists working the same story with the same adversary can reasonably make different tool choices and both be right, if their personal risk tolerance and workflow differ. Conversely, two people using identical tools can have wildly different real-world security if one has done this analysis and the other is just following a checklist. The Committee to Protect Journalists and the Electronic Frontier Foundation both publish structured worksheets for exactly this reason — the analysis is the product, not the tool list at the end of it.
A worked example
A local reporter investigating city contract fraud has a different threat model than a foreign correspondent embedded with a group under active government surveillance. The first might reasonably prioritize protecting source identity from a FOIA request or a subpoena — encrypted communication and careful metadata hygiene go a long way. The second may need to assume their devices could be physically searched at a border crossing, which changes the calculus entirely: full-disk encryption with a plausible passphrase, compartmentalized devices, and pre-planned “nothing to find” configurations become relevant in a way they simply aren’t for the first case.
The takeaway
Write your threat model down before you touch a single tool. It doesn’t need to be formal — three sentences answering the three questions above is enough to reveal whether your current setup is over-engineered, under-engineered, or actually about right.