diff options
Diffstat (limited to 'docs/introduction.md')
| -rw-r--r-- | docs/introduction.md | 119 |
1 files changed, 119 insertions, 0 deletions
diff --git a/docs/introduction.md b/docs/introduction.md new file mode 100644 index 0000000..f10c269 --- /dev/null +++ b/docs/introduction.md @@ -0,0 +1,119 @@ +# Silent Certificate Transparency + +This document introduces a silent Certificate Transparency monitor design. + +## Setting + +We consider a setting where one or more trusted _nodes_ request certificates for +a specified list of domain names. The domain names that a node requests +certificates for may overlap with the domain names of other nodes. For example, +there may be two distinct nodes that request certificates for a given domain. + + TODO: Add figure. + Figure 1: A few examples of how trusted nodes may be deployed. + +The threat we are worried about is certificate mis-issuance. Due to considering +a multi-node setting with overlapping domain names, no single node can be aware +of all legitimately issued certificates for the domain names that it manages. + +A certificate is considered mis-issued if it contains: + + 1. at least one domain name that any of the trusted nodes manage _but without + any of the trusted nodes requesting that certificate to be issued_, or + 2. at least one subdomain of the domain names that any of the trusted nodes + manage _unless that subdomain is explicitly specified as out of scope_. + +The cause of certificate mis-issuance can vary, ranging from BGP and DNS hijacks +to certificate authorities that are coerced, compromised, or actively malicious. + +## Goals and non-scope + +The goal is to detect certificate mis-issuance, not to prevent it. It is out of +scope to detect certificate mis-issuance that happened in the past. In other +words, if the architecture described herein is put into operation at time `T`, +then any certificate mis-issuance that happened before time `T` is out of scope. + +It is also out of scope to detect certificate mis-issuance that targets web +browsers without Certificate Transparency enforcement. This is because we +cannot get a concise view of all certificates without Certificate Transparency. + +To achieve the goal of certificate mis-issuance, we want a _monitor_ that: + + 1. _is easy to self-host_, because you trust yourself or can then find someone + else that is appropriate and willing to host your infrastructure, and + 2. _is silent_, so that there is little or no noise unless certificate + mis-issuance is suspected or other noteworthy log events are happening. + +## Assumptions + + - The attacker is unable to control two independent logs that count towards + the SCT checks in web browsers. So, we need not worry about split-views and + can just download the logs and verify that they are locally consistent. + - The nodes that request certificates start in good states but may be + compromised sometime in the future. Detection of certificate mis-issuance + is then out of scope for all domains that the compromised nodes managed. + - A mis-issued certificate will only be used to target connections from a + fixed set of fixed IP addresses. Any party that can distinguish between + certificates that are legitimate and mis-issued will never be targeted. + - A domain owner notices alerts about suspected certificate mis-issuance. The + monitor that generates these alerts is trusted and never compromised. + +## Architecture + +A monitor downloads all certificates that are issued by certificate authorities +from Certificate Transparency logs. The exact logs to download is automatically +updated using a list that Google publishes in signed form. All historical +updates to the list of logs is stored locally in case any issues are suspected. + +To filter out certificates that are not relevant, the monitor is configured with +a list of domains to match on. Only matching certificates will be stored, which +means there are nearly no storage requirements to run this type of monitor. + +To achieve the property of _silence_, the trusted nodes push legitimately issued +certificates to the monitor which listens for such requests using an HTTP API. +The monitor will use the feedback from these nodes to further filter the +downloaded certificates that matched based on which ones are legitimate. If any +certificates are found that no node pushed to the monitor, an alert is created. + +The communication channel between the trusted nodes and the monitor can be +tampered with. For example, it may be plain HTTP or an HTTPS connection that +the attacker trivially hijacks by obtaining yet another mis-issued certificate. +Owning that the communication channel is insecure helps avoid misconfiguration. + +A pre-shared secret is used for each node to authenticate with the monitor. +That secret is never shown on the wire (an HMAC key is derived from it), which +means that all a man-in-the-middle attacker can do is replay and block messages. +Accounting for replayed messages is either way a good thing, because it makes it +easy to overcome temporary issues like a spotty network, a monitor reboot, etc. +In other words, nodes can periodically (re)submit their legitimate certificates. +Blocked messages can not be solved by cryptography and will result in alerts. + +To keep the attack surface of the monitor down, one may optionally choose to +operate it as a Tor onion service, restrict access using WireGuard, or similar. +Or in the case of a single-node deployment, run everything on the same system. + +## Open questions + + - What do we do about certificates that appear in the logs after time `T`, but + which were issued with a `NotBefore` timestamp that is before time `T`? We + want to minimize noise, but also not open up for CA backdating threats. + - What options for generating alerts should be supported other than stdout, if + any? Most deployments already have ways to integrate with dashboards/email. + +## Further documentation + + - [HTTP API](./http-api.md): read more about how the trusted nodes + authenticate with the monitor to submit legitimately issued certificates. + - [State directory](./state.md): read more about how the trusted monitor keeps + track of its state as plaintext files in a given directory. Staying away + from a database is meant to ease debugging, silencing of alerts, etc. + +## Future ideas + + - Reduce the amount of bandwidth that the monitor spends downloading + certificates that are either way discarded (non-matches). This can be + achieved by introducing a _verifiable proxy_ supporting wildcard + (non-)membership proofs, see [verifiable light-weight monitoring][]. Ignore + the parts about changing the logs; that is easily solved by the proxy alone. + +[verifiable light-weight monitoring]: https://arxiv.org/pdf/1711.03952.pdf |