Three security events in one month reshaped our monitoring stack. Here's where Stackdog stands, and why it now installs with one command
Stackdog is not a weekend project. It is what we run on our own servers and with our own customers, and it has been through enough releases that the changelog is now the best account of its education. The short version
Stackdog is not a weekend project. It is what we run on our own servers and with our own customers, and it has been through enough releases that the changelog is now the best account of its education.
The short version of that education: October and November were brutal. PHP got two remote-code-execution bugs (PHP2Shell, late October), OpenSSH got an authentication bypass (SSHole, early November), and both were under automated exploitation within hours of disclosure. Each one taught the stack a new layer:
- PHP2Shell (v0.2.8β0.2.9) forced the version-check layer: which PHP builds predate the fix, on the host and inside every running container β and a detector for the exploit's actual footprint, the request that runs PHP out of memory while it is still being parsed.
- SSHole (v0.3.0) repeated the pattern for OpenSSH in days, and added the policy that one password login to a key-only account is already a foothold: first line bans.
- v0.3.1 generalized it: CVE watch now matches every installed package against OSV and CISA's exploited-vulnerability catalog, so next month's incident needs no release from us.
Underneath that sits what predates all of it: webshell scanning that correlates a file on disk with the request that drives it, an IP ban engine whose firewall rules are reconciled against its own database every pass, fail2ban-style SSH protection, and worker-crash watching.
What you can install today:
brew install trydirect/tap/stackdog
or
cargo install stackdog
stackdog init writes a starter config. Every setting has a safe default, so it runs with nothing configured at all. Alerts go to Slack, Discord, Telegram, email or a webhook.
It doesn't patch anything (upgrading is the fix), a ban arrives after the fact and mostly stops scanners and repeat attempts, bans are IPv4-only right now, and version-based CVE findings mean "a fix is missing", not "you are exploitable" distributions backport fixes. Every alert says to check your vendor's advisory.
The firewall list in the screenshot is from our own VPS, where the ban engine has been handling real scanner traffic since October.
Happy to answer questions about the detection side and if you've been hit by something the logs didn't catch, I genuinely want to hear about it.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.