See also: 01-security-incident.md · 03-deploy.md

Auditing untrusted files without running them

head -c 600 file.php          # first 600 *bytes* — safe peek at a huge or minified file
unzip -l index.zip            # list an archive's contents without extracting
unzip -qo index.zip -d DIR    # -q quiet, -o overwrite, -d extract into DIR (a scratch dir, never the project)

head -c beats head -n for minified/obfuscated files, where one “line” can be megabytes.

Finding injected code

grep -lniE 'eval\(|base64_decode|gzinflate|atob\(|fromCharCode' *.html *.php js/*
  • -l print only filenames, -n line numbers, -i case-insensitive, -E extended regex (so | means OR).
  • \( escapes the parenthesis, so eval\( matches the call rather than starting a regex group.
grep -hoE 'https?://[a-zA-Z0-9.-]+' *.html | sort | uniq -c | sort -rn

Extract every external domain and count it. -o prints only the match, and -h drops the
filename. sort | uniq -c | sort -rn is the classic “frequency table” pipeline. You must sort
before uniq, because uniq only collapses adjacent duplicates.

Diffing against a known-good backup

for f in backup/*.html; do b=$(basename "$f"); cmp -s "$f" "$b" && echo same $b || echo DIFF $b; done
diff -rq . cedsite            # -r recursive, -q only say *which* files differ

cmp -s is silent and just sets an exit code, so && ... || ... branches on it.

Finding files that don’t belong in a folder

find assets -type f ! -name '*.js' ! -name '*.css' ! -name '*.png' ...

! negates the next test. It’s a quick way to spot a .php hiding in an images folder.

find assets -size +20M -exec ls -lh {} \;

Find big files. This matters because GitHub rejects any single file over 100 MB.

↑ Back to Projects — Learning Logs