trust issues: a certificate's fingerprint from its record
huntress ctf 2026, day 29 · metadata forensics · all three modes
tools used: archive and metadata readers and timed-attempt pipeline.
"certificates appear in binaries, packages, mail, captures, and disk images. fingerprint each one, collapse duplicates, and identify the trust claim unsupported by evidence." day 29 opened a certificate family for the lane: one pem-encoded x509 certificate alone in the evidence zip, and one question about it, its sha-256 fingerprint.
the key insight is what a fingerprint is a digest of. a certificate's fingerprint is sha-256 over its der bytes, the binary encoding inside the pem armour, never over the pem text itself. our readers already carried an x509 record with both fingerprints among its fields, so the question's words named the record and the key, and the first run of every mode answered it.
challenge shape
single 1 field the sha-256 fingerprint of certificate-001.pem
time trial 34 fields the certificate field plus the evidence-note
fields, the roster hashes, the cumulative
properties of days 4-28 and a synthesis digest order
mastery 442 fields the same 34 per set, over thirteen sets in one zip
the field was d29-x509-001-x509-sha256-fingerprint, "determine x509.sha256_fingerprint for certificate-001.pem", answer kind pattern: "exact 64-character hexadecimal sha-256 value; use no prefix, spaces, or explanatory text", graded case-insensitively.
artifact anatomy
the evidence zip holds one file, a pem certificate, which is the der encoding of the certificate base64-encoded between a begin line and an end line. inside the der, the structure is the standard one:
Certificate
tbsCertificate version, serial number, signature algorithm,
issuer, validity (not before / not after),
subject, subject public key info, extensions
signatureAlgorithm
signatureValue
the fingerprint covers the whole Certificate sequence, signature included, so two certificates for the same subject with different signatures have different fingerprints, which is what "collapse duplicates" will turn on when that question comes.
extraction method
by hand the field is three lines:
from cryptography import x509
from cryptography.hazmat.primitives import hashes
def sha256_fingerprint(pem: bytes) -> str:
cert = x509.load_pem_x509_certificate(pem)
return cert.fingerprint(hashes.SHA256()).hex()
the standard library gives the same digest without a parser: hashlib.sha256(ssl.PEM_cert_to_DER_cert(pem.decode())).hexdigest(). openssl x509 -in certificate-001.pem -noout -fingerprint -sha256 prints it too, upper case with a colon between bytes after a Fingerprint= label; drop the label and the colons to get the answer form.
the metadata readers hold the same as a record: subject and issuer as rfc 4514 strings and as named attributes, the serial number, both validity bounds, the version, the signature algorithm and the sha-1 and sha-256 fingerprints, with the pem armour stripped when present and der accepted as is. a question is answered from the record by its words, so x509-sha256-fingerprint names the record and its sha256_fingerprint key, and the family's next question (an issuer, a validity bound, duplicates by fingerprint) needs no new code.
validation and answer formatting
the reader parses the certificate rather than searching the bytes: a pem that does not decode, or der that does not parse as a certificate, is an error, never a guess. the answer is the question's own format, exactly sixty-four hexadecimal characters with no prefix, spaces or separators; we send the lowercase .hex() of the digest, and the grader ignores letter case. b466fdbf8b8b8cb2d0b632bd24eb875576b44edb85f18f90f53dcdea1a0ec781 for the single's certificate, e58eecae5b9b3cdf8fd67a5fcc3afb178231f3c1296de288396ae08443f1ed10 for the time trial's and fb40736635c3b7b4d9d2500bbd8c111c5dd9a750edf5dd304a7a45536889a571, f7a2a7ed0fa847a63a91f5e77648be50bbe45d44f82b89190e09e06fbc8493b8, 6d7705cde33e6ddc84c4c18a1f99db783040bd7aebd3437309652a29f30d9cde and ten more for the mastery's thirteen sets.
failed approaches and pivots
none on day 29.
automation across modes
the time trial and the mastery read the certificate field through the same record and every other field through the existing readers: 0.08 seconds for the time trial's thirty-four fields, 0.94 seconds for the mastery's thirteen sets.
lessons
- a certificate's fingerprint is a digest of its der encoding; strip the pem armour before hashing, and hash the whole certificate sequence.
- a new family's first question is one key of its main structure; a reader that exposes the whole record answers the family's next days by their words.
solve times
public scoreboard · day 29 finish: below top 10
- 0:03.3singlefirst blood, +3.6 s
- 0:07.6time trialfirst blood, +3.5 s
- 0:11.9masteryfirst blood, +4.4 s
- #61team day finishof 116 teams that cleared it, in 34:38.6
- #2single orderof the day's 7, by our flag time
from the public scoreboard: 09:00 et release to our flag; the table below counts from our attempt's start.
from the mode's first attempt start to its flag (server times). attempts counts our solver runs.
| mode | work to flag | attempts | wrong submits |
|---|---|---|---|
| single | 1.1s | 1 | 0 |
| time trial | 2.3s | 1 | 0 |
| mastery | 2.2s | 1 | 0 |
the clock
server timer of the winning attempt; used is attempt start to flag, left is what remained on the clock.
| mode | time available | time used | time left |
|---|---|---|---|
| time trial | 50.0s | 2.3s | 47.7s |
| mastery | 415.0s | 2.2s | 412.8s |
where we got stuck
- nowhere on day 29: the record already carried the field, and the first run of every mode read it.
flags: single, time trial, and mastery [redacted].
{generated by a clanker}