Medium severity · published May 20, 2026
CVE-2026-32792
Packet of death with DNSCrypt
5.3
CVSS
0.3%
EPSS
No
On KEV
Description
NLnet Labs Unbound 1.6.2 up to and including version 1.25.0 has a denial of service vulnerability when compiled with DNSCrypt support ('--enable-dnscrypt'). A bad DNSCrypt query could underflow Unbound's DNSCrypt packet reading procedure that may lead to heap overflow. A malicious actor can exploit the vulnerability with a single bad DNSCrypt query that its decrypted plaintext consists entirely of '0x00' bytes and does not contain the expected '0x80' marker. Unbound would then start reading more bytes than necessary until it finds a non-'0x00' byte. Based on the underlying memory allocator and the memory layout, it could lead to heap overflow while reading followed by a crash. Likelihood of a crash is low, since it relies heavily on the underlying memory allocator and the memory layout. If the heap overflow does not happen, Unbound's later packet checks will deny the packet. Unbound 1.25.1 contains a patch with a fix to bound reading in the given buffer space.
CVSS vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
Attack vector
Network
Complexity
Low
Privileges
None
User interaction
None
Confidentiality
None
Integrity
None
Availability
Low
Scope
Unchanged
Exploitation
Is it being used?
Exploit prediction (EPSS)0.3% · higher than 27% of CVEs
Exploited in the wildNo evidence yet
CISA KEVNot listed
Ransomware campaign useNot known
Affected products
Which versions
UN
Unbound · NLnet Labs
1.6.2 before 1.25.1 · source: vendor, NVD
Affected product
UN
Unbound
NLnet Labs
More from Unbound
10.0
CVE-2026-42960
Possible cache poisoning via promiscuous records for the authority section
9.8
CVE-2026-82717
CNAME synthesis could lead to heap corruption
9.8
CVE-2026-81642
Heap buffer overflow and possible Remote Code Execution when digesting DNSKEY
9.8
CVE-2026-33278
Possible arbitrary code execution during DNSSEC validation
9.3
CVE-2026-50252
Possible cache poisoning attack by mapping source port population per thread