- It was found that because NTP's access control was based on a source IP address, an attacker could bypass source IP restrictions and send malicious control and configuration packets by spoofing ::1 addresses.
- A denial of service flaw was found in the way NTP hosts that were peering with each other authenticated themselves before updating their internal state variables. An attacker could send packets to one peer host, which could cascade to other peers, and stop the synchronization process among the reached peers.
- A flaw was found in the way the ntp-keygen utility generated MD5 symmetric keys on big-endian systems. An attacker could possibly use this flaw to guess generated MD5 keys, which could then be used to spoof an NTP client or server.
- A stack-based buffer overflow was found in the way the NTP autokey protocol was implemented. When an NTP client decrypted a secret received from an NTP server, it could cause that client to crash.
- It was found that ntpd did not check whether a Message Authentication Code (MAC) was present in a received packet when ntpd was configured to use symmetric cryptographic keys. A man-in-the-middle attacker could use this flaw to send crafted packets that would be accepted by a client or a peer without the attacker knowing the symmetric key.
- The ntpd daemon truncated symmetric keys specified in the key file to 20 bytes. As a consequence, it was impossible to configure NTP authentication to work with peers that use longer keys. The maximum length of keys has now been changed to 32 bytes.
- The ntp-keygen utility used the exponent of 3 when generating RSA keys, and generating RSA keys failed when FIPS mode was enabled. ntp-keygen has been modified to use the exponent of 65537, and generating keys in FIPS mode now works as expected.
- The ntpd daemon included a root delay when calculating its root dispersion. Consequently, the NTP server reported larger root dispersion than it should have and clients could reject the source when its distance reached the maximum synchronization distance (1.5 seconds by default). Calculation of root dispersion has been fixed, the root dispersion is now reported correctly, and clients no longer reject the server due to a large synchronization distance.
- The ntpd daemon dropped incoming NTP packets if their source port was lower than 123 (the NTP port). Clients behind Network Address Translation (NAT) were unable to synchronize with the server if their source port was translated to ports below 123. With this update, ntpd no longer checks the source port number.
- This update introduces configurable access of memory segments used for Shared Memory Driver (SHM) reference clocks. Previously, only the first two memory segments were created with owner-only access, allowing just two SHM reference clocks to be used securely on a system. Now, the owner-only access to SHM is configurable with the "mode" option, and it is therefore possible to use more SHM reference clocks securely.
- Support for nanosecond resolution has been added to the SHM reference clock. Prior to this update, when a Precision Time Protocol (PTP) hardware clock was used as a time source to synchronize the system clock (for example, with the timemaster service from the linuxptp package), the accuracy of the synchronization was limited due to the microsecond resolution of the SHM protocol. The nanosecond extension in the SHM protocol now enables sub-microsecond synchronization of the system clock.