Summary
Terrapin is a prefix truncation attack targeting the SSH protocol. More precisely, Terrapin breaks the integrity of SSH's secure channel. By carefully adjusting the sequence numbers during the handshake, an attacker can remove an arbitrary amount of messages sent by the client or server at the beginning of the secure channel without the client or server noticing it.
Mitigations
To mitigate this protocol vulnerability, OpenSSH suggested a so-called "strict kex" which alters the SSH handshake to ensure a Man-in-the-Middle attacker cannot introduce unauthenticated messages as well as convey sequence number manipulation across handshakes.
Warning: To take effect, both the client and server must support this countermeasure.
As a stop-gap measure, peers may also (temporarily) disable the affected algorithms and use unaffected alternatives like AES-GCM instead until patches are available.
Details
The SSH specifications of ChaCha20-Poly1305 (chacha20-poly1305@openssh.com) and Encrypt-then-MAC (-etm@openssh.com MACs) are vulnerable against an arbitrary prefix truncation attack (a.k.a. Terrapin attack). This allows for an extension negotiation downgrade by stripping the SSHMSGEXTINFO sent after the first message after SSHMSGNEWKEYS, downgrading security, and disabling attack countermeasures in some versions of OpenSSH. When targeting Encrypt-then-MAC, this attack requires the use of a CBC cipher to be practically exploitable due to the internal workings of the cipher mode. Additionally, this novel attack technique can be used to exploit previously unexploitable implementation flaws in a Man-in-the-Middle scenario.
The attack works by an attacker injecting an arbitrary number of SSHMSGIGNORE messages during the initial key exchange and consequently removing the same number of messages just after the initial key exchange has concluded. This is possible due to missing authentication of the excess SSHMSGIGNORE messages and the fact that the implicit sequence numbers used within the SSH protocol are only checked after the initial key exchange.
In the case of ChaCha20-Poly1305, the attack is guaranteed to work on every connection as this cipher does not maintain an internal state other than the message's sequence number. In the case of Encrypt-Then-MAC, practical exploitation requires the use of a CBC cipher; while theoretical integrity is broken for all ciphers when using this mode, message processing will fail at the application layer for CTR and stream ciphers.
For more details see https://terrapin-attack.com.
Impact
This attack targets the specification of ChaCha20-Poly1305 (chacha20-poly1305@openssh.com) and Encrypt-then-MAC (-etm@openssh.com), which are widely adopted by well-known SSH implementations and can be considered de-facto standard. These algorithms can be practically exploited; however, in the case of Encrypt-Then-MAC, we additionally require the use of a CBC cipher. As a consequence, this attack works against all well-behaving SSH implementations supporting either of those algorithms and can be used to downgrade (but not fully strip) connection security in case SSH extension negotiation (RFC8308) is supported. The attack may also enable attackers to exploit certain implementation flaws in a man-in-the-middle (MitM) scenario.
A regression in the way hashes were calculated caused rules containing the address range syntax (x.x.x.x - y.y.y.y) that only differ in the address range(s) involved to be silently dropped as duplicates. Only the first of such rules is actually loaded into pf. Ranges expressed using the address[/mask-bits] syntax were not affected.
Some keywords representing actions taken on a packet-matching rule, such as 'log', 'return tll', or 'dnpipe', may suffer from the same issue. It is unlikely that users have such configurations, as these rules would always be redundant.
Affected rules are silently ignored, which can lead to unexpected behaviour including over- and underblocking.
In order to apply a particular protection key to an address range, the kernel must update the corresponding page table entries. The subroutine which handled this failed to take into account the presence of 1GB largepage mappings created using the shmcreatelargepage(3) interface. In particular, it would always treat a page directory page entry as pointing to another page table page.
The bug can be abused by an unprivileged user to cause pmappkruupdaterange() to treat userspace memory as a page table page, and thus overwrite memory to which the application would otherwise not have access.
The implementation of TIOCNOTTY failed to clear a back-pointer from the structure representing the controlling terminal to the calling process' session. If the invoking process then exits, the terminal structure may end up containing a pointer to freed memory.
A malicious process can abuse the dangling pointer to grant itself root privileges.
An operator precedence bug in the kernel results in a scenario where a buffer overflow causes attacker-controlled data to overwrite adjacent execve(2) argument buffers.
The bug may be exploitable by an unprivileged user to obtain superuser privileges.
The BOOTP file field is written to the lease file without escaping embedded double-quotes, allowing injection of arbitrary dhclient.conf directives. When the lease file is subsequently re-parsed by dhclient, e.g., after a system restart, an attacker-controlled field from the lease is passed to dhclient-script(8), which evaluates it.
A rogue DHCP server may be able to execute arbirary code as root on a system running dhclient.
Incorrect packet validation allowed unbounded recursion parsing SCTP chunk parameters. This can eventually result in a stack overflow and panic.
Remote attackers can craft packets which cause affected systems to panic. This affects any system where pf is configured to process traffic, independent of the configured ruleset.
As dhclient is building an environment to pass to dhclient-script, it may need to resize the array of string pointers. The code which expands the array incorrectly calculates its new size when requesting memory, resulting in a heap buffer overrun.
A specially crafted packet can cause dhclient to overrun its buffer of environment entries. This can result in a crash, but it may be possible to leverage this bug to achieve remote code execution.
When exchanging data over a socket, libnv uses select(2) to wait for data to arrive. However, it does not verify whether the provided socket descriptor fits in select(2)'s file descriptor set size limit of FDSETSIZE (1024).
An attacker who is able to force a libnv application to allocate large file descriptors, e.g., by opening many descriptors and executing a program which is not careful to close them upon startup, can trigger stack corruption. If the target application is setuid-root, then this could be used to elevate local privileges.
When processing the header of an incoming message, libnv failed to properly validate the message size.
The lack of validation allows a malicious program to write outside the bounds of a heap allocation. This can trigger a crash or system panic, and it may be possible for an unprivileged user to exploit the bug to elevate their privileges.
An OpenSSL TLS server may crash if sent a maliciously crafted renegotiation ClientHello message from a client. If a TLSv1.2 renegotiation ClientHello omits the signaturealgorithms extension (where it was present in the initial ClientHello), but includes a signaturealgorithmscert extension then a NULL pointer dereference will result, leading to a crash and a denial of service attack. A server is only vulnerable if it has TLSv1.2 and renegotiation enabled (which is the default configuration). OpenSSL TLS clients are not impacted by this issue. All OpenSSL 1.1.1 versions are affected by this issue. Users of these versions should upgrade to OpenSSL 1.1.1k. OpenSSL 1.0.2 is not impacted by this issue. Fixed in OpenSSL 1.1.1k (Affected 1.1.1-1.1.1j).
OpenSSL could allow a remote attacker to bypass security restrictions, caused by a a missing check in the validation logic of X.509 certificate chains by the X509VFLAGX509STRICT flag. By using any valid certificate or certificate chain to sign a specially crafted certificate, an attacker could bypass the check that non-CA certificates must not be able to issue other certificates and override the default purpose.
Server or client applications that call the SSLcheckchain() function during or after a TLS 1.3 handshake may crash due to a NULL pointer dereference as a result of incorrect handling of the "signaturealgorithmscert" TLS extension. The crash occurs if an invalid or unrecognised signature algorithm is received from the peer. This could be exploited by a malicious peer in a Denial of Service attack. OpenSSL version 1.1.1d, 1.1.1e, and 1.1.1f are affected by this issue. This issue did not affect OpenSSL versions prior to 1.1.1d. Fixed in OpenSSL 1.1.1g (Affected 1.1.1d-1.1.1f).
A vulnerability was discovered in Linux, FreeBSD, OpenBSD, MacOS, iOS, and Android that allows a malicious access point, or an adjacent user, to determine if a connected user is using a VPN, make positive inferences about the websites they are visiting, and determine the correct sequence and acknowledgement numbers in use, allowing the bad actor to inject data into the TCP stream. This provides everything that is needed for an attacker to hijack active connections inside the VPN tunnel.
BZ2decompress in decompress.c in bzip2 through 1.0.6 has an out-of-bounds write when there are many selectors.
An information leakage flaw is found in the way Linux kernel saves and restores Floating Point Unit(FPU) state during task switch. There are two ways, one is to save & restore FPU state during task context switch. And second is to defer FPU state save & restore until an FP instruction is invoked by the current task. First is called as "Eager FPU Restore" and second is known as "Lazy FPU Restore" scheme.
Linux kernel which follows the "Lazy FPU Restore" scheme is vulnerable to the FPU state information leakage issue. An unprivileged local attacker could use this flaw to read FPU state bits by conducting targeted cache side-channel attacks, similar to Meltdown attack disclosed earlier this year.
Upstream fix: ------------- -> https://git.kernel.org/linus//58122bf1d856a4ea9581d62a07c557d997d46a19
References: ----------- -> http://www.openwall.com/lists/oss-security/2018/06/13/7
The ZFSIOCUSERSPACEMANY ioctl, used by zfs-userspace(8), truncated a 64-bit output buffer size to a 32-bit integer for the kernel allocation, but used the original 64-bit size as the buffer limit when writing records.
A local user with the "userused" delegated ZFS permission can trigger a kernel heap overflow via the ZFSIOCUSERSPACEMANY ioctl, potentially escalating privileges.
The ZFSIOCSETPROP ioctl, used by zfs-set(8), incorrectly validated the calling user such that an unprivileged user is able to set metadata on a dataset indicating that the dataset has received properties from a zfs-recv(8) stream.
Any local user can set the internal ZFS metadata flag "$hasrecvd" on datasets via ZFSIOCSETPROP.
The ZFSIOCRECVNEW ioctl, in the heal receive path, similarly truncated a 64-bit payload size to a 32-bit integer for allocation, then used the original 64-bit size as the length for a byteswap operation.
A local user with the "receive" delegated ZFS permission can trigger kernel memory corruption via ZFSIOCRECVNEW by sending a crafted receive stream in heal mode.
The RTSP handler in libalias rewrote outgoing packets into a fixed-length stack buffer without checking whether the rewritten data fit in the buffer, or whether the result fit back in the original packet.
A host sending crafted RTSP traffic from inside a NAT gateway using libalias can overflow a stack buffer, potentially achieving remote code execution in the kernel (when using ipfw(4) NAT) or in the natd(8) process (which generally runs as the root user).
The RACK setsockopt(2) handler drops the connection lock in order to copy option data from userspace, then reacquires the lock. After reacquiring, it verifies that the TCP stack had not been switched away, but did not reload its pointer to the stack's per-connection control block. If userspace switches stacks twice during this window, the check will succeed but the saved pointer will refer to freed memory.
The bug may be exploitable by an unprivileged local user to escalate privileges.
The kernel function that implements unlinkat(2) and funlinkat(2) validated the ATRESOLVEBENEATH flag but failed to pass it through to the underlying path lookup. The flag was silently dropped, so path resolution was not actually restricted.
A process that uses ATRESOLVEBENEATH with unlinkat(2) or funlinkat(2) to confine path resolution can in fact resolve paths above the starting directory. A caller relying on this flag for path containment may delete files outside the intended directory tree.
Pages belonging to largepage shared memory objects were not explicitly wired. When sendfile(2) transmitted such an object with the SFNOCACHE flag, it freed the underlying pages after transmission even though existing mappings still referred to them.
An unprivileged local user can abuse the bug to access freed kernel memory. This can be exploited to escalate privileges.
Certain system calls, such open(2) with the OTRUNC flag set, and fspacectl(2), could incorrectly free memory in largepage objects. These operations are not permitted on largepage objects, but the implementation did not verify this.
An unprivileged local user can abuse the bug to access freed kernel memory. This can be exploited to escalate privileges.
When auditing a system call executed via ptrace(PTSCREMOTE), the kernel passed the return value of an internal setup function to AUDITSYSCALLEXIT() rather than the actual result of the executed system call. As a result, committed audit records for system calls which returned an error do not reflect the true outcome of the operation. That is, they indicate that the operation succeeded when it in fact failed.
Audit records for system calls executed via ptrace(PTSCREMOTE) may show an incorrect error status. An attacker with the ability to debug a process could use this to produce misleading audit trails, potentially undermining audit-based Intrusion Detection Systems (IDS).
When building the iovec array for a received TLS 1.2 CBC record, ktlsocftlscbcdecrypt() incremented the iovec index for every mbuf in the chain, including mbufs that were skipped because they contained only TLS header bytes. This left uninitialized entries in the iovec array. The iovec array was allocated without zeroing.
A remote TLS peer can cause the kernel to read from uninitialized iovec entries during HMAC computation, resulting in a kernel panic. The peer must be able to control TCP segmentation such that the first mbuf of a CBC record contains only the 5-byte TLS record header.
When the PCM sample array needs to grow, the RACK stack doubles a 16-bit allocation counter. Repeated doublings caused the counter to wrap to zero, resulting in a near-zero-length allocation followed by a copy of the existing samples into the undersized buffer. A remote peer can trigger the expansion by sending a sufficient number of ACKs on an established connection.
A remote attacker who can establish a TCP connection to a system using the RACK stack can trigger an out-of-bounds write of approximately 512 KB in the kernel heap possibly resulting in a kernel panic.
A security regression (CVE-2006-5051) was discovered in OpenSSH's server (sshd). There is a race condition which can lead sshd to handle some signals in an unsafe manner. An unauthenticated, remote attacker may be able to trigger it by failing to authenticate within a set time period.
When msync(MSINVALIDATE) is called on a mapping of an unmanaged device object, the physical pages in the mapping range are marked invalid but remain in the pager's page list. A subsequent page fault will cause the fault handler to re-insert the page into the object's list. This corrupts the list, and on object destruction the page is freed twice.
An unprivileged local user with access to a device that provides memory-mapped I/O can trigger a use-after-free in the kernel, though this is limited to a pool of objects ("fictitious pages") that are never recycled for a different purpose. It may be possible to exploit this to escalate privileges.
During execve(2) of a SUID binary, the new virtual address space is installed before the process credentials are updated. During this window, a process running as the same user can access the target process's memory via procfs or linprocfs, because the kernel's debugging permission check still saw the original credentials.
An unprivileged local user can exploit this race to modify the address space of a SUID binary before its credentials are elevated, potentially gaining full control of the affected system.