An issue was discovered in Suricata 5.0.0. It was possible to bypass/evade any tcp based signature by faking a closed TCP session using an evil server. After the TCP SYN packet, it is possible to inject a RST ACK and a FIN ACK packet with a bad TCP Timestamp option. The client will ignore the RST ACK and the FIN ACK packets because of the bad TCP Timestamp option. Both linux and windows client are ignoring the injected packets.

An issue was discovered in Suricata 5.0.0. It is possible to bypass/evade any tcp based signature by overlapping a TCP segment with a fake FIN packet. The fake FIN packet is injected just before the PUSH ACK packet we want to bypass. The PUSH ACK packet (containing the data) will be ignored by Suricata because it overlaps the FIN packet (the sequence and ack number are identical in the two packets). The client will ignore the fake FIN packet because the ACK flag is not set. Both linux and windows clients are ignoring the injected packet.

In OISF LibHTP before 0.5.31, as used in Suricata 4.1.4 and other products, an HTTP protocol parsing error causes the httpheader signature to not alert on a response with a single \r\n ending.

An issue was discovered in Suricata 4.1.4. By sending multiple fragmented IPv4 packets, the function Defrag4Reassemble in defrag.c tries to access a memory region that is not allocated, because of a lack of headerlen checking.

An issue was discovered in Suricata 4.1.4. By sending multiple IPv4 packets that have invalid IPv4Options, the function IPV4OptValidateTimestamp in decode-ipv4.c tries to access a memory region that is not allocated. There is a check for o->len < 5 (corresponding to 2 bytes of header and 3 bytes of data). Then, "flag = (o->data + 3)" places one beyond the 3 bytes, because the code should have been "flag = (o->data + 1)" instead.

An issue was discovered in app-layer-ssl.c in Suricata 4.1.4. Upon receiving a corrupted SSLv3 (TLS 1.2) packet, the parser function TLSDecodeHSHelloExtensions tries to access a memory region that is not allocated, because the expected length of HSHelloExtensions does not match the real length of the HSHelloExtensions part of the packet.

An issue was discovered in Suricata 4.1.3. The code mishandles the case of sending a network packet with the right type, such that the function DecodeEthernet in decode-ethernet.c is executed a second time. At this point, the algorithm cuts the first part of the packet and doesn't determine the current length. Specifically, if the packet is exactly 28 long, in the first iteration it subtracts 14 bytes. Then, it is working with a packet length of 14. At this point, the case distinction says it is a valid packet. After that it casts the packet, but this packet has no type, and the program crashes at the type case distinction.

An issue was discovered in Suricata 4.1.3. The function ftppasvresponse lacks a check for the length of part1 and part2, leading to a crash within the ftp/mod.rs file.

An issue was discovered in Suricata 4.1.3. The function processreplyrecordv3 lacks a check for the length of reply.data. It causes an invalid memory access and the program crashes within the nfs/nfs3.rs file.

An issue was discovered in Suricata 4.1.3. If the network packet does not have the right length, the parser tries to access a part of a DHCP packet. At this point, the Rust environment runs into a panic in parseclientidoption in the dhcp/parser.rs file.

An issue was discovered in Suricata 4.1.3. If the function filetrackernewchunk encounters an unsafe "Some(sfcm) => { ft.newchunk }" item, then the program enters an smb/files.rs error condition and crashes.

An issue was discovered in Suricata 4.1.x before 4.1.4. If the input of the function SSHParseBanner is composed only of a \n character, then the program runs into a heap-based buffer over-read. This occurs because the erroneous search for \r results in an integer underflow.

Suricata version 4.0.4 incorrectly handles the parsing of an EtherNet/IP PDU. A malformed PDU can cause the parsing code to read beyond the allocated data because DecodeENIPPDU in app-layer-enip-commmon.c has an integer overflow during a length check.

Suricata version 4.0.4 incorrectly handles the parsing of the SSH banner. A malformed SSH banner can cause the parsing code to read beyond the allocated data because SSHParseBanner in app-layer-ssh.c lacks a length check.

The ProcessMimeEntity function in util-decode-mime.c in Suricata 4.x before 4.0.6 allows remote attackers to cause a denial of service (segfault and daemon crash) via crafted input to the SMTP parser, as exploited in the wild in November 2018.

An issue was discovered in Suricata before 3.1.2. If an ICMPv4 error packet is received as the first packet on a flow in the toclient direction, it confuses the rule grouping lookup logic. The toclient inspection will then continue with the wrong rule group. This can lead to missed detection.

Suricata before 4.0.5 stops TCP stream inspection upon a TCP RST from a server. This allows detection bypass because Windows TCP clients proceed with normal processing of TCP data that arrives shortly after an RST (i.e., they act as if the RST had not yet been received).

Suricata before 4.0.4 is prone to an HTTP detection bypass vulnerability in detect.c and stream-tcp.c. If a malicious server breaks a normal TCP flow and sends data before the 3-way handshake is complete, then the data sent by the malicious server will be accepted by web clients such as a web browser or Linux CLI utilities, but ignored by Suricata IDS signatures. This mostly affects IDS signatures for the HTTP protocol and TCP stream content; signatures for TCP packets will inspect such network traffic as usual.

Suricata before 1.4.6 allows remote attackers to cause a denial of service (crash) via a malformed SSL record.