A buffer over-read was discovered in libntlmauth in Squid 2.5 through 5.6. Due to incorrect integer-overflow protection, the SSPI and SMB authentication helpers are vulnerable to reading unintended memory locations. In some configurations, cleartext credentials from these locations are sent to a client. This is fixed in 5.7.

In Squid 3.x through 3.5.28, 4.x through 4.17, and 5.x before 5.6, due to improper buffer management, a Denial of Service can occur when processing long Gopher server responses.

An issue was discovered in Squid before 4.15 and 5.x before 5.0.6. An integer overflow problem allows a remote server to achieve Denial of Service when delivering responses to HTTP Range requests. The issue trigger is a header that can be expected to exist in HTTP traffic without any malicious intent.

Squid before 4.15 and 5.x before 5.0.6 allows remote servers to cause a denial of service (affecting availability to all clients) via an HTTP response. The issue trigger is a header that can be expected to exist in HTTP traffic without any malicious intent by the server.

An issue was discovered in Squid before 4.15 and 5.x before 5.0.6. Due to an input-validation bug, it is vulnerable to a Denial of Service attack (against all clients using the proxy). A client sends an HTTP Range request to trigger this.

An issue was discovered in Squid before 4.15 and 5.x before 5.0.6. Due to a memory-management bug, it is vulnerable to a Denial of Service attack (against all clients using the proxy) via HTTP Range request processing.

An issue was discovered in Squid before 4.15 and 5.x before 5.0.6. Due to incorrect parser validation, it allows a Denial of Service attack against the Cache Manager API. This allows a trusted client to trigger memory leaks that. over time, lead to a Denial of Service via an unspecified short query string. This attack is limited to clients with Cache Manager API access privilege.

An issue was discovered in Squid before 4.15 and 5.x before 5.0.6. Due to a buffer-management bug, it allows a denial of service. When resolving a request with the urn: scheme, the parser leaks a small amount of memory. However, there is an unspecified attack methodology that can easily trigger a large amount of memory consumption.

An issue was discovered in Squid through 4.13 and 5.x through 5.0.4. Due to improper input validation, it allows a trusted client to perform HTTP Request Smuggling and access services otherwise forbidden by the security controls. This occurs for certain uri_whitespace configuration settings.

Squid through 4.14 and 5.x through 5.0.5, in some configurations, allows information disclosure because of an out-of-bounds read in WCCP protocol data. This can be leveraged as part of a chain for remote code execution as nobody.

An issue was discovered in Squid before 4.13 and 5.x before 5.0.4. Due to incorrect data validation, HTTP Request Splitting attacks may succeed against HTTP and HTTPS traffic. This leads to cache poisoning. This allows any client, including browser scripts, to bypass local security and poison the browser cache and any downstream caches with content from an arbitrary source. Squid uses a string search instead of parsing the Transfer-Encoding header to find chunked encoding. This allows an attacker to hide a second request inside Transfer-Encoding: it is interpreted by Squid as chunked and split out into a second request delivered upstream. Squid will then deliver two distinct responses to the client, corrupting any downstream caches.

An issue was discovered in Squid before 4.13 and 5.x before 5.0.4. Due to incorrect data validation, HTTP Request Smuggling attacks may succeed against HTTP and HTTPS traffic. This leads to cache poisoning. This allows any client, including browser scripts, to bypass local security and poison the proxy cache and any downstream caches with content from an arbitrary source. When configured for relaxed header parsing (the default), Squid relays headers containing whitespace characters to upstream servers. When this occurs as a prefix to a Content-Length header, the frame length specified will be ignored by Squid (allowing for a conflicting length to be used from another Content-Length header) but relayed upstream.

Squid before 4.13 and 5.x before 5.0.4 allows a trusted peer to perform Denial of Service by consuming all available CPU cycles during handling of a crafted Cache Digest response message. This only occurs when cache_peer is used with the cache digests feature. The problem exists because peerDigestHandleReply() livelocking in peer_digest.cc mishandles EOF.

An issue was discovered in Squid before 4.12 and 5.x before 5.0.3. Due to use of a potentially dangerous function, Squid and the default certificate validation helper are vulnerable to a Denial of Service when opening a TLS connection to an attacker-controlled server for HTTPS. This occurs because unrecognized error values are mapped to NULL, but later code expects that each error value is mapped to a valid error string.

An issue was discovered in http/ContentLengthInterpreter.cc in Squid before 4.12 and 5.x before 5.0.3. A Request Smuggling and Poisoning attack can succeed against the HTTP cache. The client sends an HTTP request with a Content-Length header containing "+\ "-" or an uncommon shell whitespace character prefix to the length field-value.

An issue was discovered in Squid before 5.0.2. A remote attacker can replay a sniffed Digest Authentication nonce to gain access to resources that are otherwise forbidden. This occurs because the attacker can overflow the nonce reference counter (a short integer). Remote code execution may occur if the pooled token credentials are freed (instead of replayed as valid credentials).

An issue was discovered in Squid through 4.7 and 5. When receiving a request, Squid checks its cache to see if it can serve up a response. It does this by making a MD5 hash of the absolute URL of the request. If found, it servers the request. The absolute URL can include the decoded UserInfo (username and password) for certain protocols. This decoded info is prepended to the domain. This allows an attacker to provide a username that has special characters to delimit the domain, and treat the rest of the URL as a path or query string. An attacker could first make a request to their domain using an encoded username, then when a request for the target domain comes in that decodes to the exact URL, it will serve the attacker's HTML instead of the real HTML. On Squid servers that also act as reverse proxies, this allows an attacker to gain access to features that only reverse proxies can use, such as ESI.

An issue was discovered in Squid through 4.7. When handling the tag esi:when when ESI is enabled, Squid calls ESIExpression::Evaluate. This function uses a fixed stack buffer to hold the expression while it's being evaluated. When processing the expression, it could either evaluate the top of the stack, or add a new member to the stack. When adding a new member, there is no check to ensure that the stack won't overflow.

An issue was discovered in Squid through 4.7. When handling requests from users, Squid checks its rules to see if the request should be denied. Squid by default comes with rules to block access to the Cache Manager, which serves detailed server information meant for the maintainer. This rule is implemented via url_regex. The handler for url_regex rules URL decodes an incoming request. This allows an attacker to encode their URL to bypass the url_regex check, and gain access to the blocked resource.

An issue was discovered in Squid through 4.7. When Squid is run as root, it spawns its child processes as a lesser user, by default the user nobody. This is done via the leave_suid call. leave_suid leaves the Saved UID as 0. This makes it trivial for an attacker who has compromised the child process to escalate their privileges back to root.