FacebookCVEs & Vulnerabilities

A pre-authentication remote code execution vulnerability exists in React Server Components versions 19.0.0, 19.1.0, 19.1.1, and 19.2.0 including the following packages: react-server-dom-parcel, react-server-dom-turbopack, and react-server-dom-webpack. The vulnerable code unsafely deserializes payloads from HTTP requests to Server Function endpoints.

Sending an HTTP request/response body with greater than 2^31 bytes triggers an infinite loop in proxygen::coro::HTTPQuicCoroSession which blocks the backing event loop and unconditionally appends data to a std::vector per-loop iteration. This issue leads to unbounded memory growth and eventually causes the process to run out of memory.

A privilege escalation vulnerability existed in the Below service prior to v0.9.0 due to the creation of a world-writable directory at /var/log/below. This could have allowed local unprivileged users to escalate to root privileges through symlink attacks that manipulate files such as /etc/shadow.

Prior to v176, when opening a new project Meta Spark Studio would execute scripts defined inside of a package.json file included as part of that project. Those scripts would have the ability to execute arbitrary code on the system as the application.

Katran could disclose non-initialized kernel memory as part of an IP header. The issue was present for IPv4 encapsulation and ICMP (v4) Too Big packet generation. After a bpf_xdp_adjust_head call, Katran code didn’t initialize the Identification field for the IPv4 header, resulting in writing content of kernel memory in that field of IP header. The issue affected all Katran versions prior to commit 6a03106ac1eab39d0303662963589ecb2374c97f

The React Developer Tools extension registers a message listener with window.addEventListener('message', <listener>) in a content script that is accessible to any webpage that is active in the browser. Within the listener is code that requests a URL derived from the received message via fetch(). The URL is not validated or sanitised before it is fetched, thus allowing a malicious web page to arbitrarily fetch URL’s via the victim's browser.

The HTTP/2 protocol allows a denial of service (server resource consumption) because request cancellation can reset many streams quickly, as exploited in the wild in August through October 2023.

A lack of input validation exists in tac_plus prior to commit 4fdf178 which, when pre or post auth commands are enabled, allows an attacker who can control the username, rem-addr, or NAC address sent to tac_plus to inject shell commands and gain remote code execution on the tac_plus server.

A use-after-free related to unsound inference in the bytecode generation when optimizations are enabled for Hermes prior to commit da8990f737ebb9d9810633502f65ed462b819c09 could have been used by an attacker to achieve remote code execution. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

netconsd prior to v0.2 was vulnerable to an integer overflow in its parse_packet function. A malicious individual could leverage this overflow to create heap memory corruption with attacker controlled data.

A bytecode optimization bug in Hermes prior to commit e6ed9c1a4b02dc219de1648f44cd808a56171b81 could be used to cause an use-after-free and obtain arbitrary code execution via a carefully crafted payload. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

A type confusion bug in TypedArray prior to commit e6ed9c1a4b02dc219de1648f44cd808a56171b81 could have been used by a malicious attacker to execute arbitrary code via untrusted JavaScript. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

A use-after-free in BigIntPrimitive addition in Hermes prior to commit a6dcafe6ded8e61658b40f5699878cd19a481f80 could have been used by an attacker to leak raw data from Hermes VM’s heap. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

A null pointer dereference bug in Hermes prior to commit 5cae9f72975cf0e5a62b27fdd8b01f103e198708 could have been used by an attacker to crash an Hermes runtime where the EnableHermesInternal config option was set to true. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

There is a vulnerability in the fizz library prior to v2023.01.30.00 where a CHECK failure can be triggered remotely. This behavior requires the client supported cipher advertisement changing between the original ClientHello and the second ClientHello, crashing the process (impact is limited to denial of service).

An error in Hermes' algorithm for copying objects properties prior to commit a00d237346894c6067a594983be6634f4168c9ad could be used by a malicious attacker to execute arbitrary code via type confusion. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

An error in BigInt conversion to Number in Hermes prior to commit a6dcafe6ded8e61658b40f5699878cd19a481f80 could have been used by a malicious attacker to execute arbitrary code due to an out-of-bound write. Note that this bug is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

HHVM 4.172.0 and all prior versions use TLS 1.0 for secure connections when handling tls:// URLs in the stream extension. TLS1.0 has numerous published vulnerabilities and is deprecated. HHVM 4.153.4, 4.168.2, 4.169.2, 4.170.2, 4.171.1, 4.172.1, 4.173.0 replaces TLS1.0 with TLS1.3. Applications that call stream_socket_server or stream_socket_client functions with a URL starting with tls:// are affected.

Anchor tag hrefs in Lexical prior to v0.10.0 would render javascript: URLs, allowing for cross-site scripting on link clicks in cases where input was being parsed from untrusted sources.

A vulnerability was found in zstd v1.4.10, where an attacker can supply empty string as an argument to the command line tool to cause buffer overrun.

DexLoader function get_stringidx_fromdex() in Redex prior to commit 3b44c64 can load an out of bound address when loading the string index table, potentially allowing remote code execution during processing of a 3rd party Android APK file.

An integer conversion error in Hermes bytecode generation, prior to commit 6aa825e480d48127b480b08d13adf70033237097, could have been used to perform Out-Of-Bounds operations and subsequently execute arbitrary code. Note that this is only exploitable in cases where Hermes is used to execute untrusted JavaScript. Hence, most React Native applications are not affected.

A write-what-where condition in hermes caused by an integer overflow, prior to commit 5b6255ae049fa4641791e47fad994e8e8c4da374 allows attackers to potentially execute arbitrary code via crafted JavaScript. Note that this is only exploitable if the application using Hermes permits evaluation of untrusted JavaScript. Hence, most React Native applications are not affected.

An out of bounds write in hermes, while handling large arrays, prior to commit 06eaec767e376bfdb883d912cb15e987ddf2bda1 allows attackers to potentially execute arbitrary code via crafted JavaScript. Note that this is only exploitable if the application using Hermes permits evaluation of untrusted JavaScript. Hence, most React Native applications are not affected.

It was possible to trigger an infinite recursion condition in the error handler when Hermes executed specific maliciously formed JavaScript. This condition was only possible to trigger in dev-mode (when asserts were enabled). This issue affects Hermes versions prior to v0.12.0.

Instagram iOS 106.0 and prior and Android 107.0.0.11 and prior user interface does not properly represent URI messages to the user, which results in URI spoofing via specially crafted messages

The Facebook Messenger app for iOS 227.0 and prior and Android 228.1.0.10.116 and prior user interface does not properly represent URI messages to the user, which results in URI spoofing via specially crafted messages.

By passing invalid javascript code where await and yield were called upon non-async and non-generator getter/setter functions, Hermes would invoke generator functions and error out on invalid await/yield positions. This could result in segmentation fault as a consequence of type confusion error, with a low chance of RCE. This issue affects Hermes versions prior to v0.10.0.

A type confusion vulnerability could be triggered when resolving the "typeof" unary operator in Facebook Hermes prior to v0.10.0. Note that this is only exploitable if the application using Hermes permits evaluation of untrusted JavaScript. Hence, most React Native applications are not affected.

HHVM supports the use of an "admin" server which accepts administrative requests over HTTP. One of those request handlers, dump-pcre-cache, can be used to output cached regular expressions from the current execution context into a file. The handler takes a parameter which specifies where on the filesystem to write this data. The parameter is not validated, allowing a malicious user to overwrite arbitrary files where the user running HHVM has write access. This issue affects HHVM versions prior to 4.56.2, all versions between 4.57.0 and 4.78.0, as well as 4.79.0, 4.80.0, 4.81.0, 4.82.0, and 4.83.0.

parlai is a framework for training and evaluating AI models on a variety of openly available dialogue datasets. In affected versions the package is vulnerable to YAML deserialization attack caused by unsafe loading which leads to Arbitary code execution. This security bug is patched by avoiding unsafe loader users should update to version above v1.1.0. If upgrading is not possible then users can change the Loader used to SafeLoader as a workaround. See commit 507d066ef432ea27d3e201da08009872a2f37725 for details.

Due to use of unsafe YAML deserialization logic, an attacker with the ability to modify local YAML configuration files could provide malicious input, resulting in remote code execution or similar risks. This issue affects ParlAI prior to v1.1.0.

Passing an attacker controlled size when creating an IOBuf could cause integer overflow, leading to an out of bounds write on the heap with the possibility of remote code execution. This issue affects versions of folly prior to v2021.07.22.00. This issue affects HHVM versions prior to 4.80.5, all versions between 4.81.0 and 4.102.1, all versions between 4.103.0 and 4.113.0, and versions 4.114.0, 4.115.0, 4.116.0, 4.117.0, 4.118.0 and 4.118.1.

A use after free in hermes, while emitting certain error messages, prior to commit d86e185e485b6330216dee8e854455c694e3a36e allows attackers to potentially execute arbitrary code via crafted JavaScript. Note that this is only exploitable if the application using Hermes permits evaluation of untrusted JavaScript. Hence, most React Native applications are not affected.

A regular expression denial of service (ReDoS) vulnerability in the validateBaseUrl function can cause the application to use excessive resources, become unresponsive, or crash. This was introduced in react-native version 0.59.0 and fixed in version 0.64.1.

An invalid free in Thrift's table-based serialization can cause the application to crash or potentially result in code execution or other undesirable effects. This issue affects Facebook Thrift prior to v2021.02.22.00.

The wp_ajax_save_fbe_settings and wp_ajax_delete_fbe_settings AJAX actions of the Facebook for WordPress plugin before 3.0.4 were vulnerable to CSRF due to a lack of nonce protection. The settings in the saveFbeSettings function had no sanitization allowing for script tags to be saved.

The run_action function of the Facebook for WordPress plugin before 3.0.0 deserializes user supplied data making it possible for PHP objects to be supplied creating an Object Injection vulnerability. There was also a useable magic method in the plugin that could be used to achieve remote code execution.

A packet of death scenario is possible in mvfst via a specially crafted message during a QUIC session, which causes a crash via a failed assertion. Per QUIC specification, this particular message should be treated as a connection error. This issue affects mvfst versions prior to commit a67083ff4b8dcbb7ee2839da6338032030d712b0 and proxygen versions prior to v2021.03.15.00.

When unserializing an object with dynamic properties HHVM needs to pre-reserve the full size of the dynamic property array before inserting anything into it. Otherwise the array might resize, invalidating previously stored references. This pre-reservation was not occurring in HHVM prior to v4.32.3, between versions 4.33.0 and 4.56.0, 4.57.0, 4.58.0, 4.58.1, 4.59.0, 4.60.0, 4.61.0, 4.62.0.

The unserialize() function supported a type code, "S", which was meant to be supported only for APC serialization. This type code allowed arbitrary memory addresses to be accessed as if they were static StringData objects. This issue affected HHVM prior to v4.32.3, between versions 4.33.0 and 4.56.0, 4.57.0, 4.58.0, 4.58.1, 4.59.0, 4.60.0, 4.61.0, 4.62.0.

The fb_unserialize function did not impose a depth limit for nested deserialization. That meant a maliciously constructed string could cause deserialization to recurse, leading to stack exhaustion. This issue affected HHVM prior to v4.32.3, between versions 4.33.0 and 4.56.0, 4.57.0, 4.58.0, 4.58.1, 4.59.0, 4.60.0, 4.61.0, 4.62.0.

The fbgames protocol handler registered as part of Facebook Gameroom does not properly quote arguments passed to the executable. That allows a malicious URL to cause code execution. This issue affects versions prior to v1.26.0.

Due to incorrect string size calculations inside the preg_quote function, a large input string passed to the function can trigger an integer overflow leading to a heap overflow. This issue affects HHVM versions prior to 4.56.3, all versions between 4.57.0 and 4.80.1, all versions between 4.81.0 and 4.93.1, and versions 4.94.0, 4.95.0, 4.96.0, 4.97.0, 4.98.0.

In the crypt function, we attempt to null terminate a buffer using the size of the input salt without validating that the offset is within the buffer. This issue affects HHVM versions prior to 4.56.3, all versions between 4.57.0 and 4.80.1, all versions between 4.81.0 and 4.93.1, and versions 4.94.0, 4.95.0, 4.96.0, 4.97.0, 4.98.0.

Incorrect bounds calculations in substr_compare could lead to an out-of-bounds read when the second string argument passed in is longer than the first. This issue affects HHVM versions prior to 4.56.3, all versions between 4.57.0 and 4.80.1, all versions between 4.81.0 and 4.93.1, and versions 4.94.0, 4.95.0, 4.96.0, 4.97.0, 4.98.0.

In-memory file operations (ie: using fopen on a data URI) did not properly restrict negative seeking, allowing for the reading of memory prior to the in-memory buffer. This issue affects HHVM versions prior to 4.56.3, all versions between 4.57.0 and 4.80.1, all versions between 4.81.0 and 4.93.1, and versions 4.94.0, 4.95.0, 4.96.0, 4.97.0, 4.98.0.

xbuf_format_converter, used as part of exif_read_data, was appending a terminating null character to the generated string, but was not using its standard append char function. As a result, if the buffer was full, it would result in an out-of-bounds write. This issue affects HHVM versions prior to 4.56.3, all versions between 4.57.0 and 4.80.1, all versions between 4.81.0 and 4.93.1, and versions 4.94.0, 4.95.0, 4.96.0, 4.97.0, 4.98.0.