Vulnerability in Keras Model.load_model Leading to Arbitrary Local File Loading and SSRF

The Keras.Model.load_model method, including when executed with the intended security mitigation safe_mode=True, is vulnerable to arbitrary local file loading and Server-Side Request Forgery (SSRF). This vulnerability stems from the way the StringLookup layer is handled during model loading from a specially crafted .keras archive. The constructor for the StringLookup layer accepts a vocabulary argument that can specify a local file path or a remote file path. * Arbitrary Local File Read: An attacker can create a malicious .keras file that embeds a local path in the StringLookup layer's configuration. When the model is loaded, Keras will attempt to read the content of the specified local file and incorporate it into the model state (e.g., retrievable via get_vocabulary()), allowing an attacker to read arbitrary local files on the hosting system. * Server-Side Request Forgery (SSRF): Keras utilizes tf.io.gfile for file operations. Since tf.io.gfile supports remote filesystem handlers (such as GCS and HDFS) and HTTP/HTTPS protocols, the same mechanism can be leveraged to fetch content from arbitrary network endpoints on the server's behalf, resulting in an SSRF condition. The security issue is that the feature allowing external path loading was not properly restricted by the safe_mode=True flag, which was intended to prevent such unintended data access.

geyang ml-logger Ping server.py log_handler deserialization

A vulnerability was determined in geyang ml-logger up to acf255bade5be6ad88d90735c8367b28cbe3a743. Affected is the function log_handler of the file ml_logger/server.py of the component Ping Handler. This manipulation of the argument data causes deserialization. It is possible to initiate the attack remotely. The exploit has been publicly disclosed and may be utilized. This product is using a rolling release to provide continious delivery. Therefore, no version details for affected nor updated releases are available.

A vulnerability in Crater Invoice allows an unauthenticated attacker with knowledge of the APP_KEY to achieve remote command execution on the server by manipulating the laravel_session cookie, exploiting arbitrary deserialization through the encrypted session data. The exploitation vector of this vulnerability relies on an attacker obtaining Laravel's secret APP_KEY, which would allow them to decrypt and manipulate session cookies (laravel_session) containing serialized data. By altering this data and re-encrypting it with the APP_KEY, the attacker could trigger arbitrary deserialization on the server, potentially leading to remote command execution (RCE). The vulnerability is primarily exploited by accessing an exposed cookie and manipulating it using the secret key to gain malicious access to the server.

Insecure Deserialization in some workflows of the IPS Manager allows unauthenticated remote attackers to perform arbitrary code execution and access to the vulnerable Trellix IPS Manager.

Flipbox Builder <= 1.5 - Authenticated (Contributor+) PHP Object Injection

The Flipbox Builder plugin for WordPress is vulnerable to PHP Object Injection in all versions up to, and including, 1.5 via deserialization of untrusted input in the flipbox_builder_Flipbox_ShortCode function. This makes it possible for authenticated attackers, with Contributor-level access and above, to inject a PHP Object. No known POP chain is present in the vulnerable software. If a POP chain is present via an additional plugin or theme installed on the target system, it could allow the attacker to delete arbitrary files, retrieve sensitive data, or execute code.

ORIPA LoaderXML.java deserialization

A vulnerability was found in ORIPA up to 1.72. It has been declared as critical. Affected by this vulnerability is an unknown functionality of the file src/main/java/oripa/persistence/doc/loader/LoaderXML.java. The manipulation leads to deserialization. The attack can be launched remotely. Upgrading to version 1.80 is able to address this issue. It is recommended to upgrade the affected component. The identifier VDB-270169 was assigned to this vulnerability.

H2O deserializes ML models without filtering, potentially allowing execution of malicious code

The H2O machine learning platform uses "Iced" classes as the primary means of moving Java Objects around the cluster. The Iced format supports inclusion of serialized Java objects. When a model is deserialized, any class is allowed to be deserialized (no class whitelist). An attacker can construct a crafted Iced model that uses Java gadgets and leads to arbitrary code execution when imported to the H2O platform.

The Next <= 1.1.0 - Authenticated (Contributor+) PHP Object Injection

The The Next theme for WordPress is vulnerable to PHP Object Injection in all versions up to, and including, 1.1.0 via deserialization of untrusted input from the wpeden_post_meta post meta value. This makes it possible for authenticated attackers, with Contributor-level access and above, to inject a PHP Object. No known POP chain is present in the vulnerable software. If a POP chain is present via an additional plugin or theme installed on the target system, it could allow the attacker to delete arbitrary files, retrieve sensitive data, or execute code.

Unauthenticated RCE via .NET Deserialization in Topal Finance Software

Deserialization of Untrusted Data vulnerability in Topal Solutions AG Topal Finanzbuchhaltung on Windows allows Remote Code Execution.This issue affects at least Topal Finanzbuchhaltung: 10.1.5.20 and is fixed in version 11.2.12.00

SEAT Queue Ticket Kiosk Java RMI Registry deserialization

A flaw has been found in SEAT Queue Ticket Kiosk up to 20250827. This affects an unknown part of the component Java RMI Registry Handler. This manipulation causes deserialization. The attack can only be done within the local network. The attack is considered to have high complexity. It is indicated that the exploitability is difficult. The vendor was contacted early about this disclosure but did not respond in any way.

Deseriliazation of untrusted data can occur in versions 3.7.0 or newer of Ydata's ydata-profiling open-source library, enabling a maliciously crafted dataset to run arbitrary code on an end user's system when loaded.

A local privilege escalation vulnerability in Sophos Intercept X for Windows with Central Device Encryption 2024.2.0 and older allows writing of arbitrary files.

Deserialization of untrusted data can occur in versions 3.7.0 or newer of Ydata's ydata-profiling open-source library, enabling a malicously crafted report to run arbitrary code on an end user's system when loaded.

Remote Code Execution in kedro-org/kedro

A Remote Code Execution (RCE) vulnerability has been identified in the Kedro ShelveStore class (version 0.19.8). This vulnerability allows an attacker to execute arbitrary Python code via deserialization of malicious payloads, potentially leading to a full system compromise. The ShelveStore class uses Python's shelve module to manage session data, which relies on pickle for serialization. Crafting a malicious payload and storing it in the shelve file can lead to RCE when the payload is deserialized.

Aridius XYZ News loadMore deserialization

A vulnerability has been found in Aridius XYZ up to 20240927 on OpenCart and classified as critical. This vulnerability affects the function loadMore of the component News. The manipulation leads to deserialization. The attack can be initiated remotely. The exploit has been disclosed to the public and may be used. It is recommended to upgrade the affected component.

PixelYourSite 10.1.1.1 - Insecure deserialization

PixelYourSite - Your smart PIXEL (TAG) and API Manager 10.1.1.1 was found to be vulnerable. Unvalidated user input is used directly in an unserialize function in myapp/modules/facebook/facebook-server-a sync-task.php.

VITA-MLLM Freeze-Omni utils.py torch.load deserialization

A vulnerability, which was classified as problematic, has been found in VITA-MLLM Freeze-Omni up to 20250421. This issue affects the function torch.load of the file models/utils.py. The manipulation of the argument path leads to deserialization. It is possible to launch the attack on the local host.

NewPipe has potential security vulnerability when importing settings

NewPipe is an Android app for video streaming written in Java. It supports exporting and importing backups, as a way to let users move their data to a new device effortlessly. However, in versions 0.13.4 through 0.26.1, importing a backup file from an untrusted source could have resulted in Arbitrary Code Execution. This is because backups are serialized/deserialized using Java's Object Serialization Stream Protocol, which can allow constructing any class in the app, unless properly restricted. To exploit this vulnerability, an attacker would need to build a backup file containing the exploit, and then persuade a user into importing it. During the import process, the malicious code would be executed, possibly crashing the app, stealing user data from the NewPipe app, performing nasty actions through Android APIs, and attempting Android JVM/Sandbox escapes through vulnerabilities in the Android OS. The attack can take place only if the user imports a malicious backup file, so an attacker would need to trick a user into importing a backup file from a source they can control. The implementation details of the malicious backup file can be independent of the attacked user or the device they are being run on, and do not require additional privileges. All NewPipe versions from 0.13.4 to 0.26.1 are vulnerable. NewPipe version 0.27.0 fixes the issue by doing the following: Restrict the classes that can be deserialized when calling Java's Object Serialization Stream Protocol, by adding a whitelist with only innocuous data-only classes that can't lead to Arbitrary Code Execution; deprecate backups serialized with Java's Object Serialization Stream Protocol; use JSON serialization for all newly created backups (but still include an alternative file serialized with Java's Object Serialization Stream Protocol in the backup zip for backwards compatibility); show a warning to the user when attempting to import a backup where the only available serialization mode is Java's Object Serialization Stream Protocol (note that in the future this serialization mode will be removed completely).

image-optimizer before 1.7.3 allows PHAR deserialization, e.g., the phar:// protocol in arguments to file_exists().

NLTK through 3.8.1 allows remote code execution if untrusted packages have pickled Python code, and the integrated data package download functionality is used. This affects, for example, averaged_perceptron_tagger and punkt.