In Apache Hadoop 3.1.0 to 3.1.1, 3.0.0-alpha1 to 3.0.3, 2.9.0 to 2.9.1, and 2.0.0-alpha to 2.8.4, the user/group information can be corrupted across storing in fsimage and reading back from fsimage.
A vulnerability has been found in the way developers have implemented the archive extraction of files. An arbitrary file write vulnerability, that can be achieved using a specially crafted zip archive (affects other archives as well, bzip2, tar,xz, war, cpio, 7z), that holds path traversal filenames. So when the filename gets concatenated to the target extraction directory, the final path ends up outside of the target folder. Of course if an executable or a configuration file is overwritten with a file containing malicious code, the problem can turn into an arbitrary code execution issue quite easily. This affects multiple libraries that lacks of a high level APIs that provide the archive extraction functionality.
References: https://snyk.io/research/zip-slip-vulnerability
Patch: https://github.com/apache/hadoop/commit/745f203e577bacb35b042206db94615141fa5e6f#diff-19cfa944d1bb9ce1daa46aa4223f3642
Vulnerability in Apache Hadoop 0.23.x, 2.x before 2.7.5, 2.8.x before 2.8.3, and 3.0.0-alpha through 3.0.0-beta1 allows a cluster user to expose private files owned by the user running the MapReduce job history server process. The malicious user can construct a configuration file containing XML directives that reference sensitive files on the MapReduce job history server host.
Apache Hadoop before 0.23.4, 1.x before 1.0.4, and 2.x before 2.0.2 generate token passwords using a 20-bit secret when Kerberos security features are enabled, which makes it easier for context-dependent attackers to crack secret keys via a brute-force attack.
Apache Hadoop 0.23.x before 0.23.11 and 2.x before 2.4.1, as used in Cloudera CDH 5.0.x before 5.0.2, do not check authorization for the (1) refreshNamenodes, (2) deleteBlockPool, and (3) shutdownDatanode HDFS admin commands, which allows remote authenticated users to cause a denial of service (DataNodes shutdown) or perform unnecessary operations by issuing a command.
The YARN NodeManager daemon in Apache Hadoop 0.23.0 through 0.23.11 and 2.x before 2.5.2, when using Kerberos authentication, allows remote cluster users to change the permissions of certain files to world-readable via a symlink attack in a public tar archive, which is not properly handled during localization, related to distributed cache.
It was reported [1] that:
"The Apache Hadoop RPC protocol is intended to provide bidirectional authentication between clients and servers. However, a malicious server or network attacker can unilaterally disable these authentication checks. This allows for potential reduction in the configured quality of protection of the RPC traffic, and privilege escalation if authentication credentials are passed over RPC."
This flaw only affects users who have enabled Hadoop's kerberos security features. This is corrected in upstream versions 0.23.9, 1.2.1, and 2.0.6-alpha.
[1] http://seclists.org/fulldisclosure/2013/Aug/251
DataNodes in Apache Hadoop 2.0.0 alpha does not check the BlockTokens of clients when Kerberos is enabled and the DataNode has checked out the same BlockPool twice from a NodeName, which might allow remote clients to read arbitrary blocks, write to blocks to which they only have read access, and have other unspecified impacts.