CVE-2020-15202: critical vulnerability in tensorflow
Integer truncation in Shard API usage
Published · Updated
E-mail alert when it lands in KEV, gets a public PoC or its score goes up.
No sign of exploitation. No public exploitation artifact known so far.
TensorFlow's Shard API has a bug where functions expecting 64-bit integers receive 32-bit integers instead, causing data to be truncated when processing large amounts of work. This can lead to crashes, memory corruption, or data being read/written to the wrong locations.
Integer truncation vulnerability in TensorFlow's Shard API occurs when lambdas with int32 parameters are passed instead of required int64 parameters; when parallelized workload exceeds 32-bit range, truncation causes out-of-bounds heap access, stack overflow, or memory corruption. Requires TensorFlow versions before 1.15.4, 2.0.3, 2.1.2, 2.2.1, or 2.3.1 with affected Shard API calls.
In the same product, most dangerous first.