CVE-2022-36022

Deeplearning4J is a suite of tools for deploying and training deep learning models using the JVM. Packages org.deeplearning4j:dl4j-examples and org.deeplearning4j:platform-tests through version 1.0.0-M2.1 may use some unclaimed S3 buckets in tests in examples. This is likely affect people who use some older NLP examples that reference an old S3 bucket. The problem has been patched. Users should upgrade to snapshots as Deeplearning4J plan to publish a release with the fix at a later date. As a workaround, download a word2vec google news vector from a new source using git lfs from here.

Weakness

The product uses insufficiently random numbers or values in a security context that depends on unpredictable numbers.

Affected Software

Potential Mitigations

• Use a well-vetted algorithm that is currently considered to be strong by experts in the field, and select well-tested implementations with adequate length seeds.
• In general, if a pseudo-random number generator is not advertised as being cryptographically secure, then it is probably a statistical PRNG and should not be used in security-sensitive contexts.
• Pseudo-random number generators can produce predictable numbers if the generator is known and the seed can be guessed. A 256-bit seed is a good starting point for producing a “random enough” number.

Related Attack Patterns

• https://cwe.mitre.org/data/definitions/112.html
• https://cwe.mitre.org/data/definitions/485.html
• https://cwe.mitre.org/data/definitions/59.html

References

• https://github.com/eclipse/deeplearning4j/security/advisories/GHSA-rc39-g977-687w
• https://github.com/mmihaltz/word2vec-GoogleNews-vectors