Spectra Assure
Community
warningRisk: Secrets
Scanned: 20 days ago

jax-cuda12-pjrt

Artifact:
latest
Top 10k
JAX XLA PJRT Plugin for NVIDIA GPUs
License: Permissive (Apache-2.0)
Published: 20 days ago




SAFE Assessment

Compliance

Licenses
No license compliance issues
Secrets
1 debugging symbols found

Security

Vulnerabilities
No known vulnerabilities detected
Hardening
1 execution hijacking concerns

Threats

Tampering
No evidence of software tampering
Malware
No evidence of malware inclusion

INCIDENTS FOR THIS VERSION:

Popularity

6.61M
Total Downloads
Contributor
Declared Dependencies
3
Dependents

Top issues

Problem

On Linux, external symbols are resolved via the procedure linkage table (PLT) and the global offset table (GOT). Without any protection, both are writable at runtime and thus leave the executable vulnerable to pointer hijacking - an attack where the function address is overwritten with an address of a malicious function. Pointer hijacking can be mitigated by using full read-only relocations, which instruct the compiler to unify global offset tables into a single read-only table. This requires that all external function symbols are resolved at load-time instead of during execution, and may increase loading time for large programs.

Prevalence in PyPI community

26 packages
found in
Top 100
135 packages
found in
Top 1k
847 packages
found in
Top 10k
17.09k packages
in community

Next steps

In most cases, it's recommended to use full read-only relocations (in GCC: -Wl,-z,relro,-z,now).
If the executable load-time is an issue, you should use partial read-only relocations.

Problem

Common compilers often embed source code information into executables for debugging purposes, usually by mapping symbols to source filenames or paths. While this is typically desirable in open-source software and standard tools, that information can be used to determine security weaknesses, code repository layout, trade secrets and similar sensitive information. Such symbols make it easier to reverse-engineer a closed source application.

Prevalence in PyPI community

27 packages
found in
Top 100
138 packages
found in
Top 1k
922 packages
found in
Top 10k
19.5k packages
in community

Next steps

Strip out such information in the linking phase by using compiler options like the -s flag in GCC, or in the post-build phase by using the strip tool.

Problem

Operating systems allow multiple user accounts to coexist on a single computer system. Each registered user has identity information associated with their account. At the very least, user accounts consist of a user name and an optional password. In some cases, user account data may also include personally identifiable information. Extended personal information may include user's given and last name, their email and mailing address, personal photo and their telephone number. Financially motivated attackers may seek to collect personal information for purposes of selling the private data to a third-party. Malicious code that typically exhibits these behavior traits is commonly referred to as an information stealer. While the presence of code that accesses identity information does not necessarily imply malicious intent, all of its uses in a software package should be documented and approved. Accessing identity information is a very common behavior for software packages. One example of acceptable use for such functions is verifying that the active user has purchased a software license that allows them to run the application.

Prevalence in PyPI community

16 packages
found in
Top 100
113 packages
found in
Top 1k
669 packages
found in
Top 10k
19.67k packages
in community

Next steps

Investigate reported detections as indicators of software tampering.
Consult Mitre ATT&CK documentation: T1033 - System Owner/User Discovery.

Top behaviors

Prevalence in PyPI community

Behavior often found in this community (Common)
10 packages
found in
Top 100
49 packages
found in
Top 1k
275 packages
found in
Top 10k
5.88k packages
in community

Prevalence in PyPI community

Behavior often found in this community (Common)
12 packages
found in
Top 100
75 packages
found in
Top 1k
385 packages
found in
Top 10k
8.9k packages
in community

Prevalence in PyPI community

Behavior often found in this community (Common)
34 packages
found in
Top 100
181 packages
found in
Top 1k
1324 packages
found in
Top 10k
54.83k packages
in community

Prevalence in PyPI community

Behavior often found in this community (Common)
69 packages
found in
Top 100
506 packages
found in
Top 1k
3612 packages
found in
Top 10k
163.85k packages
in community

Prevalence in PyPI community

Behavior often found in this community (Common)
37 packages
found in
Top 100
167 packages
found in
Top 1k
1008 packages
found in
Top 10k
21.72k packages
in community

Top vulnerabilities

No vulnerabilities found.