Resolve Vulnerabilities in Runtime Image #1491

codyharris-h2o-ai · 2024-03-20T18:11:27Z

Hello!
As part of our ongoing to ensure the security of our products, one or more vulnerabilties requiring redmediation have been identified.

Vulnerability	Severity	Image	Package	Description
CVE-2023-37920	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	certifi:2022.12.7	Certifi is a curated collection of Root Certificates for validating the trustworthiness of SSL certificates while verifying the [...]
CVE-2023-48022	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	ray:2.9.3	Anyscale Ray 2.6.3 and 2.8.0 allows a remote attacker to execute arbitrary code via the job submission API. NOTE: the vendor's p[...]

To resolve this, we recommend the following approach:

Install trivy (https://aquasecurity.github.io/trivy)
Scan the current version of the image using a command like trivy image --scanners vuln --severity CRITICAL,HIGH --timeout 60m [...image address...]
Validate that the CVEs are detected using trivy. The provided scans were taken using a different scanner (ECR), so the first step should be to validate that trivy can see them as well.
Iterate to resolve the vulnerabilities. trivy enables you to scan the image without pushing them, so it should help in finding the resolution
Test and publish the fix version, and let us know where we can find the fixed image(s) so we can validate the fixes on our side as well.

Note that we disregard the severity levels assigned by various tools and operate soley on CVSS in line with NIST guidelines. Also note that this scan was performed by ECR, so the results will likely be different. It is in our experience that Trivy produces more results than ECR or Prisma.

The text was updated successfully, but these errors were encountered:

pseudotensor · 2024-03-20T19:20:01Z

The certifi package is 2024.2.2 in image 0.2.0 408. The older vulnerable version being detected is in a "pkgs" folder that is unused and just part of conda base installation before installing other packages. So the notice is a false positive on the wrong version.
There's no resolution for the ray package, no new version is specified, no action can be taken as it's required part of vLLM. Ray is not exposed directly, only the vLLM port that is not ray directly, so there's no real issue.

codyharris-h2o-ai · 2024-03-20T21:56:17Z

@pseudotensor Thanks! For certifi, then can we remove it from the filesystem during the build process?

codyharris-h2o-ai · 2024-03-20T22:00:34Z

There are also a handful of HIGH severities, some of these may or may not be real

Vulnerability	Severity	Image	Package	Description
CVE-2022-3996	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	If an X.509 certificate contains a malformed policy constraint and policy processing is enabled, then a write lock will be taken[...]
CVE-2022-40898	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	wheel:0.37.1	An issue discovered in Python Packaging Authority (PyPA) Wheel 0.37.1 and earlier allows remote attackers to cause a denial of s[...]
CVE-2022-4450	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	The function PEM_read_bio_ex() reads a PEM file from a BIO and parses and decodes the "name" (e.g. "CERTIFICATE"), any header da[...]
CVE-2023-0215	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	The public API function BIO_new_NDEF is a helper function used for streaming ASN.1 data via a BIO. It is primarily used internal[...]
CVE-2023-0216	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to load malformed PKCS7 data with the d2i_PKCS[...]
CVE-2023-0217	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to check a malformed DSA public key by the EVP[...]
CVE-2023-0286	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	There is a type confusion vulnerability relating to X.400 address processing inside an X.509 GeneralName. X.400 addresses were p[...]
CVE-2023-0401	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	A NULL pointer can be dereferenced when signatures are being verified on PKCS7 signed or signedAndEnveloped data. In case the ha[...]
CVE-2023-38325	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	The cryptography package before 41.0.2 for Python mishandles SSH certificates that have critical options.
CVE-2023-43804	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	urllib3:1.26.14	urllib3 is a user-friendly HTTP client library for Python. urllib3 doesn't treat the `Cookie` HTTP header special or provide any[...]
CVE-2023-4807	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	Issue summary: The POLY1305 MAC (message authentication code) implementation contains a bug that might corrupt the internal stat[...]
CVE-2023-49083	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	cryptography is a package designed to expose cryptographic primitives and recipes to Python developers. Calling `load_pem_pkcs7_[...]
CVE-2023-50782	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	A flaw was found in the python-cryptography package. This issue may allow a remote attacker to decrypt captured messages in TLS [...]
CVE-2023-5363	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	cryptography:38.0.4	Issue summary: A bug has been identified in the processing of key and initialisation vector (IV) lengths. This can lead to pote[...]
CVE-2023-6730	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	transformers:4.28.1	Deserialization of Untrusted Data in GitHub repository huggingface/transformers prior to 4.36.
CVE-2023-7018	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-408	transformers:4.28.1	Deserialization of Untrusted Data in GitHub repository huggingface/transformers prior to 4.36.

pseudotensor · 2024-03-20T22:01:44Z

@achraf-mer Can you add the removal of pkgs folders for the h2ogpt/vllm installs like we have for DAI?

pseudotensor · 2024-03-20T22:05:05Z

Just randomly, @codyharris-h2o-ai For transformers, I only see 4.38.2 in the image, not 4.28.1. I don't know where it is getting the versions.

codyharris-h2o-ai · 2024-03-20T22:05:52Z

It's picking it up from workspace/spaces/demo/requirements.txt

codyharris-h2o-ai · 2024-03-20T22:07:04Z

findings.json
Attaching the raw report from ECR

Search for "filePath" in the JSON

pseudotensor · 2024-03-20T22:40:49Z

Ok, that's old code, could be updated, not part of image really.

pseudotensor · 2024-03-20T23:20:10Z

@codyharris-h2o-ai I pushed those changes to remove those unnecessary files. Try again tomorrow on 0.2.0-410

pseudotensor · 2024-03-21T19:53:36Z

@codyharris-h2o-ai Please check again.

codyharris-h2o-ai · 2024-03-21T20:06:24Z

@pseudotensor thanks,
I scanned 412 with the following results:

Vulnerability	Severity	Image	Package	Description
CVE-2023-48022	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	ray:2.9.3	Anyscale Ray 2.6.3 and 2.8.0 allows a remote attacker to execute arbitrary code via the job submission API. NOTE: the vendor's p[...]
CVE-2024-0964	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	gradio:3.50.2	A local file include could be remotely triggered in Gradio due to a vulnerable user-supplied JSON value in an API request.
SNYK-PYTHON-GRADIO-6263801	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	gradio:3.50.2	## Overview gradio is a Python library for easily interacting with trained machine learning m[...]
CVE-2022-3996	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	If an X.509 certificate contains a malformed policy constraint and policy processing is enabled, then a write lock will be taken[...]
CVE-2022-40898	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	wheel:0.37.1	An issue discovered in Python Packaging Authority (PyPA) Wheel 0.37.1 and earlier allows remote attackers to cause a denial of s[...]
CVE-2022-4450	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	The function PEM_read_bio_ex() reads a PEM file from a BIO and parses and decodes the "name" (e.g. "CERTIFICATE"), any header da[...]
CVE-2023-0215	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	The public API function BIO_new_NDEF is a helper function used for streaming ASN.1 data via a BIO. It is primarily used internal[...]
CVE-2023-0216	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to load malformed PKCS7 data with the d2i_PKCS[...]
CVE-2023-0217	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to check a malformed DSA public key by the EVP[...]
CVE-2023-0286	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	There is a type confusion vulnerability relating to X.400 address processing inside an X.509 GeneralName. X.400 addresses were p[...]
CVE-2023-0401	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	A NULL pointer can be dereferenced when signatures are being verified on PKCS7 signed or signedAndEnveloped data. In case the ha[...]
CVE-2023-38325	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	The cryptography package before 41.0.2 for Python mishandles SSH certificates that have critical options.
CVE-2023-4807	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	Issue summary: The POLY1305 MAC (message authentication code) implementation contains a bug that might corrupt the internal stat[...]
CVE-2023-49083	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	cryptography is a package designed to expose cryptographic primitives and recipes to Python developers. Calling `load_pem_pkcs7_[...]
CVE-2023-50782	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	A flaw was found in the python-cryptography package. This issue may allow a remote attacker to decrypt captured messages in TLS [...]
CVE-2023-51449	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	gradio:3.50.2	Gradio is an open-source Python package that allows you to quickly build a demo or web application for your machine learning mod[...]
CVE-2023-5363	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	cryptography:38.0.4	Issue summary: A bug has been identified in the processing of key and initialisation vector (IV) lengths. This can lead to pote[...]
CVE-2023-6572	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-412	gradio:3.50.2	Command Injection in GitHub repository gradio-app/gradio prior to main.

pseudotensor · 2024-03-21T20:09:13Z

Sorry 512 is gradio 3 for k8 and 513 failed during push due to some network issue. Need to avoid the gradio 3 builds we make for k8 issue.

codyharris-h2o-ai · 2024-03-21T20:16:43Z

Ok will try 410

pseudotensor · 2024-03-21T20:44:39Z

I'm building new one, 414.

achraf-mer · 2024-03-21T21:24:26Z

@achraf-mer Can you add the removal of pkgs folders for the h2ogpt/vllm installs like we have for DAI?

I see done in 98e390b and you are building a new image, so will wait and see how to address new findings, thanks.

pseudotensor · 2024-03-21T21:38:13Z

@achraf-mer I already removed the items, I unassigned you thanks!

codyharris-h2o-ai · 2024-03-22T16:09:34Z

Latest scan of 414:

Vulnerability	Severity	Image	Package	Description
CVE-2023-48022	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	ray:2.10.0	Anyscale Ray 2.6.3 and 2.8.0 allows a remote attacker to execute arbitrary code via the job submission API. NOTE: the vendor's p[...]
CVE-2022-3996	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	If an X.509 certificate contains a malformed policy constraint and policy processing is enabled, then a write lock will be taken[...]
CVE-2022-40898	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	wheel:0.37.1	An issue discovered in Python Packaging Authority (PyPA) Wheel 0.37.1 and earlier allows remote attackers to cause a denial of s[...]
CVE-2022-4450	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	The function PEM_read_bio_ex() reads a PEM file from a BIO and parses and decodes the "name" (e.g. "CERTIFICATE"), any header da[...]
CVE-2023-0215	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	The public API function BIO_new_NDEF is a helper function used for streaming ASN.1 data via a BIO. It is primarily used internal[...]
CVE-2023-0216	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to load malformed PKCS7 data with the d2i_PKCS[...]
CVE-2023-0217	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	An invalid pointer dereference on read can be triggered when an application tries to check a malformed DSA public key by the EVP[...]
CVE-2023-0286	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	There is a type confusion vulnerability relating to X.400 address processing inside an X.509 GeneralName. X.400 addresses were p[...]
CVE-2023-0401	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	A NULL pointer can be dereferenced when signatures are being verified on PKCS7 signed or signedAndEnveloped data. In case the ha[...]
CVE-2023-38325	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	The cryptography package before 41.0.2 for Python mishandles SSH certificates that have critical options.
CVE-2023-4807	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	Issue summary: The POLY1305 MAC (message authentication code) implementation contains a bug that might corrupt the internal stat[...]
CVE-2023-49083	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	cryptography is a package designed to expose cryptographic primitives and recipes to Python developers. Calling `load_pem_pkcs7_[...]
CVE-2023-50782	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	A flaw was found in the python-cryptography package. This issue may allow a remote attacker to decrypt captured messages in TLS [...]
CVE-2023-5363	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-414	cryptography:38.0.4	Issue summary: A bug has been identified in the processing of key and initialisation vector (IV) lengths. This can lead to pote[...]

wrt ray, we must mitigate the functionality by removing the offending source files in the package (such as overwriting or deleting or stubbing out the appropriate functions), or remove ray altogether.

pseudotensor · 2024-03-22T19:29:10Z

Where is cryptography==38.0.04 from? I only see we install any latest version, unconstrained. Should be 42.0.5.

codyharris-h2o-ai · 2024-03-25T13:36:36Z

@pseudotensor, hey it appears to be coming from h2ogpt_conda/lib/python3.10/site-packages/cryptography-38.0.4.dist-info/METADATA

pseudotensor · 2024-03-25T14:40:34Z

I think it's because docker build was using fixed miniconda version, not latest, so should be ok tomorrow.

codyharris-h2o-ai · 2024-04-02T15:08:42Z

Vulnerability	Severity	Image	Package	Description
CVE-2023-48022	critical	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-446	ray:2.10.0	Anyscale Ray 2.6.3 and 2.8.0 allows a remote attacker to execute arbitrary code via the job submission API. NOTE: the vendor's p[...]
SNYK-PYTHON-PILLOW-6514866	high	223008754879.dkr.ecr.us-east-1.amazonaws.com/h2ogpt-runtime:0.2.0-446	pillow:10.2.0	## Overview Affected versions of this package are vulnerable to Buffer Overflow via the `strcpy` function in `_imagingcms.c`, d[...]

achraf-mer · 2024-04-02T16:43:22Z

@codyharris-h2o-ai is the ray:2.10.0 issue a case of a bad report?
according to https://nvd.nist.gov/vuln/detail/CVE-2023-48022 and https://bishopfox.com/blog/ray-versions-2-6-3-2-8-0 the CVE only applies to 2.6.3 and 2.8.0.

codyharris-h2o-ai · 2024-04-03T16:42:06Z

I discussed this with @YogevMaty and it sounds like it is still an issue

YogevMaty · 2024-04-03T17:06:41Z

Apparently this CVE is very similar to the one we had in h2o3 .
The default installation does not require authentication and is listening on 0.0.0.0
The company behind Ray is saying it is not a CVE it's by design this is the reason it is not visible in some scanners.
Currently they are not planing of fixing this issue.

what to do
Security and isolation must be enforced outside of the Ray Cluster. Ray expects to run in a safe network environment and to act upon trusted code. Developers and platform providers must maintain the following invariants to ensure the safe operation of Ray Clusters.

https://docs.ray.io/en/latest/ray-security/index.html#best-practices

more info in -https://www.oligo.security/blog/shadowray-attack-ai-workloads-actively-exploited-in-the-wild

codyharris-h2o-ai added the area/security Security issues label Mar 20, 2024

pseudotensor assigned achraf-mer Mar 20, 2024

pseudotensor added a commit that referenced this issue Mar 20, 2024

Remove pkgs for Issue #1491

98e390b

pseudotensor added a commit that referenced this issue Mar 20, 2024

More cleanup for docker image for Issue #1491

fbde9e1

pseudotensor unassigned achraf-mer Mar 21, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Resolve Vulnerabilities in Runtime Image #1491

Resolve Vulnerabilities in Runtime Image #1491

codyharris-h2o-ai commented Mar 20, 2024

pseudotensor commented Mar 20, 2024 •

edited

Loading

codyharris-h2o-ai commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024 •

edited

Loading

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 21, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 21, 2024

pseudotensor commented Mar 21, 2024 •

edited

Loading

achraf-mer commented Mar 21, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 22, 2024 •

edited

Loading

pseudotensor commented Mar 22, 2024

codyharris-h2o-ai commented Mar 25, 2024

pseudotensor commented Mar 25, 2024

codyharris-h2o-ai commented Apr 2, 2024

achraf-mer commented Apr 2, 2024 •

edited

Loading

codyharris-h2o-ai commented Apr 3, 2024

YogevMaty commented Apr 3, 2024

Resolve Vulnerabilities in Runtime Image #1491

Resolve Vulnerabilities in Runtime Image #1491

Comments

codyharris-h2o-ai commented Mar 20, 2024

pseudotensor commented Mar 20, 2024 • edited Loading

codyharris-h2o-ai commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024

codyharris-h2o-ai commented Mar 20, 2024 • edited Loading

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 20, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 21, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 21, 2024

pseudotensor commented Mar 21, 2024 • edited Loading

achraf-mer commented Mar 21, 2024

pseudotensor commented Mar 21, 2024

codyharris-h2o-ai commented Mar 22, 2024 • edited Loading

pseudotensor commented Mar 22, 2024

codyharris-h2o-ai commented Mar 25, 2024

pseudotensor commented Mar 25, 2024

codyharris-h2o-ai commented Apr 2, 2024

achraf-mer commented Apr 2, 2024 • edited Loading

codyharris-h2o-ai commented Apr 3, 2024

YogevMaty commented Apr 3, 2024

pseudotensor commented Mar 20, 2024 •

edited

Loading

codyharris-h2o-ai commented Mar 20, 2024 •

edited

Loading

pseudotensor commented Mar 21, 2024 •

edited

Loading

codyharris-h2o-ai commented Mar 22, 2024 •

edited

Loading

achraf-mer commented Apr 2, 2024 •

edited

Loading