msModelSlim Security Statement

System Security Hardening

You are advised to enable the address space layout randomization (ASLR) (level 2) in the system. Run the following command to enable it:

echo 2 > /proc/sys/kernel/randomize_va_space

Running User Recommendations

1. Use the principle of least privilege. For example, prevent other users from writing data by disabling permissions such as 666 and 777.

2. Before installing or using the tools, ensure the user's umask is set to 0027 or a more restrictive value. Otherwise, issues may occur, such as source code compilation failures, configuration file read failures, or excessive permissions for generated directories and files.

3. All tools in this repository are designed to run with minimal permissions. For security reasons, do not use root or other privileged accounts. Always install and execute tools as a regular user.

4. If a tool depends on CANN, use the CANN package installed by the same non-privileged user. After running the source command, do not modify environment variables in set_env.sh.

File Permission Control

1. Ensure that the directory permissions are 750.

2. When providing input files to the tool as command inputs, it is recommended that the file owner matches the process owner of the tool and that file permissions restrict write access for group and others. By default, tool files written to the drive are not writable by others. You can manually control the permissions for the generated files as needed.

3. Proper permission control is essential during installation and use. For details, see the following table.

File Permission Reference

Type	Maximum Linux Permission
Home directory	750 (rwxr-x---)
Program files (including scripts and libraries)	550 (r-xr-x---)
Program file directory	550 (r-xr-x---)
Configuration files	640 (rw-r-----)
Configuration file directory	750 (rwxr-x---)
Log files (recorded or archived)	440 (r--r-----)
Log files (being recorded)	640 (rw-r-----)
Log file directory	750 (rwxr-x---)
Debug files	640 (rw-r-----)
Debug file directory	750 (rwxr-x---)
Temporary file directory	750 (rwxr-x---)
Maintenance and upgrade file directory	770 (rwxrwx---)
Service data files	640 (rw-r-----)
Service data file directory	750 (rwxr-x---)
Key components, private keys, certificates, and ciphertext file directory	700 (rwx------)
Key components, private keys, certificates, and ciphertext files	600 (rw-------)
APIs and scripts for encryption and decryption	500 (r-x------)

Vulnerability Security Statement

The MindStudio community highly values the security of community versions. Vulnerability management specialists are specifically designated to handle vulnerability-related matters. To build a more secure AI full-process toolchain, we look forward to your participation.

Vulnerability Handling Process

For each security vulnerability, the MindStudio community assigns personnel to follow up and handle it. The end-to-end vulnerability handling process is shown in the following figure.

The following sections explain the vulnerability reporting, assessment, and disclosure processes.

Vulnerability Reporting

You can contact the MindStudio community team by submitting an issue. We will arrange for a security vulnerability specialist to contact you promptly. Note that to ensure security, do not include specific information about security privacy in the issue.

Response to Reports

1. The MindStudio community will confirm, analyze, and report security vulnerability issues within three working days, while initiating the security handling process.
2. The MindStudio security team will assign confirmed security vulnerability issues to dedicated personnel and follow up on them.
3. During the process of classifying, confirming, and fixing security vulnerabilities, as well as releasing patches, we will provide timely updates on the report.

Vulnerability assessment

The industry widely uses the CVSS standard to assess vulnerability severity. When using CVSS v3.1 for vulnerability assessment, MindStudio sets specific attack scenarios and performs assessments based on the actual impact within those scenarios. Vulnerability severity assessment involves assessing the difficulty of exploitation as well as the impact on confidentiality, integrity, and availability after exploitation, resulting in a numerical score.

Vulnerability Assessment Metrics

MindStudio assesses vulnerability severity levels by using the following vector metrics:

• Attack vector (AV): indicates the "remoteness" of an attack and how a vulnerability can be exploited.
• Attack complexity (AC): describes the difficulty of executing an attack and the factors required for a successful attack.
• User interaction (UI): determines whether the attack requires user participation.
• Privileges required (PR): records the level of user authentication required for a successful attack.
• Scope (S): determines whether an attack can affect components with different permission levels.
• Confidentiality (C): measures the impact resulting from information disclosure to unauthorized parties.
• Integrity (I): measures the impact resulting from information tampering.
• Availability (A): measures the impact on users' access to data or services when needed.

Assessment Principles

• Assess the severity level of a vulnerability, not the risk.
• The assessment must be based on an attack scenario where a successful attack can compromise the confidentiality, integrity, and availability of the system.
• When a security vulnerability has multiple attack scenarios, the attack scenario with the highest CVSS score (that is, with the greatest impact) shall prevail in the assessment.
• If a vulnerability exists in an embedded or invoked library, perform the assessment after determining the attack scenario based on how the library is used in the product.
• If a security defect cannot be triggered or does not affect confidentiality, integrity, or availability (CIA), the CVSS score is 0.

Assessment Procedure

To assess the severity level of a vulnerability, perform the following steps:

1. Set a possible attack scenario and score based on this attack scenario.

2. Identify the vulnerable component and affected components.

3. Select values for base metrics.

   • Select values for the exploitability metrics (attack vector, attack complexity, privileges required, user interaction, and scope) based on the vulnerable component.

   • Ensure impact metrics (confidentiality, integrity, and availability) reflect the impact on either the vulnerable component or the affected components, whichever is more severe.

Severity Rating

Severity Rating	CVSS Score	Vulnerability Fix Time
Critical	9.0~10.0	7 days
High	7.0~8.9	14 days
Medium	4.0~6.9	30 days
Low	0.1~3.9	30 days

Vulnerability Disclosure

After a security vulnerability is fixed, the MindStudio community will release a security advisory (SA) and security notice (SN). The SA includes technical details of the vulnerability, type, reporter, CVE ID, affected versions, and fixed versions. To ensure security for MindStudio users, the MindStudio community will not publicly disclose, discuss, or confirm security issues until after investigation and fixing are complete and an SA has been released.

Appendix

MindStudio SA

Currently maintained versions have no security vulnerabilities.

MindStudio SN

No vulnerability notices are currently available for third-party open-source components.

Data Security Statement

Loading and saving data during the use of the tools may involve data risks.

Build Security Statement

To reduce security risks, you can perform permission control on files within the source code directory. During the build process, you can modify build scripts as needed to avoid security risks and ensure the security of the build results.

Runtime Security Statement

If an exception occurs during operation, the tool will exit the process and print error messages. This is expected behavior. You are advised to locate the specific cause of the error based on the error prompts, such as by viewing log files or result files generated during the collection and parsing process.

Public Network Address Statement

For details, see Public URLs.

Public API Statement

msModelSlim is developed in C++ and Python. All public APIs are documented.

For scripting languages such as Python where source code is released, use the public APIs specified in the documentation. Do not directly call source code APIs that are not explicitly disclosed.

Usage of Safe Functions

While unsafe functions are not forcibly disabled, you are advised to use their safe variants that explicitly take a buffer length parameter, for example, memset_s and memcpy_s.

Communication Matrix

No.	Code Repository	Function	Source Device	Source IP Address	Source Port	Destination Device	Destination IP Address	Destination Port (Listening)	Protocol	Port Description	Port Configuration	Listening Port Configurable (Yes/No)	Authentication Mode	Encryption Mode	Plane	Version	Special Scenarios	Remarks
1	msModelSlim	Inter-card communication for quantization	Single-node multi-device	127.0.0.1		Single-node multi-device	127.0.0.1	Dynamic search. The default value is 29500-29599. The value ranges from 1024 to 65535.	TCP	Dynamically searches for an available port for inter-card communication. Search starts at 29500 by default, and the tool tries up to 100 ports. You can specify the port through the MASTER_PORT environment variable. The port must be in the range from 1024 to 65535.	N/A	Yes (using the MASTER_PORT environment variable)			Service plane	master	None
2	msModelSlim	Multi-process context sharing	Local on a single node	N/A (Unix domain socket)		Local on a single node	N/A (Unix domain socket)	N/A (using a file system socket path)	Unix domain socket	Shares context data between processes through the unix domain socket. The socket file is created by default in the system temporary directory (/tmp/peercred_manager_<pid>.sock), and the file permissions are 0600. It uses SO_PEERCRED for kernel-level UID authentication and allows connections only from processes with the same UID.	N/A	The socket file path can be specified in code.	Kernel-level UID authentication with SO_PEERCRED	N/A (local communication)	Service plane	master	Linux only