Platform/SAST
SAST: data path from input to vulnerability
Static analysis of your code. Shows the full path of user data to the dangerous sink.
uid = request.args["id"]
sql = "SELECT … WHERE id=" + uid
cursor.execute(sql)
SQL injection
01How it works
From code to decision
- 01Codefrom GitLab or as an archive
- 02Parsingall languages are reduced to one code model
- 03Data pathfrom input to the dangerous sink, through functions and files
- 04Findingstep-by-step path, file and line
- 05DecisionAppSec confirms or dismisses with a reason
02In the platform
What a finding is based on
- path proven
- A data path from input to the dangerous sink is found. Every step is visible in the code.
- by code structure
- A dangerous construct is visible without a data path: for example, TLS certificate verification is disabled.
- suspicion
- The analysis did not confirm the path. Such findings are not included in the compliance report.
Vulnerabilities · payments-api group by by finding
SQL injectiontransfer/TransferDao.java:118
needs triage Deserialization of untrusted dataapi/ImportService.java:64
confirmed TLS certificate verification disabledclient/HttpFactory.java:27
needs triage Server-side request forgery (SSRF)preview/LinkLoader.java:52
needs triage Secret in a logauth/LoginFilter.java:88
fixed exampleseverity · basis · status
- The full data path
- Through functions and files, from input to the dangerous sink. You can walk the steps right in the code.
- Place in code
- File and line at every step of the path.
- Fix deadline
- The organization sets days to fix by severity. Overdue items are shown separately.
- Tests and examples
- Findings in tests and examples are hidden by default and do not count as vulnerabilities in the report.
- Decision
- AppSec confirms or dismisses with a reason. "Fixed" is set by a re-scan.
Languages
JavaPL/SQLC#JavaScriptTypeScriptPython1C
04Rules
Sample rules by language
Java
- SQL injection
- JNDI injection
- Deserialization of untrusted data
- Expression injection: SpEL, OGNL, MVEL
- XML external entities (XXE)
- JWT without signature verification
C#
- XSS in a view (Html.Raw)
- Deserialization with a type taken from the request
- Mass assignment of an entity from the request
- API handler without an authorization check
- TLS certificate verification disabled
- Secret or personal data in a log
JavaScript and TypeScript
- OS command injection
- Prototype pollution
- NoSQL injection
- DOM-based cross-site scripting
- Server-side request forgery (SSRF)
- Regular expression with nested quantifiers (ReDoS)
Python
- Code execution from user data
- Insecure deserialization
- Server-side template injection
- Unpacking an archive from the request without a filter
- Debug mode in production settings
- Jinja2 autoescaping disabled
PL/SQL
- SQL injection in dynamic SQL
- Caller's command run with the code owner's privileges
- Constant encryption key in a DBMS_CRYPTO call
- Audit disabled from code
1C
- Execute (Выполнить) and Eval (Вычислить) over user data
- External data processor without safe mode
- Privileged mode without a permission check
- HTTP service without an access check
05Report and perimeter
For the auditor and for your network
What goes into the report
- ARDFM No. 53, cl. 148. Static analysis of source code completed without failures. The basis states the scan number.
- ARDFM No. 53, cl. 149. Languages of the checked version: which are checked, which are not, and what share of code that is.
- Order No. 111/NQ. Code analysis with a software tool and a decision on every vulnerability found.
Inside your perimeter
- No internet. The check runs on your server, with no outbound requests.
- Nothing is executed. No build tools or scripts from the repository are run: files are only read.
- What was not checked. The scan report names languages without rules, unparsed and oversized files.