WeasyPrint: CSS Injection via Presentational Hints
WeasyPrint helps web developers to create PDF documents. Prior to 69.0, WeasyPrint embeds unescaped HTML presentational-hint attribute values into CSS in weasyprint/css/__init__.py when presentational_hints=True. The background attribute is inserted into a background-image:url() declaration and parsed by tinycss2.parse_blocks_contents(), allowing untrusted HTML to inject additional CSS declarations. Applications that render untrusted HTML with presentational hints enabled can be affected by CSS injection and server-side requests through injected url() values. This issue is fixed in version 69.0.
aiosmtplib: SMTP command injection via CR/LF in sender/recipient address
aiosmtplib is an asynchronous SMTP client for use with asyncio. Prior to 5.1.1, SMTP.mail(), SMTP.rcpt(), SMTP.vrfy(), and SMTP.expn() send caller-supplied addresses without rejecting embedded CR or LF bytes. Data after the line break is framed as additional standalone SMTP command lines, allowing an attacker who influences an envelope sender or recipient to inject commands such as MAIL FROM, RCPT TO, RSET, DATA, or AUTH. SMTP.sendmail() and SMTP.send() without a Message object pass addresses through the affected methods, while SMTP.send_message() is not affected. Successful injection can desynchronize the command-response pipeline, hang the
MobSF: CSRF checks not enforced after Django migration
MobSF is a mobile application security testing tool used. Prior to 4.5.1, mobsf/MobSF/settings.py places django.middleware.csrf.CsrfViewMiddleware only in the deprecated MIDDLEWARE_CLASSES setting and omits it from the active MIDDLEWARE tuple, allowing a remote attacker to make a logged-in victim submit cross-site POST requests to authenticated web endpoints including /delete_scan/, /upload/, /download_scan/, /change_password/, /create_user/, and /delete_user/. This can delete scans, upload or download applications, change passwords, or manage users with the victim account permissions. This issue is fixed in version 4.5.1.
MobSF: SSRF port restriction bypass in assetlinks_check
MobSF is a mobile application security testing tool used. Prior to 4.5.1, get_browsable_activities in mobsf/StaticAnalyzer/views/android/manifest_analysis.py validates only an Android manifest android:host value with valid_host before appending a separately supplied android:port to the URL fetched by _check_url, allowing an authenticated user to upload a crafted APK that makes requests to an attacker-selected nonstandard port at /.well-known/assetlinks.json. With an attacker-controlled hostname and DNS rebinding between validation and the requests.get connection, the request can reach an internal service, although redirects remain disabled an
MobSF: Arbitrary File Read via Path Traversal in ZIP Uploads
MobSF is a mobile application security testing tool used. Prior to 4.5.1, find_icon_path_zip in mobsf/StaticAnalyzer/views/android/icon_analysis.py uses the Android manifest android:icon value to construct paths under the scan resource directory without rejecting traversal or verifying containment, allowing an authenticated user to upload a crafted ZIP or APK that reads a server file with an ALLOWED_EXTENSIONS suffix, copies it to DWD_DIR as the predictable name -icon., and retrieves it through the /download/ endpoint. The same behavior provides a file-existence oracle through the icon_path report field. This issue is fixed in version 4.5.1.
MobSF: Zip Bomb Denial of Service via Per-File Size Limit Bypass in ZIP/APK Extraction
MobSF is a mobile application security testing tool used. Prior to 4.5.1, the unzip function in mobsf/StaticAnalyzer/views/common/shared_func.py logs that an archive member exceeding ZIP_MAX_UNCOMPRESSED_FILE_SIZE is being skipped but does not continue to the next member, so an authenticated user can upload a crafted ZIP or APK whose oversized member is extracted to disk when the aggregate ZIP_MAX_UNCOMPRESSED_TOTAL_SIZE limit has not yet been reached, potentially exhausting disk space and preventing further scans. This issue is fixed in version 4.5.1.
sh _uid does not drop supplementary groups (incomplete privilege drop)
sh provides Python process launching. Prior to 2.2.4, the _uid option in sh.py performs an incomplete privilege drop on Linux and Unix-like systems. When sh runs from an elevated process and launches a command with _uid set to an unprivileged user, the child changes its UID but can retain the parent process's supplementary groups because the privilege-drop sequence does not fully establish the target user's UID, primary GID, and supplementary groups. The child can therefore retain access to files or resources granted to privileged groups such as root, docker, disk, shadow, or sudo, violating the expected _uid privilege boundary. This issue is
`HTTPPasswordMgr` can send saved HTTPS credentials via HTTP because of incorrect scheme matching
The HTTPPasswordMgr class in the urllib.request module, along with its subclasses HTTPPasswordMgrWithDefaultRealm and HTTPPasswordMgrWithPriorAuth, did not take the URL scheme into account when matching stored credentials against a requested URL. Credentials added for an https:// URL were also used for requests to the same host over http://, so an attacker able to redirect or downgrade a client to plain HTTP (for example, via an HTTPS-to-HTTP redirect or an on-path position) could capture credentials in cleartext. Credentials added for http:// URLs could likewise be sent over https://.
Credential matching is now scoped by URL scheme. Credent
Copyparty is a portable file server. Prior to 1.20.17, copyparty volumes with the dk or dks directory-key flag combined with the fk or fka file-key flag can convert a valid file key into a directory key, granting read access to the containing folder. This vulnerability was only reachable if both types of keys (filekeys and dirkeys) were manually enabled in the volume flags simultaneously. This issue is fixed in version 1.20.17.
ONNX: Heap-Buffer-Overflow READ in Gemm Version Converter Adapter via Undersized Input Shape
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.3.0 until 1.22.0, onnx.version_converter.convert_version() can perform an out-of-bounds read in Gemm_7_6::adapt_gemm_7_6() in onnx/version_converter/adapters/gemm_7_6.h when a Gemm node has input tensors with fewer than two dimensions because B_shape[1], A_shape[0], or A_shape[1] is accessed without a rank check, potentially causing a process crash during an opset 7 to 6 downgrade. This issue is fixed in version 1.22.0.
stringprep.map_table_b2() deviates from RFC 3454 Table B.2
The "stringprep" module didn't process characters from RFC 3454 tables
B.2 or B.3 correctly: the latest Unicode codepoint attributes were used
instead of the specified Unicode 3.2.0. This behavior would cause
mismatches when processing domain names using IDNA 2003 (the "idna"
codec) and the in_table_b2() function of the "stringprep" module. This
only affects domain names containing characters that were not previously
registered or had their Unicode attributes such as case-folding
behavior updated since Unicode 3.2.0.
Critical · CVSS 9.3EPSS 10.02%In KEVPublic exploitFix available
MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding)
MLflow is an open source AI engineering platform for agents, large language models, and machine learning models. Starting in 3.3.0 and prior to 3.15.0, the unauthenticated POST /api/2.0/mlflow/webhooks/{id}/test endpoint calls _validate_webhook_url() in mlflow/utils/validation.py only for the original URL while mlflow/webhooks/delivery.py follows redirects and re-resolves the hostname without pinning the validated address, allowing attackers to reach internal or cloud metadata services and receive response_status and response_body. This issue is fixed in version 3.15.0.
vLLM: SSRF + arbitrary local file read in MiMoV2OmniMultiModalProcessor `_fetch_image` and audio loader bypass MediaConnector protections
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the MiMoV2OmniMultiModalProcessor in vllm/transformers_utils/processors/mimo_v2_omni.py passes attacker-controlled image and audio strings through _fetch_image, requests.get, and Image.open instead of MediaConnector, bypassing allowed_media_domains and allowed_local_media_path protections and allowing server-side requests and reads of arbitrary files accessible to the vLLM process. This issue is fixed in version 0.26.0.
vLLM is an inference and serving engine for large language models. Prior to 0.26.0, the /v1/completions/derender and /v1/chat/completions/derender endpoints accept caller-supplied GenerateResponse objects whose generate_responses, choices, token_ids, prompt_logprobs, logprobs.content, top_logprobs, and routed_experts structures are processed by OnlineDerenderer and tokenizer.decode before max_model_len, max_tokens, max_num_seqs, or response-size limits are enforced, allowing an authenticated API client to consume excessive CPU and memory and produce oversized responses. This issue is fixed in version 0.26.0.
Kolibri has Unauthenticated Server-Side Request Forgery (SSRF) in RemoteFacilityUserViewset
Kolibri is an offline-first education platform. Prior to version 0.19.4, several Kolibri API endpoints accept an unvalidated `baseurl` parameter and fetch attacker-controlled URLs from the Kolibri server, reflecting the response body back to the caller. The original report identified two endpoints on the `RemoteFacilityUser*` viewsets; remediation review found two further reflection points on the same pattern. The GET endpoint was unauthenticated. Version 0.19.4 fixes the vulnerability.
compliance-trestle - jinja has an Arbitrary File Write via Path Traversal
compliance-trestle is a tooling platform for managing compliance as code. Prior to versions 3.12.2 and 4.0.3, the `-o/--output` argument in `trestle author jinja` allows writing files outside the intended workspace. The application does not properly validate, `../`, `..\`, or absolute paths. This allows arbitrary file write to attacker-controlled locations. Versions 3.12.3 and 4.0.3 patch the issue.
sqlparse: TokenList.__init__ materializes O(subtree) value per group, causing CPU DoS before depth/token caps trigger
sqlparse is a non-validating SQL parser module for Python. Prior to 0.6.0, TokenList construction and string conversion in sqlparse/sql.py repeatedly flatten nested token subtrees constructed by group_parenthesis and group_case, causing quadratic CPU consumption through sqlparse.parse(), sqlparse.format(), and sqlparse.split() before depth and token limits terminate processing. This issue is fixed in version 0.6.0.
sqlparse: Inefficient Regex Handling of Dollar-Quoted SQL Literals Leads to ReDoS (Denial of Service)
sqlparse is a non-validating SQL parser module for Python. Prior to 0.6.0, SQL_REGEX in sqlparse/keywords.py and the per-position loop in sqlparse/lexer.py repeatedly scan unmatched dollar-quoted literal and multiline-comment delimiters, causing quadratic CPU consumption through sqlparse.parse(), sqlparse.format(), and sqlparse.split(). This issue is fixed in version 0.6.0.
sqlparse is a non-validating SQL parser module for Python. Prior to 0.6.0, group_comments in sqlparse/engine/grouping.py repeatedly rescans comment-only statements before the MAX_GROUPING_TOKENS guard, causing quadratic CPU consumption through sqlparse.parse() and sqlparse.format(sql, strip_comments=True). This issue is fixed in version 0.6.0.
sqlparse: Generated Python and PHP snippets allow SQL string breakout through unescaped backslashes
sqlparse is a non-validating SQL parser module for Python. Prior to 0.6.0, sqlparse/filters/output.py fails to escape existing backslashes before quotes in sqlparse.format output_format='python' and output_format='php' and the corresponding sqlformat -l modes, allowing crafted SQL to terminate the generated string and inject Python or PHP code when a downstream consumer executes or imports the generated source. This issue is fixed in version 0.6.0.