SShortSingh.
Back to feed

AI-Generated Kubernetes Manifest Passed Schema Check but Had 21 Security Flaws

0
·2 views

A GitHub Copilot-generated Kubernetes Deployment manifest cleared strict schema validation via kubeconform without any errors, yet a subsequent policy scanner revealed 21 security findings. The core issue is that schema validators only confirm structural correctness — they do not evaluate whether a configuration is safe or operationally sound. Key gaps in the generated manifest included missing resource limits, absent security contexts such as runAsNonRoot and readOnlyRootFilesystem, and no liveness or readiness probes. The article also warns against conflating liveness and readiness probes, since routing them to the same endpoint can trigger full deployment restarts during incidents rather than simply removing unhealthy pods from traffic rotation. The author recommends running schema checks as pre-commit hooks and policy scans in CI, treating both as complementary rather than interchangeable steps.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Attestly Automates EU AI Act Annex IV Documentation from Agent Traces

A new tool called Attestly aims to simplify compliance with the EU AI Act by generating Annex IV technical documentation directly from AI agents' operational traces. The platform ingests data from sources such as OpenTelemetry, LangSmith, AgentOps, and MCP logs, mapping them into structured evidence trails. Each section of the generated documentation is labeled as AI-generated, user-provided, or missing, and must be reviewed and approved by a human before export. Attestly is designed primarily for startups and teams deploying AI agents within the European Union. A free tier is available for users to get started.

0
ProgrammingDEV Community ·

Developer Builds Live Playground to Compare Four Authorization Engines Side by Side

A developer created an interactive browser-based tool to test four authorization engines — Cedar, Casbin, OpenFGA (ReBAC), and OPA/Rego — against identical access-control rules and every possible request combination. The experiment revealed that ReBAC struggled with time-based conditions, since its core check function has no concept of time, causing it to allow edits at 3am while the other three engines correctly denied them. This limitation is acknowledged by OpenFGA itself, which later introduced Conditions specifically to address time-sensitive scenarios. The project also exposed a subtle testing blind spot: the test suite sampled only four times of day, missing business-hours boundary cases entirely, meaning a misconfigured time window went undetected despite all tests passing. The tool runs fully in the browser using WebAssembly bindings, with no backend, and is publicly available as an open-source repository.

0
ProgrammingDEV Community ·

Developer Builds Low-Cost Epistemic Gate to Counter LLM Data Poisoning in Fine-Tuning

An independent developer has created an 'epistemic gate' system designed to prevent data poisoning attacks during large language model fine-tuning. The project was tested across five different model architectures and orchestrated entirely on a 2006 Toshiba laptop at zero cost. The work spans 16 documented experiments, with fully reproducible code and notebooks published on Kaggle. Supporting materials include a technical whitepaper and a corrective manual with formal mathematical loss specifications. The developer is inviting indie researchers, developers, and startups interested in safe local fine-tuning to review and run the notebooks.

0
ProgrammingDEV Community ·

Three Thai Open-Weight LLMs Compared: Typhoon, OpenThai, and Pathumma

Three locally developed Thai large language models — Typhoon, OpenThai, and Pathumma — offer free, downloadable alternatives to foreign AI subscriptions for Thai-language tasks. Typhoon, built by SCB 10X, provides the broadest portfolio, covering speech recognition, document reading, translation, and even Isan dialect support. OpenThai, maintained by the Thai AI Entrepreneur Association and iApp Technology, is licensed under Apache 2.0 and includes a specialized 30-billion-parameter legal variant capable of citing Thai law by name and section. Pathumma, developed by researchers at VISTEC, focuses on cultural and contextual Thai understanding rather than grammatical accuracy alone. All three models trail frontier foreign models in multi-step reasoning but outperform them on Thai-specific tasks and can be run locally, though hardware requirements vary by model size.