mlsecops-devsecops-awesome

Addresses data poisoning and supply-chain risk in open ML datasets and models by enabling fully attestable ML pipelines. Atlas combines open provenance specifications, trusted hardware, and transparency logs to record artifact authenticity and end-to-end lineage while preserving confidentiality, with a prototype evaluated on two case-study pipelines.

Modern LLM systems assemble third-party weights, adapters, datasets, packages, and images through automated pipelines, exposing them to compromised dependencies, malicious hub artifacts, unsafe deserialization, forged provenance, and backdoored models. Training and release claims are rarely cryptographically bound to the artifacts they describe. This work proposes an attestation-aware promotion gate that verifies claim evidence, enforces safe loading and static scanning, applies secure-by-default deployment constraints, and optionally ingests runtime signals before artifacts enter trusted training or deployment environments.

Deep neural network (DNN) classifiers are potent instruments that can be used in various security-sensitive applications. Still, they are dangerous to certain attacks that impede or distort their learning process. For example, backdoor attacks involve polluting the DNN learning set with a few samples from one or more source classes, which are then labeled as target classes by an attacker. Even if the DNN is trained on clean samples with no backdoors, this attack will still be successful if a backdoor pattern exists in the training data. Backdoor attacks are difficult to spot and can be used to make the DNN behave maliciously, depending on the target selected by the attacker. In this study, we survey the literature and highlight the latest advances in backdoor attack strategies and defense mechanisms. We finalize the discussion on challenges and open issues, as well as future research opportunities.

Multi-agent AppSec scanning with test oracles across a large Rails estate.

Aligning OpenSSF baselines and SLSA with EU CRA compliance.

Agentic vulnerability remediation from Slack finding to verified merge.

GUAC, Scorecard, AllStar, SLSA L3, OpenVEX, and S2C2F in production supply-chain defense.

AI security research, AICM, and agentic guidance.

OpenSSF working material for securing AI/ML systems and pipelines.

SLSA, Sigstore, Scorecard, and related supply-chain security communities.

Global community for GenAI and agentic application security guidance.

Five-phase SDL for agent systems informed by MAESTRO and OWASP Agentic Top 10.

Control set for AI system security and governance.

Companion profile for governing, mapping, measuring, and managing GenAI risk.