Comments URL: https://news.ycombinator.com/item?id=47664112

Points: 4

# Comments: 0

We’ve been working on security for multi-turn agent loops and noticed most detection approaches operate on isolated prompts. This paper introduces a framework for modeling intent trajectories across sequences in real time (<20ms), enabling enforcement before harmful actions occur.

Happy to answer technical questions or discuss assumptions in the paper.

Comments URL: https://news.ycombinator.com/item?id=47094095

Points: 1

# Comments: 1

Highflame’s new tool Palisade brings a zero-trust approach to the AI model supply chain. It validates format and structural integrity, detects hidden malicious patterns, verifies provenance via Sigstore/SLSA, and can even trigger behavioral checks to surface backdoors that only activate under certain inputs. Built in Rust for speed and scalability, Palisade makes it feasible to gate models before they hit inference servers or CI/CD pipelines, turning “download and hope” into a verifiable trust boundary.

Author here — happy to answer questions about threat models, performance tradeoffs, or how this fits into CI/CD.

Comments URL: https://news.ycombinator.com/item?id=46371718

Points: 2

# Comments: 1

Its free, extensible with yara support and can be adapted to the needs of an enterprise.

uv run palisade --verbose scan-dir "models/bad_qwen3_sft_playwright_gguf_v2/" --format json Scanning directory: models/bad_qwen3_sft_playwright_gguf_v2 Recursive: False Policy: Default security policy

  Running ToolCallSecurityValidator (3.8s) - 1 critical warning found
  Detection Details:
  - Risk Score: 1.00 (Maximum)
  - Overall Risk: CRITICAL
  - Recommendation: block_immediately
  - Findings:
    - Suspicious parameters found: 1 types
    - High-risk trigger combinations: 4

   Detected Model behavioral backdoor (ToolCallSecurityValidator)
   Identified format string vulnerabilities (BufferOverflowValidator)
   Found injection indicators (ModelIntegrityValidator)
   Discovered tampering evidence (ModelIntegrityValidator)
   Located data exfiltration patterns(SupplyChainValidator)

Comments URL: https://news.ycombinator.com/item?id=44661116

Points: 4

# Comments: 1

Recent advances in transformer architectures, including compact BERT(Devlin et al. 2019) variants (e.g., ModernBERT (Warner et al. 2024)), allow us to build highly accurate classifiers with as few as approximately 400M parameters that achieve rapid inference speeds even on standard CPU hardware. We systematically explore five progressively sophisticated transformer-based architectures: Sharanga (baseline transformer classifier), Mahendra (enhanced attention-weighted pooling with deeper heads), Vaishnava and Ashwina (hybrid neural ensemble architectures), and Raudra (an advanced multi-task framework with specialized loss functions).

Our models are rigorously benchmarked across nine diverse adversarial datasets, including popular sets like the NotInject series, BIPIA, Garak, ImprovedLLM, ToxicChat, WildGuard, and our newly introduced JavelinBench, specifically crafted to test generalization on challenging borderline and hard-negative cases.

Additionally, we compare our architectures against leading open-source guardrail models as well as large decoder-only LLMs such as gpt-4o, demonstrating superior cost-performance trade-offs in terms of accuracy, and latency. Our findings reveal that while Raudra's multi-task design offers the most robust performance overall, each architecture presents unique trade-offs in speed, interpretability, and resource requirements, guiding practitioners in selecting the optimal balance of complexity and efficiency for real-world LLM security applications.

Comments URL: https://news.ycombinator.com/item?id=44238404

Points: 29

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=43119944

Points: 1

# Comments: 0

Full disclosure, I am one of the co-founders.