#security News & Analysis

DMind Benchmark: Toward a Holistic Assessment of LLM Capabilities across the Web3 Domain

What Was That Again? Certified Robustness for Automatic Speech Recognition

Towards Reliable and Robust LLM Planning: Symbolic Feedback-Driven Iterative Self-Refinement Framework

Seven Security Challenges That Must be Solved in Cross-domain Multi-agent LLM Systems

Agent-Native Immune System: Architecture, Taxonomy, and Engineering

On the Inseparability of Instructions and Data in Shared-Embedding Sequence Models

AI Healthcare Chatbots as Information Infrastructure: A Large-Scale Study of User-Reported Breakdowns

Byzantine-Robust Aggregation for Securing Decentralized Federated Learning

Fortress and Gatekeeper: Theorizing Transitive Trust in Third-Party Cybersecurity Risk Governance

Chai: Agentic Discovery of Cryptographic Misuse Vulnerabilities

ShareLock: A Stealthy Multi-Tool Threshold Poisoning Attack Against MCP

Adaptive Evaluation of Out-of-Band Defenses Against Prompt Injection in LLM Agents

Inherited Circuits, Learned Semantics: How Fine-Tuning Creates Evasion Vulnerabilities Invisible to Standard Evaluation

TEMPO-Diffusion: Temporally Exposed Malicious Poisoning of Diffusion Models

CyberChainBench: Can AI Agents Secure Smart Contracts Against Real-World On-Chain Vulnerabilities?

BitGo cuts 15% of workforce, sharpens focus on security, trading, and AI infrastructure

BitGo announced a 15% workforce reduction while strategically pivoting toward security, trading, and AI infrastructure development. The restructuring positions the company to capitalize on emerging crypto sector trends while streamlining operations, signaling broader industry consolidation around high-value service categories.

Revolut enhances fraud detection with PRAGMA model on Nvidia platform

Revolut has deployed its PRAGMA fraud detection model on Nvidia's platform to enhance financial security and operational efficiency. The development represents a significant advancement in AI-driven banking solutions, potentially establishing new industry standards for fraud prevention across the financial sector.

Attention in Motion: Secure Platooning via Transformer-based Misbehavior Detection

FALCON: Transforming Cyber Threat Intelligence into Deployable IDS Rules with Self-Reflection

RIFT-Bench: Dynamic Red-teaming For Agentic AI Systems

Red-Teaming the Agentic Red-Team

MuTRAP: Multi-trigger Trojans Attacking Robot Task Planning Systems

Tuning without Peeking: Provable Generalization Bounds and Robust LLM Post-Training

CNnotator: LLM-Guided Memory Safety Annotation Synthesis

CNnotator, an LLM-powered tool, automatically generates memory safety annotations for legacy C code by synthesizing specifications that help identify security vulnerabilities. OpenAI's o3 model achieved 90% first-attempt success rates, suggesting AI-assisted code annotation is becoming practical for real-world systems migration and security analysis.

SCRUB-FL: Sanitizing and Cleansing Representations via Unlearning of Backdoors

Researchers introduce SCRUB-FL, a post-training defense mechanism against backdoor attacks in federated learning systems that reduces attack success rates to 3.88% while preserving model accuracy. The method uses spectral analysis and machine unlearning to remove trigger-target associations without requiring prior knowledge of attack patterns or clean datasets.