Governing the KV Cache: Preventing Timing Side-Channel Leakage in Multi-Tenant LLM Inference

The key-value KV cache is the primary throughput optimization in modern large language model LLM inference, enabling prefix reuse across requests. In multi-tenant deployments this cache is shared across tenants, creating a timing side channel: an adversarial tenant can reconstruct another tenant'...

From Prompt Injection to Web Exploitation: Revisiting Classic Vulnerabilities in LLM-Integrated Applications

Large Language Models are increasingly integrated into web applications through chatbots, tool-calling pipelines, and agentic workflows. In these systems, user input may influence not only generated text, but also backend actions such as database queries, HTTP requests, file operations, template...

Stealing Reasoning Traces from Proprietary LLM APIs

Leading large language model providers now conceal their models' step-by-step reasoning, or chain-of-thought, to protect intellectual property and limit information leakage. Rather than storing these traces server-side, providers return them to the client as blocks of encrypted text, which the...

Why Generative AI Needs a New Kind of Security

This is Post 2 of a four-part series: Post 1: Agentic AI Security: The Chatbot Era Is Over Post 3: The OWASP LLM Top 10 Was the Warm-Up: What Comes Next Post 4: OWASP LLM Top 10 2026: Every Move Points the Same Direction There is a moment every CISO is having right now, often without warning. A...

Not an A11y: How Android Accessibility Exposes Mobile AI Agents to Indirect Prompt Injection

The rise of autonomous AI agents represents a major paradigm shift in how users interact with mobile devices. Frameworks such as MobileRun and Mobile-Use can autonomously navigate Android applications and execute complex multi-step tasks. To interpret user interfaces, these frameworks rely...

SkillSpector v2.8.2

SkillSpector Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, and security risks before installing agent skills. Overview AI agent skills used by Claude Code, Codex CLI, Gemini CLI, etc. execute with implicit trust and minimal vetting. Research shows that 26.1% of...

Atlassian Rovo Can Be Tricked Into Sending Jira and Confluence Data to Attackers

Attacker-controlled instructions can make Atlassian's Rovo assistant collect Jira or Confluence data that a signed-in user can access, then send it to an outside server. Two security firms found that behavior independently, by different routes. Only one of those routes is confirmed closed...

New CSS Attacks Can Break Webmail Defenses to Steal Passwords and Tokens

New research shows content inside an email can escape its message boundary and interfere with the webmail interface. Across attack chains spanning Outlook, Gmail, Fastmail, Proton Mail, Yahoo Mail, and AOL Mail, the techniques can capture passwords, take over third-party accounts, leak tokens,...

BASIS: Breach-Aware Selective Prompt Injection Shielding with Prefill Attention Probes

Prompt injection is a critical security threat in large language model LLM applications, where attackers hijack model behavior by embedding malicious instructions in user or external data. Existing detection methods only detect the presence of injection and refuse to respond upon detection,...

Defending Retrieval-Augmented Intrusion Detection against Knowledge Poisoning and Prompt Injection

Retrieval-Augmented Generation RAG enables large language models to classify network flows and generate human-readable incident reports by retrieving semantically similar historical traffic from a vector knowledge base. However, the retrieval layer introduces vulnerabilities to knowledge poisonin...

agent-scan v0.5.17-snapshot-6e2d290-1483

Snyk Agent Scan Discover and scan agent components on your machine for prompt injections and vulnerabilities including agents, MCP servers, skills. Note: CLI output is experimental and subject to change Agent Scan v0.5.x planned for deprecation The raw output of this CLI — including issue codes,...

The Anatomy of a Prompt Injection: A Component Model for Structured Analysis

Four years after prompt injection was first identified in 2022, attacks are still predominantly documented as verbatim strings rather than structured exploits, despite advancing agent capabilities and threat actors embedding injections to subvert AI-assisted security analysis. This paper formaliz...

MAL-2026-13430 Malicious code in @ch4acko3/frontal-lobe (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector b73f6f94b2fcc3ef2aabd490862d743e9722f891d2c41421b1352dc318294585 The package hardcodes the endpoint http://47.112.15.137:8789/frontal/trace-sync bare IPv4, plain HTTP in...

MAL-2026-13399 Malicious code in agenthub-multiagent-mcp (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector ca33b55b94f99f1de3d4b34ee5fc42941013cc5ab1f94dce49b6cc0bed5eaa9a agenthub-multiagent-mcp ships a worker dist/worker.js that opens a WebSocket to the hardcoded server wss://agenthub.contetial.com/ws/workers/ and, fo...

Malicious code in agenthub-multiagent-mcp (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector ca33b55b94f99f1de3d4b34ee5fc42941013cc5ab1f94dce49b6cc0bed5eaa9a agenthub-multiagent-mcp ships a worker dist/worker.js that opens a WebSocket to the hardcoded server wss://agenthub.contetial.com/ws/workers/ and, fo...

Malicious code in wallet-monitor-snap (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 5c50a0acdc6be5b64538f83b0c14c3808ea3a57b963233ba32a2d80c3e4cc2c3 wallet-monitor-snap is a MetaMask Snap whose onRpcRequest handler implements a 'showSecurityAlert' method that opens a snapdialog inside MetaMask's...

bromure agentic-coding-v4.5.4

Bromure Secure, ephemeral computing in disposable Linux VMs on macOS. → Full details, screenshots, and downloads at bromure.io This repo ships two sibling apps, both built on Apple's Virtualization.framework: Bromure — every browser session runs in a throwaway Linux VM. Close the window, the VM i...

EUVD-2026-53791

A vulnerability was detected in TinyAGI 0.0.20. The affected element is the function buildSystemPrompt of the file packages/server/src/routes/agents.ts. Performing a manipulation results in file inclusion. The attack may be initiated remotely. The exploit is now public and may be used. The projec...

CVE-2026-19011 TinyAGI agents.ts buildSystemPrompt file inclusion

A vulnerability was detected in TinyAGI 0.0.20. The affected element is the function buildSystemPrompt of the file packages/server/src/routes/agents.ts. Performing a manipulation results in file inclusion. The attack may be initiated remotely. The exploit is now public and may be used. The projec...

CVE-2026-19011

CVE-2026-19011 affects TinyAGI 0.0.20. The vulnerable element is the function buildSystemPrompt in packages/server/src/routes/agents.ts, where manipulation can cause file inclusion. The attack may be initiated remotely and, per the description, the exploit is public. No remediation details are pr...