Apache Storm DRPC doAsUser skips impersonation checks (CVE-2026-124496)
An authenticated Storm client can run DRPC functions limited to another user; 3.2.0 fixes storm-server.
Stories per day · 30 days
8 RSS
Relatedapache6information-disclosure6agent-security5apache-answer5access-control4ai-agents4oss-security4apache-airflow3apache-kylin3authentication3openjdk3ubuntu3asset-api2benchmark2bypass2guardrails2java2llm-agents2
81 stories, newest first · one card per story
Stories tagged #authorization
Ubuntu fixes OpenJDK 21 flaws that can expose or alter data
Remote attackers could modify data via Security and JSSE issues; one Compiler flaw needs no authentication.
Airflow async connection test can overwrite later connections
A create-only user can replace a connection created by someone else when commit_on_success runs after the test.
Airflow queued-event delete ignores per-Dag edit checks
Edit rights on one Dag let a user delete queued asset events for every Dag they can read.
Authorization drift occurs in LLM agents during multi-step tool use
Benchmark shows frontier LLMs drift from authorization boundaries in 43% to 61% of tasks, but clean-success paths exist for 86% of cases.
Provenance Identity Continuity proves confused-deputy exclusion in causal security graphs
PIC and COSS generalize security-state continuity over causal execution graphs, proving authority cannot be resurrected across branches.
Relational theory of authority separates proof of possession from proof of relationship
The paper introduces Proof of Relationship alongside PoP, using causal histories to classify executors and prove authorization is occurrence-sensitive.
Apache ActiveMQ LDAPAuthorizationMap allows authorization bypass via DN injection (CVE-2026-108500)
Authenticated attackers can create destinations with crafted DN names containing special characters in versions before 6.3.3 and 5.19.12.
Apache ZooKeeper SASL authorization bypass on client port (CVE-2026-100252)
Affects 3.8.0-3.9.6; client-supplied authzid bypasses SASL checks
Quest Foglight for Databases runScript incorrect authorization RCE
ZDI-26-859 CVSS 8.8 flaw allows remote code execution on authenticated Quest Foglight for Databases
Intent-Bound Authorization Protocol Limits Exposure in Compromised Cryptocurrency Systems
IWAP enforces customer intent and spending reservations to prevent unauthorized withdrawals even when backend signatures are controlled.
Cross-Evidence Learning Detects Broken Object-Level Authorization in Web Applications
The method reconstructs authorization relations from request-response transactions to detect BOLA in black-box systems.
Apache Pinot before 1.6.0 can read unauthorized tables (CVE-2026-73322)
With broker access control on, lookup() can name a dimension table that the FROM-clause check never authorizes.
Kafka lets unauthorized principals read broker files (CVE-2026-112375)
An authenticated principal with no grants can make a broker return a host file or environment variable.
Paper specifies a commit guardrail for agentic entity resolution
The design is conceptual only, with no implementation, empirical results, or proof of semantic correctness.
Claude retrieved paid Georgia vote totals without operator approval
Anthropic says the model requested a dashboard token during an internal statistics project and did not touch individual voter records.
Synthetic social engineering pushes AI agents into unauthorized requests
In 1,260 Qwen3-4B episodes, a prior-agreement appeal raised unauthorized requests from 43.3% to 96.7%.
US banks curb third-party AI agents tied to shared customer logins
Axios says the issue is customers authorizing agents, not a breach of bank systems or stolen credentials.
Many minimal agent tool grants cannot be ranked
For 29.8–33.3% of Microsoft Graph tasks, minimal sufficient grants form an antichain with no smaller choice.
IDENTOS makes PolicyArc generally available for AI agent access
Agents can receive a scoped, time-limited copy of a person's access that ends when the task or window ends.
Per-action checks miss relational policy violations in LLM agents
A SAT/SMT synthesizer produces witnesses that per-action monitors approve but paired-execution review flags.
AuthorizationBench: LLM agents exceed authority in 19.2% of runs
Across 6,864 trajectories, violation rates ranged from 10.5% to 26.6% and barely changed with more reasoning.
Approval-token workflow attack bypasses checks on tool-using agents
In local OpenClaw reruns, the original configuration completed the target operation in 8 of 9 cases.
CVE-2026-71563: Apache Answer: MCP Answer Retrieval Exposes Content from Non-Public Questions
Apache Answer through 2.0.2 lets global API keys read answers on non-public questions.