New defense system ORAN-DEFEND targets backdoor attacks in Open RAN

Researchers have developed ORAN-DEFEND, a new system designed to protect Open Radio Access Networks (O-RAN) from backdoor attacks embedded in third-party deep reinforcement learning (DRL) xApps. This defense mechanism operates without retraining the compromised xApp by projecting KPI windows onto a safe subspace identified through singular value decomposition of trusted data. The system's effectiveness is contingent on the backdoor trigger's energy concentrating in a subspace orthogonal to the safe one, with empirical results on the Colosseum COLORAN dataset showing a 100% return recovery and over 99.5% defense success rate against various DRL backdoor attacks. AI

IMPACT This research introduces a novel defense against sophisticated supply-chain attacks targeting AI components in critical infrastructure like Open RAN.

RANK_REASON The item is a research paper detailing a new defense mechanism for a specific technological domain. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• BadRL
• Colosseum COLORAN
• deep reinforcement learning
• Mohammad Raihan Uddin
• Open Radio Access Network
• ORAN-DEFEND
• Q-Incept
• singular value decomposition
• SleeperNets
• TrojDRL
• xApps

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →