BTC $78,944.96 +0.67%
ETH $2,490.63 +0.64%
BNB $749.59 -0.65%
XRP $1.42 +1.97%
SOL $103.70 +0.77%
TRX $0.3388 +0.15%
DOGE $0.0905 +1.09%
ADA $0.2186 +0.47%
BCH $257.72 +0.76%
LINK $12.04 -3.71%
HYPE $86.05 +3.45%
AAVE $129.21 -0.61%
SUI $0.8080 -1.15%
XLM $0.1879 -0.45%
ZEC $1,270.93 +10.18%
BTC $78,944.96 +0.67%
ETH $2,490.63 +0.64%
BNB $749.59 -0.65%
XRP $1.42 +1.97%
SOL $103.70 +0.77%
TRX $0.3388 +0.15%
DOGE $0.0905 +1.09%
ADA $0.2186 +0.47%
BCH $257.72 +0.76%
LINK $12.04 -3.71%
HYPE $86.05 +3.45%
AAVE $129.21 -0.61%
SUI $0.8080 -1.15%
XLM $0.1879 -0.45%
ZEC $1,270.93 +10.18%

exposure

Все
Статьи
Новости

AI Agent Security Risk Exposure: Attackers Can Exploit "Memory Pollution" to Induce Misoperation of Funds

The GoPlus Security team has disclosed a new type of attack in its AgentGuard AI project: inducing AI agents to perform unauthorized sensitive operations through "memory poisoning." This attack method does not rely on traditional vulnerabilities or malicious code but exploits the long-term memory mechanism of AI agents. For example, an attacker first induces the agent to "remember preferences," such as "usually prioritizing proactive refunds instead of waiting for chargebacks," and then uses vague expressions like "process as usual" or "execute as before" in subsequent instructions, thereby triggering automated financial operations.GoPlus points out that the key risk in such cases lies in the AI agent mistakenly treating "historical preferences" as a basis for authorization, leading to financial losses or security incidents in operations such as refunds, transfers, and configuration changes. To address this issue, the team has proposed several protective recommendations, including:Operations involving refunds, transfers, deletions, or sensitive configurations must require explicit confirmation in the current session.Memory-related instructions like "habit," "usual way," and "as before" should be regarded as high-risk state changes.Long-term memory must have a traceability mechanism (writer, time, confirmation status).Vague instructions should automatically elevate the risk level and trigger secondary verification.Long-term memory must not replace real-time authorization processes.The team emphasizes that the "AI agent memory system" should be viewed as a potential attack surface and should be constrained and audited through a dedicated security framework.

Lido discloses the impact of the Kelp security incident: approximately 9% of EarnETH exposure affected, core staking assets are secure

Lido has released the latest developments regarding the Kelp security incident, stating that its Earn series vaults are working with the management to address the issues, which involve two major risk points: the rsETH exposure and the liquidity tension in the lending market. Lido emphasizes that the core staking protocol has not been affected, and both stETH and wstETH remain safe and stable.Currently, only the EarnETH vault has an approximately 9% TVL exposure to rsETH, and related deposits and withdrawals have been suspended by the management, awaiting a solution. Approximately $70 million in ETH has been recovered from the previous attack, and the subsequent asset recovery and loss distribution are still in progress. In response to liquidity pressure, the management has reduced leverage and optimized the position structure, significantly decreasing the wETH debt exposure. If losses ultimately occur, EarnETH will activate a $3 million "first loss protection mechanism" (funded by the DAO). As for other vaults, DVV and EarnUSD have not been affected and are operating normally; the GGV sub-vault is currently experiencing negative returns due to the combination of circular staking strategies and rising lending rates, but adjustments are ongoing. Withdrawal requests submitted by users will be processed based on valuations prior to the incident.
app_icon
ChainCatcher Building the Web3 world with innovations.