September Web3 Security Monthly Report: 48 Major Security Incidents, Over $760 Million in Stolen Funds

The deepest trend in Web3 security in 2026 is the systematic expansion of attack surfaces. Attacks are occurring simultaneously at the code level, in daily operations, and during interactions, with enhanced AI capabilities accelerating vulnerability discovery.

Written by: Beosin

According to monitoring data from the Beosin Alert platform, in September 2026, the total loss from various security incidents amounted to approximately $760 million, with a total of 48 major security incidents occurring. Among them, the Bitget security incident resulted in losses of about $387.5 million, and the Liquid Network suffered losses of approximately $320 million (with around $285 million returned subsequently), while the total loss from the remaining 46 incidents was about $52.95 million.

Top 10 Losses in September

On September 24, the Bitget platform's hot wallet was illegally hacked, with this attack spanning multiple mainstream blockchain networks, resulting in total losses exceeding $387.5 million, making it the most significant security incident in September. On September 6, the Bitcoin sidechain Liquid Network was maliciously minted with approximately 4,000 BTC (worth about $320 million) due to a verification cache vulnerability in the underlying software Elements. On September 8, the attacker returned 3,400 BTC, while the remaining 598.5 BTC is still under dispute as to whether it constitutes a "bug bounty." On September 22, Neutron suffered a governance attack that led to the theft of approximately 1.73 million ATOM from DeFi protocols such as Astroport, resulting in a total loss of $9.4 million, ranking third.

Additionally, among the Top 10 losses, there were 3 phishing/poisoning incidents: a phishing attack targeting address 0x40e93, app poisoning of FomoPeek, and a malicious backdoor in the iToken wallet. User security remains one of the weakest links in the Web3 ecosystem.

Types of Attacked Projects and Losses by Chain

This month, the attacked targets included exchanges, public chains, lending protocols, wallet applications, token contracts, cross-chain bridges, and ordinary users, among others. The exchanges suffered the highest losses, totaling $387.5 million; public chains had total losses of about $334 million; attacks targeting individual users/addresses caused losses of approximately $15.57 million, a decrease of about 45.2% compared to the previous month. DeFi contracts were attacked the most, with a total of 13 incidents; token contracts were attacked 10 times, ranking second.

The chain with the highest losses in September was Liquid Network. Ripple suffered losses of about $157 million due to the Bitget incident, ranking second, while Ethereum incurred losses of approximately $146 million, with a total of 14 security incidents. Currently, most DeFi protocols and phishing attacks targeting whales still predominantly involve Ethereum. The chain with the second most security incidents was BNB Chain, but the attack targets were still mainly token contracts, resulting in smaller losses.

Moreover, Cosmos ecosystem public chains such as Neutron and Injective also faced attacks in September, indicating a multi-chain attack trend.

Analysis of Major Security Incidents

For an analysis of the two major security incidents involving Bitget and Liquid Network, please refer to the previous articles published by Beosin. Below is an analysis of other attack methods this month:

1. Neutron: Governance Attack

The Neutron governance attack is a typical case of governance mechanism manipulation. The attacker took advantage of the time difference and staking snapshot mechanism in the governance rules to take over contract management rights at a very low cost. The attack path is as follows:

(1) The attacker prepared a proposal titled "AIATO: AI Agent Takeover. Phase 1: Agent Admin Registration," which superficially described an experiment on "autonomous AI agents taking over control of the test network operations." The proposal summary cited academic papers on decentralized governance by AI agents, aiming to confuse the governance parties:

(2) The proposal actually contained 11 "Update Admin" messages, which functioned to transfer the administrator rights of multiple contracts on the Neutron chain to the proposal address one by one. The involved contracts belonged to Astroport (a decentralized exchange) and Drop (a liquidity staking protocol).

(3) Low-cost purchase of voting rights. The attacker spent about 20,199 USDC to buy NTRN in five transactions on Astroport, then staked 31.6 million NTRN. This staking occurred 11 minutes and 39 seconds before the voting deadline, precisely within the effective window of the voting snapshot in the Cosmos SDK.

This proposal passed with an 82% approval rate (the fast track threshold is 67%, and the quorum is 30%), allowing the attacker to replace the contract code and withdraw funds. It was reported that Astroport lost approximately $4.9 million, and Drop lost about $4.5 million.

This attack is also a very typical on-chain protocol governance attack. For on-chain governance, project teams should establish the following checkpoints for prevention:

• Set absolute vote or capital thresholds: Governance proposals should not rely solely on relative proportions for approval. A hard threshold based on absolute numbers must be set, such as requiring the approval votes to reach a certain amount (e.g., $1 million) or a number of independent addresses.
• Equip time locks with guardians or cancellation paths: Although governance execution usually has delays, this only retains a certain reaction time. Project teams must equip effective guardians or proposal cancellation paths for the execution delay period. If a malicious proposal is discovered during the delay period, the guardian can intervene immediately and cancel it.
• Monitor governance participation: Protocols should establish real-time monitoring of governance participation across various treasuries. When a treasury's governance token total supply or voting participation rate is found to be abnormally low, timely alerts should be issued, and even automatic protective measures should be triggered.

2. Injective: Options Settlement Vulnerability

The Injective attack incident is a binary options settlement vulnerability attack targeting its native protocol module. The attacker exploited a failed oracle to trigger the protocol's "no price refund" logic, extracting approximately $4.9 million at a very low cost. The specific attack path is as follows:

(1) The attacker created an oracle named Frontrunner, whose data source was set to empty and would never return a price. At the same time, 299 binary options markets were created, all pointing to this oracle.

(2) Since the oracle could not provide a price, the market could not settle normally, forcing the system to enter the "no price refund" (NO_PRICE_FOR_REFUND) path. The design logic of this path is to refund double the margin instead of just the principal.

(3) The attacker simultaneously bought and sold in the market through multiple sub-accounts, acting as the counterparty. Since there were no real trading counterparts, the funds only circulated between the attacker's own accounts, but the refund mechanism would additionally generate profits "out of thin air." Within a window of about 30 minutes, the attacker deposited approximately 105,199 USDC but was able to withdraw 204,699 USDC.

The attacker then exchanged USDC for ETH and transferred it to Ethereum via the CCTP cross-chain bridge.

3. ASI Alliance: Private Key Leak

From September 20 to 21, the ASI Alliance ecosystem (Fetch.ai, NuNet, and SingularityNET) was attacked by the same attacker in succession. The attacker first exploited a leaked private key and contract design flaws to extract 8.7 million FET from Fetch.ai's TokenConversionManagerV3 contract, then used NuNet's minting permissions to issue 408.5 million NTX tokens, and finally minted approximately 896 million AGIX, 500.5 million WMTx, and 492.4 million CGV on Ethereum through the SingularityNET cross-chain bridge.

According to disclosures from ASI Alliance, the attacker leveraged the leaked private key to gain minting rights for multiple projects, resulting in a large amount of minting. Upon reviewing the on-chain attack process, Beosin found that in addition to the private key leak, there were also design flaws in the key functions of their contracts:

This function relied solely on the signature of a single external account (EOA) as authorization proof. The function neither implemented the checkLimits(amount) limit control mechanism present in conversionOut() nor required verification that the target asset had been locked or destroyed on another network. Therefore, as long as the signing key was leaked, all assets within the contract were exposed to risk without any buffer. The attacker exploited this flaw to call conversionIn() using the leaked authorization signature, clearing out 8,721,530 FET tokens from the contract in one transaction.

Web3 Security Threat Trends

The deepest trend in Web3 security in 2026 is the systematic expansion of attack surfaces. Attacks are occurring simultaneously at the code level, in daily operations, and during interactions, with enhanced AI capabilities accelerating vulnerability discovery. Relying solely on several security audits or tools cannot cover operational security, on-chain governance, and business logic vulnerabilities. This poses new challenges for exchanges, public chains, and Web3 project teams in building security defense systems.

Additionally, attacks targeting DeFi contracts and individual users are frequent. Contract vulnerabilities or authorizations can easily be exploited by attackers. Contract developers or operators should review the security of contracts, and for contracts handling core business, multiple and diverse security audits should be conducted. For individual users, it is advisable to regularly use blockchain explorers or revocation tools to check and cancel unused contract authorizations, and to stay informed about common and emerging phishing techniques, as well as the latest security alerts and developments, to enhance security awareness.