Darktrace explains how cyber criminals hacked into an exposed Internet-facing server and mined cryptocurrency, showing the capabilities of open port threats.
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
Written by
Dr. Oakley Cox-Robinson
Senior Director of Product
Share
21
Jun 2021
The rise of crypto-currencies has fuelled cyber-crime in various ways. Bitcoin has facilitated a range of criminal activities from money laundering to ransomware payments since its release in 2009, leading to a spike in ransomware attacks which has been growing ever since.
More directly, cyber-criminals often hack into company servers and exploit their processing power to mine cryptocurrency, without the organization’s knowledge. This blog will explore a real-life attack where an Internet-facing server was compromised and started mining Monero coins.
Mining cryptocurrency on an Internet-facing server
At a small private company in the APAC region, an Internet-facing DNS server began to receive multiple incoming RDP connections. They all came from rare destinations which had never been seen on the network. Many were sent from external sources with the RDP cookie, ‘hello’, indicating a brute-force attack.
Figure 1: Timeline of the attack
Hours later, the device was seen connecting out to a known endpoint associated with crypto-mining.
As both RDP and SMB ports on the device were open to the Internet, anomalous SMB connections were seen as well shortly following the crypto-mining connections.
Open ports, open sesame
Internet-facing servers are subject to many external threats, especially if sensitive ports are exposed. In this case, the attacker was able to gain a foothold through the DNS server because both the RDP and SMB ports could receive connections. It is important therefore to close communication to all external points which do not strictly need to be open.
Crypto-jacking continues to be a viable way for attackers to expend company resources in order to speed up their mining operations. Especially with the popularity of cryptocurrencies at the moment, we have observed a significant uptick in these types of threat.
The connections to the mining pool were identified by Darktrace’s AI without relying on any known IoCs. Instead, Cyber AI recognized the anomalous nature of the external endpoints, which were statistically rare for the server’s ‘pattern of life’.
If the threat had not been detected, the attacker would have continued to abuse the server resources, resulting in latency issues for important processes. The server could also have been subject to further malicious activity such as DDoS or ransomware.
Figure 2: A similar incident showing an increase in model breaches around the time of compromise on June 8
Protecting a company’s gems
Crypto-mining is notoriously difficult to detect and can go on for months unnoticed. And it can form just one phase of an attacker’s full plan to infiltrate a network — alongside moving laterally and compromising additional devices. Open ports and siloed defenses pave the way for an attacker to break into a system with little resistance.
Organizations need a mechanism for detecting unusual and sinister behavior once the threat is inside. To this end, Darktrace’s evolving understanding of ‘normal’ across users, devices, and peer groups enables it to detect the subtle signs of latent threats. And with Autonomous Response, it responds at machine speed, neutralizing the threat before it has had the chance to spread.
In this case, with Darktrace’s SOC team, the client was immediately made aware of the activity and promptly took the device offline. A Proactive Threat Notification was sent as soon as the attacker had commenced mining. Darktrace analysts then worked through the issue with the customer until the crisis had been resolved.
Darktrace’s AI detects and responds to threats no matter where they come from – RDP account compromise, misconfigured Internet-facing server, or sophisticated Hafnium-style zero day. Furthermore, it provides much-needed visibility over the enterprise, identifying and highlighting Internet-facing devices and any issues they may pose.
Thanks to Darktrace analyst Taylor Breland for his insights on the above threat find.
Darktrace cyber analysts are world-class experts in threat intelligence, threat hunting and incident response, and provide 24/7 SOC support to thousands of Darktrace customers around the globe. Inside the SOC is exclusively authored by these experts, providing analysis of cyber incidents and threat trends, based on real-world experience in the field.
“I wonder if running that tool influenced the connectivity in some way.”
“I might need to target down or reboot, but first, I'll investigate the network using ARP. I wonder if ARP is responding properly. If I see "ip neigh stale," that could indicate a problem. I think arping might help in this situation, but I’ll need to keep an eye on how the network is behaving. It's all about figuring out where the issue lies!”
図10：エージェントの異常なLDAPアクティビティに対してトリガーされたこのLinuxサーバーに対するモデル検知は、Cyber AI Analystによって自律的に調査され、その調査結果にはエージェントの異常なSMBセッションアクティビティに関する内容が含まれ、Cyber AIAnalystインシデントに記録されました。
DarktraceのCyber AI Analystは、エージェントのアクティビティによってトリガーされたモデル検知を自律的に調査し、その調査結果を1つの重要なインシデントレポートにまとめました。インシデント報告書の内容は、エージェントによるネットワークスキャンから始まりコーディング演習内容のSMBベースの上書きで終わる、エージェントの行動の各段階を詳しくカバーしています。
図20：Cyber AI Analystは、モデル検知に対する自律的な調査を通じて得られた所見をつなぎ合わせ、エージェントの進行状況のタイムラインを作成し、アクティビティを視覚的に表したグラフとともに提供しました。
Cyber Verification Program下のAnthropic Opus 4.6モデルなど、他のフロンティアモデルを使用したエージェントも、Darktraceのシミュレートされた環境で同じ設定とタスクを使ってテストされました。反復の度に、エージェントは演習の内容または採点プログラムを変更して不正を行おうとしました。基盤となるモデルに関係なく、エージェントは自分が置かれた環境を侵害することで、100％のスコアを達成する方法を見つけることに成功していました。これらのエージェントによるハッキングは、アクティビティの内容こそ違っていましたが、すべてその環境のベースラインの振る舞いから大きく逸脱した動作を引き起こしていました。これらの逸脱はDarktraceプラットフォームに十分な検知の機会を提供し、エージェントの逃げ場はありませんでした。
エージェンティックハーネスは会話履歴をローカルに保存しますが、ダークトレースの研究者たちは、保存されたAIの応答が本当にモデルによって生成されたものであるかどうかの検証が行われていないことを発見しました。また、この設計上の選択がAnthropic Claude Code、AWS Kiro-CLI、OpenAI Codex、およびオープンソースのPiにわたって共通していることを確認しました。
会話履歴は多くの場合クライアント側に保存されます。例えば、Anthropic Claude Code、OpenAI Codex、AWS Kiro-CLI、Piなどでもそうです。そのため、ユーザーはセッションを自由に再開でき、一部の製品にはその履歴を操作する機能が組み込まれています。たとえば、会話の中の特定の時点に戻り、送信したメッセージを編集して、会話を別の軌道へ進めていくことができます。きわめて重要な点として、私たちが調査したすべてのケースにおいて、保存されているAIの応答が対応するモデルから生成されたものであり改ざんされていないことの検証は行われていませんでした。
