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
Max Heinemeyer
Global Field CISO
Share
27
Jul 2020
First observed in 2015, LeChiffre ransomware works by encrypting files and changing their extensions to .lechiffre (translated as ‘cipher’ in French). Unlike other variants of malware, hackers must run LeChiffre manually on a compromised system. Cyber-criminals automatically scan networks in search of poorly secured desktops, and then log in remotely and manually run an instance of the virus.
To profit from the current wave — and success — of ransomware, many low-level threat groups are using mutations of known ransomware. Despite the multiple recovery tools available online, LeChiffre remains a persistent and highly disruptive threat across all industries, with new mutations enabling it to bypass many legacy security solutions.
Recent activity in a Darktrace customer environment
Darktrace recently discovered a resurgence of LeChiffre ransomware whilst deployed in the network of a wholesale distributor in the US.
Over the course of a week, several devices were seen making HTTP requests for Mimikatz Powershell script, then authenticating with administrative credentials for the first time. The infected devices then exhibited network scanning activity on port 445. Shortly after, the devices began writing multiple files (svcctl, PsExec) over SMB to hidden file shares on multiple other devices.
The activity took place early on a Sunday morning, when the security team wasn’t closely monitoring the network and didn’t have eyes on the activity. However, Darktrace’s AI detected the highly suspicious activity in real time, monitoring 19.7k of SMB move successes, with .LEV62C_LeChiffre extension and almost 38.8k of SMB write successes with LEV62C_LeChiffre_ReadMe.txt — the ransom note.
This threatening activity escaped detection from the rest of the organization’s security stack, likely due to the fact that the download of the Mimikatz Powershell script was from a redirected, legitimate domain - githubusercontent[.]com. The use of this legitimate domain allowed it to go unnoticed by firewall and endpoint detection, while this particular variant of the LeChiffre malware was not yet recognized by signature-based tools.
However, with a continuously updating understanding of what constitutes ‘normal’ behavior for every network and device in an organization, Darktrace’s AI quickly detected the unusual activity and flagged it as highly threatening. The Cyber AI Analyst then fully investigated the attack, autonomously connecting the dots and surfacing the five stages of the attack chain ready for incident review. This greatly reduced the incident response time for the security team, who were able to isolate the infected devices and quickly take them offline.
A sample of the AI Analyst report of the SMB file encryption is shown below.
Figure 1: Darktrace’s Cyber AI Analyst autonomously investigates the attack
Overview of infected device
Figure 2: The graph showing the Darktrace detections cluster on June 21
Figure 3: The device event log showing the list of model breaches on June 21
This incident happened while the customer was participating in Darktrace’s free 30-day Proof of Value. Had Darktrace’s Autonomous Response technology, Antigena, been enabled, the AI would have responded immediately, taking action when the malware attempted lateral movement to prevent the breach device from carrying out encryption. Even without Antigena, Darktrace’s real-time detection and AI-powered investigation enabled the security team to quickly de-escalate the compromise.
The barrier to entry has been significantly lowered by the commoditization of malware. Many forms of ransomware are easily available to low-level threat actors via Darknet purchases, and minor changes made by cyber-criminals can easily enable the attacks to go undetected by most security tools. Now even low-skilled threat actors have the potential to cause significant damage and evade the traditional security stack.
The events described above demonstrate that ransomware continues to present a very real threat across multiple industries. And with ransomware capable of encrypting files across hundreds of devices in minutes, the need for machine-speed response is more important than ever. Darktrace Antigena enables security teams to get ahead of a threat by autonomously responding to threatening activity, taking devices offline or enforcing their ‘pattern of life’ to keep businesses operational.
Thanks to Darktrace analyst Fiona Goh for her 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。
