How and Why Bybit Was Hacked Despite Its High Security: The Human Factor Unveiled

How and Why Bybit Was Hacked Despite Its High Security: The Human Factor Unveiled

Good morning, friends!

Today, we’re diving into a fascinating yet cautionary tale: the massive hack of Bybit, one of the world’s leading cryptocurrency platforms, which took place on February 21, 2025. Despite its robust security measures, Bybit lost approximately 401,000 ETH—worth $1.4 to $1.5 billion—from its cold wallet. How did this happen? Spoiler alert: it wasn’t a flaw in the tech but a masterclass in exploiting the human element. Let’s break it down, based on information circulating as of March 29, 2025.

How the Bybit Hack Unfolded

Bybit relied on a cold wallet secured with the Safe Protocol (formerly Gnosis Safe), a multisig tool requiring multiple approvals—in this case, three signers—to authorize any transaction. This setup is considered one of the gold standards in crypto security. Yet, the hackers didn’t need to crack encryption or exploit a coding bug. Instead, they turned the platform’s own processes against it. Here’s how they pulled it off:

Social Engineering and UI Trickery
The attackers, suspected to be the infamous Lazarus Group from North Korea, used social engineering to deceive the signers. During what seemed like a routine transfer from the cold wallet to a warm wallet, they presented a fake user interface (UI). This masked UI mimicked the legitimate Safe interface, complete with a convincing URL and the expected destination address. The signers, trusting what they saw, approved the transaction.
The Malicious Contract Sneak
Hidden within the approved transaction was a malicious smart contract. This contract leveraged a function called delegatecall, which allows one contract to borrow code from another. Once executed, it swapped out the wallet’s legitimate logic with the hackers’ own version. Functions like sweepETH and sweepERC20 in the rogue contract gave them full control, letting them drain the 401,000 ETH into dozens of wallets they controlled.
Execution and Escape
With the signers’ approval secured, the hackers moved swiftly, splitting the stolen funds across multiple addresses to complicate tracking efforts. By the time the breach was noticed, the damage was done.

Why It Worked Despite Top-Notch Security

You’d think a multisig cold wallet would be impenetrable, right? Here’s why it wasn’t:

The Human Factor Trumped Tech
The hackers didn’t need to break the system—they broke the people operating it. The signers trusted the UI and didn’t double-check the underlying transaction data. Routine and familiarity dulled their vigilance, proving that even the best tech is only as strong as its human overseers.
High Security, Low Suspicion
Bybit’s reliance on multisig and offline storage bred confidence, but that confidence may have led to complacency. Routine transfers became mechanical, and the signers didn’t scrutinize the details closely enough.
A Sophisticated Adversary
The Lazarus Group is known for its advanced tactics—phishing, malware, and UI spoofing. They likely studied Bybit’s operations for months, targeting the signers with precision and patience.

How They Studied and Outsmarted the Humans

The hackers didn’t strike blindly. Their success hinged on meticulous preparation:

Reconnaissance
They monitored Bybit’s processes, learning who the signers were, how they used Safe, and when transfers happened. This could have involved phishing emails, intercepted communications, or even an inside leak.
Psychological Manipulation
The fake UI was a psychological trap—perfectly mimicking the real thing, down to the URL and destination address. It gave the signers no reason to doubt it.
Timing the Routine
Daily cold-to-warm wallet transfers are standard for exchanges. The hackers waited for a busy moment when the signers were likely distracted or rushed, lowering their guard.

How They Hijacked the Contract with an Update

The technical brilliance of this hack lies in how they altered the contract:

The Delegatecall Trick
Safe’s flexibility with delegatecall became its Achilles’ heel. The hackers crafted a “Trojan horse” contract that, once approved, replaced the wallet’s logic. By mapping variables like _transfer to the wallet’s storage slots (e.g., SLOT[0]), they rewrote the rules.
Seamless Update
The signers’ approval triggered this update, handing the hackers unrestricted access. No further checks were needed—the funds were theirs to sweep away.

The Takeaway

This breach wasn’t a failure of technology but a triumph of human exploitation. The hackers turned Bybit’s security into their own Trojan horse by studying behavior, exploiting trust, and striking at the perfect moment. It’s a stark reminder that even the most secure systems are vulnerable if the humans behind them let their guard down.

What do you think—how can crypto platforms better protect against these human-centric attacks? Let me know in the comments!

Comments

Turn Your Old PC That Can’t Upgrade to Windows 11 into a Powerful Tool for Preppers & Tech Savers

Turn Your Old PC That Can’t Upgrade to Windows 11 into a Powerful Tool for Preppers & Tech Savers Have an old PC gathering dust because it doesn’t support Windows 11 due to TPM 2.0 or hardware limitations? Don’t worry—you can give it a new lease on life! Instead of throwing it away, transform it into a secure, offline tool for prepping or tech-savvy projects. In this guide, we’ll show you how to install Lubuntu, a lightweight Linux distribution, and DeepSeek R1, an offline AI model, to create a system ready for blackouts, crises, or everyday use. With a strong focus on cybersecurity, this setup is perfect for preppers gearing up for the unexpected and tech savers looking to repurpose old hardware. Why Do This? Older PCs (from 2015-2018, e.g., with Intel 6th/7th Gen CPUs or 8GB RAM) are still capable of many tasks. In scenarios like the 2021 Spain blackout, access to information without internet and data security are critical. With Linux and DeepSeek, you can build a secure, offl...

Linux time for some time

Benefits of Using Linux Free and Open-Source No license fees—ever. You can download, use, and even modify Linux distros (distributions) like Ubuntu or Linux Mint at no cost. This is a huge win for budget-conscious users compared to Windows’ price tag. Lightweight and Efficient Linux can run smoothly on older hardware. Distros like Lubuntu or Xubuntu are designed for low-spec machines, often needing just 1-2 GB of RAM and a basic CPU—way less than Windows 11’s demands (4 GB RAM, TPM 2.0, etc.). Highly Customizable Users can tweak everything: desktop environments (e.g., GNOME, KDE, XFCE), themes, and even the kernel itself. Want a Windows-like interface? Linux Mint with Cinnamon has you covered. Prefer something sleek and modern? Try Pop!_OS. Security and Privacy Linux is less prone to viruses and malware due to its architecture and smaller user base (less of a target). Plus, it doesn’t harvest your data like some proprietary OSes—updates are about fixes, not ads. Regular Updates...

Convolutional Neural Networks

Convolutional Neural Networks (CNNs or ConvNets) Convolutional Neural Networks, are a class of deep neural networks most commonly applied to analyze visual imagery. They have revolutionized the field of computer vision and are widely used in tasks like image recognition, image classification, object detection, and even in some aspects of natural language processing and time series analysis. Here's a breakdown of their key features and components: Key Features: Local Receptive Fields : CNNs maintain the spatial relationship between pixels by learning features using small squares of input data (local patches). This reduces the number of parameters and computations. Shared Weights : The same weights (or filters) are used for several locations in the input, which means the network learns features that are invariant to translation. Pooling : Typically, CNNs include pooling layers (like max pooling or average pooling) which reduce spatial size, thus reducing computation, memory usage, an...

Indirect Prompt Injections

ALEXICACUS BLOGGER CYBERSECURITY ISSUES INDIRECT PROMPT INJECTIONS Recent Kaspersky Lab's investigation into indirect prompt injection highlights a significant cybersecurity concern for systems utilizing large language models (LLMs). Here's a breakdown of the issue: What is Indirect Prompt Injection? Definition : Indirect prompt injection involves embedding special phrases or commands within texts (like websites or documents) that are accessible online. These commands are designed to manipulate the behavior of AI models when they process these texts. Mechanism : When an AI, particularly those using LLMs like chatbots, processes content from these sources, it might inadvertently include these injections in its response generation process. This can lead to: Manipulation of Output : The AI might provide responses that serve the interests of the party who embedded the injection rather than the user's query. Privacy Concerns : Potentially sensitive data could be extracted or ...

AI detection accuracy of security solutions

AI Detection Accuracy of Cyber Security Solutions Comparing AI detection accuracy for phishing and email security solutions like Proofpoint, Mimecast, Barracuda, Sentinel, Abnormal Security, Cofense, Ironscales, and SlashNext involves looking at several reports, user reviews, and independent assessments. Here's a comparative analysis based on available data: Proofpoint : Detection Accuracy: Known for high accuracy in detecting a broad spectrum of email threats, including sophisticated phishing and BEC attacks. Proofpoint uses AI, machine learning, and dynamic analysis for threat detection. False Positives: Efforts are made to keep false positives low, but user feedback sometimes mentions a need for tuning to reduce them. Mimecast : Detection Accuracy: Mimecast employs AI to analyze emails for phishing and other malicious content. It's praised for its effectiveness but can have issues with false positives, particularly with new or emerging threats. False Positives: Users ...

AI security measures to protect AI systems

AI security measures are crucial to protect AI systems from various threats, including data breaches, adversarial attacks, model poisoning, and the kind of prompt injection discussed previously. Here's a comprehensive overview of key security measures for AI: Data Security Encryption : Encrypt data both at rest and in transit to protect against unauthorized access. Access Control : Implement strict access controls, ensuring only authorized users or systems can interact with or modify data used by AI models. Model Security Secure Model Development : Adversarial Training : Train models with adversarial examples to make them more robust against attacks that aim to mislead the AI. Regular Updates : Update models with new data and retrain them to adapt to new threats or attack vectors. Model Monitoring : Anomaly Detection : Use systems to detect unusual behavior or outputs from AI models which might indicate a security breach or model manipulation. Audit Trails : Keep logs of all model ...

The "best" AI search engine

Searching... Asking the Right Questions: How to Get the Best Answers from AI Artificial Intelligence is transforming the way we learn, work, and explore the tech world. Whether you’re diving into convolutional neural networks, bolstering your cybersecurity defenses, or just curious about the latest AI trends, tools like AI assistants can be game-changers. But here’s the catch: to get the right answers from AI, you need to ask the right questions. On Alexicacus, we’re all about empowering you with tech knowledge, so let’s break down how to master the art of asking questions to unlock AI’s full potential. Why Asking the Right Questions Matters AI systems, like the ones you might interact with on this blog (shoutout to our friend Grok!), are designed to process vast amounts of data and provide answers based on patterns and logic. But they’re not mind readers. The quality of the answer you get depends heavily on how you frame your question. A vague or poorly structured question can le...

Alexicacus

Search This Blog