AI-Driven Campaign Exposes Over 5,200 Indian Academic Records via Brute Force

A threat actor operating under the handle BUGSTER employed DeepSeek V4 to automate a cybercrime campaign targeting Indian organizations, including universities. According to Unit 42, Palo Alto Networks’ threat-intelligence team, the operator successfully breached a university website and exfiltrated a database containing more than 5,200 student and faculty records. The stolen data included personally identifiable information, demonstrating how accessible AI tools can lower the barrier for executing targeted intrusions against educational institutions.

How did the attacker gain access?

The intrusion began with AI-generated reconnaissance that mapped 12 subdomains of the target’s online education platform. Unit 42 states the actor used WordPress REST API data to identify 13 user accounts. The threat operator then generated a password-guessing tool using 117 usernames and 28 common passwords. This automated effort resulted in a valid login credential. In a separate vector, the actor accessed an administrator account protected by a weak password to exfiltrate a MongoDB database.

What other capabilities were observed?

Beyond the university breach, Unit 42 found the actor developed credential-scanning tools to search for exposed cloud keys and private secrets. The operation included scripts for cryptocurrency-holder reconnaissance and a command-and-control listener supporting reverse-shell connections. Additionally, the actor built an automated pipeline for cracking Wi-Fi passwords. A nearby Linux device captured authentication material, which was sent to cloud GPU instances for processing, suggesting local physical proximity to the targeted wireless networks.

What tools supported the operation?

Unit 42 researchers report the actor ran the Hermes Agent framework with DeepSeek V4 and used Cursor, an AI coding environment, as an additional utility. The operator controlled specialized Hermes profiles through Telegram, including one named “alice” that advised on VoIP fraud and another called “pandarouter” for researching carriers. The threat actor also tested other AI models and used prompt-level overrides to reduce safety restrictions, exploring the setup of fraudulent Voice over Internet Protocol operations.

What is the current status?

No patch or fix has been confirmed by the affected university or the software vendors mentioned in the report. The findings highlight the risks of weak administrative credentials and exposed API endpoints when combined with automated AI reconnaissance. Organizations should review their exposure to common password attacks and ensure that administrative interfaces are protected by strong authentication mechanisms.

What analysts watch for

Teams usually examine logs for rapid, automated login attempts against known administrative endpoints. Analysts check for unusual API usage patterns that might indicate enumeration of user accounts. Persistence mechanisms often involve reverse shells or unauthorized cloud resource usage. Investigators look for connections to known command-and-control infrastructure and review access logs for exfiltration of sensitive databases. Focus on identifying compromised credentials and ensuring that API endpoints are properly secured against brute-force attacks.

Background: Weak password policies

Common password rules often reduce security. Forcing regular changes, banning symbols, and limiting length create predictable patterns. True security relies on checking against known breached credentials, allowing long passphrases, and enforcing multi-factor authentication rather than complex character requirements.

Read the full guide: Weak Password Policy Myths: What Actually Breaks Your Security

What to do and how to stay safe: BUGSTER

• Enforce strong, unique passwords for all administrative accounts and implement multi-factor authentication to prevent brute-force successes.
• Monitor authentication logs for unusual spikes in failed login attempts or access from unfamiliar IP addresses.
• Restrict public access to administrative APIs and ensure that subdomains are properly secured and monitored for reconnaissance activity.
• Review cloud infrastructure for exposed keys or secrets that could be harvested by automated scanning tools.

Step-by-step guide: Account Takeover vs Identity Theft: The Critical Difference

General security guidance from the Sandbox Desk newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which AI model was used in this campaign?

The threat actor used DeepSeek V4 as the reasoning engine for their AI-assisted cybercrime operation, according to Unit 42.

How many records were stolen from the university?

The attacker stole a database containing more than 5,200 student and faculty records with personally identifiable information.

Did the attack involve physical presence?

The Wi-Fi cracking workflow suggests local physical proximity to targeted networks, as a nearby device captured authentication material.