Sandeep

Cyber Security | Foodie | Philosopher to be

[SS @just_infosec_]

An open-source goal-based adaptive red teaming tool for LLM applications.

Attacker LLM vs Target LLM + Judge LLM in a smart evolving loop.

Strengthen your LLMs against real attacks.

https://t.co/ldztiQCToL

#LLMSecurity #RedTeaming #OpenSource #AI

[SS @just_infosec_]

I just published AI vs AI: An Open-Source Adaptive Red Teaming Tool for LLM Applications

https://t.co/fnX4IRUYNX

#promptinjection #llm #redteam

[SS @just_infosec_]

DeepProbe: Open Source Memory Forensics Automation
Deterministic Rule Based Engine + Local AI to explain.

Attack Chain + MITRE Mapping in Minutes

Full Write-up: https://t.co/VJyOnzLLI3

#dfir #volatility #forensics

[SS @just_infosec_]

Most LLM security tools replay static jailbreak prompts.

AIvsAI actually adapts.

Paste a raw HTTP/cURL request from Burp Suite and it automatically parses + tests for "Goal Based Prompt Injection" to the LLM behind it using an adaptive AI-vs-AI loop.

The attacker LLM learns from refusals, mutates prompts, and keeps iterating toward the objective.

https://t.co/VB73JQe39C

Write-up:
https://t.co/3WQBci5fsg

#AISecurity #PromptInjection #RedTeam

[SS @just_infosec_]

Cyber Security:
My advice to Students:

- Master the basics
- Fancy stuff can wait

The industry will teach you the rest.

[SS @just_infosec_]

AI ML Engineers should know a little about best security practices

Infosec guys shud understand AI for cyber security & Cyber Security for AI

I tried to develop a developer centric tool for AI specific Threat Modeling.

Try: https://t.co/EIL148rmFk

-Free Hosting, slow loading

[SS @just_infosec_]

Just dropped DeepProbe: An open-source, rule-driven memory forensics toolkit with automated detections, correlation, and MITRE ATT&CK context.

Ready to turn raw memory into actionable intelligence.

Collab & feedback welcome

🆕 GitHub: https://t.co/3esjNRYS5g

#DFIR #SOC

[Sandeep @eistrack]

Threat modeling AI systems is hard because traditional frameworks weren’t designed for them.
That’s why this tool exists.

[Sandeep @eistrack]

@NoseNothin @techspence 😂

[Sandeep @eistrack]

@hetmehtaa In general I ask, "You are supposed to protect a web portal, how will u start? how will u assess progress, 30-60-90 days plan"

[Sandeep @eistrack]

"Penetration Testing of AI: Why and How?"

https://t.co/CC63Z1gNdU

#cybersecurity #infosec #AI #ML #PenetrationTesting

[Sandeep @eistrack]

@techspence A Fool with a tool is still a Fool

[Sandeep @eistrack]

@hetmehtaa Yes

[Sandeep @eistrack]

@4n6lady Also a little hope that attacker will honor our exception list for EDR/Vulnerability Management and wont go there.

[Sandeep @eistrack]

@IceSolst Wen it comes to big web portals, nobody is actually able to have an inventory of all endpoints.
Some zombie endpoints disconnected from portal but still alive, endpoints being sent in emailers.
And u get compromised by these missed endooints.

[Sandeep @eistrack]

@techspence U install edr on every new machine(cloud too) as per the process.
But for some reason agent is not working or reporting back.
Few companies do this reconciliation quarterly, few monthly.

Real time identification of machines, not running edr is not difficult though.

[InfoSec Community @InfoSecComm]

🚨 New Writeup Alert! 🚨

"Bad Bots: The Unseen Cyber Threat and the Fight to Secure the Internet" by SS is now live on IW!

Check it out here: https://t.co/7XBNfOpIAK

#scraping #infosec #cybersecurity #badbots #bots

[InfoSec Community @InfoSecComm]

🚨 New Writeup Alert! 🚨

"Building an Integrated Threat Intelligence Platform Using Python and Kibana" by Sandeep Saxena is now live on IW!

Check it out here: https://t.co/OAQSbzfZHM

#threatintel #threatdetection #threatintelligence