Skip to content

OPEN TO SOC / DFIR INTERNSHIPS · 56 HANDS-ON LABS · 20 VERIFIED CREDENTIALS

Manichand Gupta

Security Analyst

Security analyst working the full defense loop — 36 CyberDefenders and 20 HackTheBox Sherlock investigations, a self-built seven-node SOC lab, and open-source tooling for detection, IOC enrichment, malware and phishing analysis.

analyst@meerut:~
SCROLL

The brief.

SECTION 02 / ABOUT

Analyst behind the glass

Security analyst based in Meerut, UP, finishing a BCA in Cloud & Cybersecurity (2024–2027) at IIMT University. Works across the full defense loop: 36 CyberDefenders and 20 HackTheBox Sherlock investigations, a seven-node SOC lab with 148 Wazuh rules and 87 suppressions, seven published ATT&CK-mapped cases, and open-source tooling for detection, IOC enrichment, malware and phishing analysis. KC7 World #836 of 174,154; CyberDefenders World #4,017 / India #229; 20 verified credentials.

Color shifts by section, mapped to Defense-in-Depth layers — perimeter to data.

KC7
Under 0.5%
World #836 of 174,154 players
Global rank #836 of 174,154 players on the KC7 profile ↗
Verify proof ↗
Investigations
0
Hands-on labs
36 CyberDefenders investigations and 20 HackTheBox Sherlock labs across web, Active Directory, memory, disk and log forensics
Threat Intelligence
0
YARA Signatures
Aggregated across 40+ threat intel feeds and source repositories
CyberDefenders
India #229
World #4,017
India #229 / World #4,017 on the CyberDefenders profile ↗
Verify proof ↗
Research
0
Write-ups Published
Published DFIR investigation write-ups with IOCs mapped to ATT&CK

SECTION 02A / DEFENSE IN DEPTH

Layers I work in

The tools behind each layer are listed in the toolkit and mapped in MITRE ATT&CK coverage.

Network layer

SIEM correlation, traffic inspection, and protocol-level threat detection

Wazuh Microsoft Sentinel Splunk Elastic Stack Wireshark tcpdump Nmap

Click any ring · Tab + Enter works too.

PERIMETER NETWORK ENDPOINT APP DATA

SECTION 03 / EXPERIENCE

Hands on the queue

Two threads of practice: the analyst workflow inside the seven-node SOC lab, and the public investigations and released tooling built alongside it.

Self-Built SOC Detection & Monitoring Lab

Independent

JUL 2026 — OCT 2026
  • Monitored Wazuh and Suricata alerts across a seven-node hybrid lab spanning VMware and Azure
  • Triaged a 9,567-alert feed down to 46 alerts across three incident cases using 148 rules and 87 suppressions
  • Contained a live intrusion — severed reverse shells, quarantined webshells, and invalidated forged Kerberos tickets via a double krbtgt reset
  • Documented IOCs, root causes, containment, and ATT&CK mappings for each case, then hardened the root-cause flaws
Wazuh Suricata Sysmon TheHive Velociraptor MITRE ATT&CK

Independent Investigations & Security Tooling

Self-directed

2025 — PRESENT
  • Completed 36 CyberDefenders investigations and 20 HackTheBox Sherlock labs across web compromise, Active Directory, memory, disk, log and network forensics
  • Published seven ATT&CK-mapped investigations — Stealc infostealer, a live job-phishing campaign, WebStrike and PoisonedCredentials PCAP walkthroughs
  • Built and released RuleForge (8 SIEM dialects, 1,247 tests), an IOC enrichment platform (33 sources, 31 indicator types) and a 2,778-rule malware analyzer
  • KC7 World #836 of 174,154 (under 0.5%) · CyberDefenders World #4,017 / India #229 · 20 verified credentials
CyberDefenders HackTheBox KC7 YARA Sigma Python

SECTION 03A / SOC LAB

A fake company, built to be attacked

Seven machines that act like a small business: a firewall, a Windows domain with real user accounts, a website to break into, and a place to keep the evidence. Five run on my own computer, two run in Azure, and a private tunnel joins the two halves.

I ran a 74-minute attack against it, then worked the alerts like a real analyst. The tabs below say what each machine does. Everything after that — the detections, the three cases, the cleanup — came out of this one network.

The 46 alerts it produced were triaged into three cases in Hands on the queue.

WireGuard · UDP 51820 AZURE VNET · 10.122.0.0/16 ATTACK SEGMENT VICTIM NETWORK · 192.168.56.0/24 CASE MGMT · DFIR TheHive + Velociraptor SIEM · INDEXER Wazuh (Azure) · 4.14.8 ATTACKER Kali NIDS · ROUTER · VPN Suricata Gateway 8.0.6 DC · AD DS · SYS Win Server 2022 192.168.56.10 WIN 10 · SYSMON Workstation 192.168.56.20 WEB · AUDITD DVWA · Ubuntu 192.168.56.30

Wazuh SIEM (Azure)

Runs in Azure · version 4.14.8

This is the SIEM: the one place every machine sends its logs, and the tool that decides what deserves an alert. Wazuh checks each log against 148 rules I wrote: 61 that spot attacker behaviour, and 87 that stay quiet when normal work happens. Anything serious (Level 8 or above on its severity scale) goes straight to the case tracker.

148 rules87 noise filtersLog correlation

7

Nodes

148

Detection rules

46

Alerts / 3 cases

DETECTION ENGINEERING

Four sets of eyes, each covering the others

No single sensor sees the whole attack. Each one has a blind spot, and the next one is placed to catch what it misses. Four layers, in the order the evidence passes through them.

L1 · THE NETWORK

Suricata NIDS

Suricata 8.0.6 reading raw packets on the gateway

Watches the traffic itself. It sees Nmap scanning for open ports, SQL injection hidden inside a web request, and Kerberos ticket requests — including the ones no machine on the network ever writes down.

Blind spot — cannot read encrypted HTTPS or SSH, or commands typed on a host.

L2 · THE LINUX BOX

Auditd + bashcmd

Auditd inside the Linux kernel + Rule 100220

Watches the Linux machine from the inside. Every program the web server user starts is recorded, and a custom rule saves the exact command lines typed into a shell — pipes, built-ins and chained commands included.

Blind spot — Linux only; it sees no packets and nothing on Windows.

L3 · THE WINDOWS BOXES

Sysmon + Security logs

Sysmon events 1 and 3 · Windows events 4624 / 4720 / 4728

Watches the Windows machines from the inside: every program that starts with its full command line, every connection it opens, registry changes, who logged in, and who was handed admin rights.

Blind spot — Windows only; blind to the wire and to Linux.

L4 · THE SIEM

Wazuh SIEM

Wazuh · 148 rules · 87 noise filters

Puts all four streams in one place and decides what matters. It reads the 1,300+ lines of rules I wrote, uses lookup lists to mute noise it already knows is harmless, and sends anything serious (Level 8 or above) to the case desk.

Blind spot — only sees what the agents and decoders forward.

ALERT PIPELINE

The alert pipeline I had to rebuild

My first pipeline used Shuffle, a no-code automation tool. It worked fine in testing and fell over under a real attack: 400 automated alert calls (webhooks) arrived in 10 seconds, its listeners overflowed, the scripts it launched blocked waiting on the network, and the SIEM’s own engine (wazuh-analysisd) fell behind and started dropping alerts.

So I replaced it with a two-part Python pipeline. Part one writes every Level 8+ alert into a small local database file (SQLite, in a write mode called WAL that never makes the SIEM wait) — no network call, saved in under 0.2 ms. Part two is a background service that wakes every 5 s, works out which tool caused the burst (Nmap, Nikto, Hydra, sqlmap), groups related alerts together, and sends one clean batch to TheHive.

If new alerts land while a batch is on its way, a version counter marks the group as changed and it gets sent again — so nothing is lost between ticks.

< 0.2 msto save one alert
5 show often the sender wakes
4.6 salert to case, typical (fastest 1.4 s)
0alerts lost during floods
18.9 msto store 500 alerts (95% under 23.0 ms)
17 / 17pipeline self-tests passing

DETECTION SCORECARD

An honest score: 14 / 2 / 2 / 2

Claiming 100% detection would be a lie. Of the 20 attack steps, here is what the lab caught directly, what it only caught from a second machine, and what it could not catch at all.

14Caught directly70.0% · steps 01–03, 05–08, 10, 13, 15, 17–20
2Caught only from another machine10.0% · step 14 DCSync, step 16 Golden Ticket
2Missed — nothing was logging it10.0% · step 09 L2 drops, step 12 LDAP event 1644
2Missed — happened off the network10.0% · steps 04 and 11 GPU cracking
16 / 20Covered in practice · 80%No rule or decoder failed — nothing was lost silently

9,567 raw log events → 808 worth keeping (11 filters) → 46 alerts in TheHive → 3 closed cases

THE CAMPAIGN

74 minutes, 20 steps, three cases

One controlled break-in, start to finish: from the first port scan outside the network to full control of the Windows domain. Every step is tagged with a MITRE ATT&CK technique — the industry list of attacker behaviours — 20 in total, across 9 tactics, then written up as three cases in TheHive, all resolved and closed.

CAS-001-WEBCOMPSteps 01–07

Breaking in through the website, then becoming root

Nmap scanned for open ports. Hydra guessed a password. SQL injection dumped the user table. A PHP web shell went up, then a Netcat reverse shell gave a remote command prompt. A sudo misconfiguration let that shell become root, and an SSH key was planted so the attacker could come back later.

Resolved · Closed

CAS-002-LATMOVESteps 08–14

Moving inward, stealing credentials, climbing the domain

The attacker scanned the internal network. Responder poisoned Windows name lookups (LLMNR/NBT-NS) to capture password hashes, which Hashcat then cracked offline. BloodHound mapped who could reach what over LDAP. Kerberoasting pulled a service account ticket (HANK_SHERMAN), and DCSync copied the domain’s password database (NTDS).

Resolved · Closed

CAS-003-DOMIMPSteps 15–20

Taking the domain, then causing damage

Pass-the-Hash over WMI reused a stolen hash to reach another machine. A forged Kerberos Golden Ticket let the attacker sign in as any account, admins included, without a password. A second Domain Admin account was created. Documents were gathered and copied out over SCP, the wallpaper was replaced, and fake ransomware encrypted files and forced a reboot.

Resolved · Closed

RESPONSE

Stop the bleeding, then fix the cause

7 containment actions

  1. Blocked the attacker’s address at the gateway, both incoming and passing through
  2. Killed the reverse shell and quarantined shell.php
  3. Deleted the attacker’s root SSH key and closed the sessions it opened
  4. Disabled the backdoor account and removed it from Domain Admins
  5. Reset krbtgt twice (to KVNO 3), which makes every forged ticket worthless
  6. Quarantined the staged files and put the tampered wallpaper back
  7. Changed the DVWA, MariaDB and web-app passwords

7 root causes erased

  1. Removed the sudo rule that let python3 run as root with no password (NOPASSWD)
  2. Turned off direct root login and password login over SSH (PermitRootLogin no + PasswordAuthentication no)
  3. Stopped the uploads folder from running scripts
  4. Demoted the service account, deleted the name it advertised to other services (its SPN), and forced stronger AES encryption for Kerberos tickets
  5. Moved privileged accounts into the Protected Users group
  6. Disabled LLMNR and NetBIOS-over-TCP/IP — the old Windows name-lookup protocols — on every machine
  7. Split the flat network into firewall zones that allow only the traffic that is needed

EVIDENCE

The evidence register

Every clue was checked against real logs before it went into the report, then sorted into six groups so the next person can search for it.

  • NetworkAddresses and ports seen in traffic — attacker, pivot and victim IPs; C2 4444; admin 22; SMB 445; RPC 135; Kerberos 88; LDAP 389; LLMNR/NetBIOS
  • Host & filesystemFiles left on disk — web shell, named pipe, staged hash dumps, hacked.bmp, ransom note, dummy .enc files
  • RegistryWindows settings that were changed — Wallpaper / WallpaperStyle tampering on the DC and workstation
  • IdentityAccounts involved — admin, dvwa, www-data, root, win10, HANK_SHERMAN, Administrator, backdoor, krbtgt
  • Hashes & secretsPasswords and keys stolen — MD5, NetNTLMv2, Kerberoast TGS, krbtgt NTLM, a Golden Ticket, an injected ED25519 key
  • Tooling & commandsWhat the attacker ran — Nmap, sqlmap, Netcat, Responder, Impacket (SPNs, secretsdump, wmiexec, ticketer), Hashcat, reg.exe

THE DIAGRAMS

Eight figures from the lab

Nothing here is decoration. Each figure was built from the lab’s own config files, rules, logs and case notes. Click any one to open it full size.

Every figure above comes from the lab itself. The full write-up is at soc-lab.eclipsemanic.me · the source code is at github.com/EclipseManic/SOC-Lab.

SECTION 04 / SKILLS & TOOLS

The toolkit

Filter by role
Order
Depth Pro Working Learning Hover a tool for how it is used. Pick a role above to narrow the list.

140 tools across 15 categories. The full inventory, with context, is in the resume.

SECTION 05 / MITRE ATT&CK

Mapped to the matrix

Detection coverage mapped across 14 MITRE ATT&CK tactics from lab telemetry and published investigations.

Detection coverage across 14 MITRE ATT&CK tactics, full kill-chain — open on desktop for the interactive matrix.

SECTION 06 / ORIGINALS

Security Originals

Five tools and investigation projects. Open any title for its full case file, methods, evidence, and technical details.

PROJECT 01 / 05
MALWARE TRIAGE

Malware
Analyzer

Malware triage tool scanning PE, ELF, Mach-O, APK, Office, PDF, scripts, and archives via a Flask/WebSocket dashboard. 2,778 YARA rules from 40 sources. Connects VirusTotal, Hybrid Analysis, Tria.ge, and FileScan.io with live scan progress. Drop a file — or a whole folder — and the queue triages everything in one pass, streaming verdicts to the browser as each engine reports back. Verdicts arrive with matched rules and engine reports side by side, so one screen answers the first triage question. Every scan leaves a shareable record you can revisit later.

Challenge: Manual triage across many file formats is slow — analysts jump between separate tools for each type.
  • Multi-format static analysis: PE, ELF, Mach-O, APK, Office, PDF, scripts, archives
  • 2,778 YARA rules aggregated from 40 intel sources
  • Live WebSocket scan queue with per-file verdicts
  • Multi-engine reputation lookup: VirusTotal, Hybrid Analysis, Tria.ge, FileScan.io
YARA Flask WebSocket VirusTotal
GitHub ↗
PROJECT 02 / 05
T1566 PHISHING

Phishing Mail
Detector

11-step Flask tool validating SPF, DKIM, DMARC, and ARC to catch spoofed emails. Brand-check system covering ~50 commonly faked brands plus QR-code (quishing) detection. ~40-signal risk-scoring model mapped to MITRE ATT&CK T1566. Paste any raw email or .eml and get a full header breakdown, an impersonation verdict, and plain-English reasoning for every signal — evidence an analyst can act on, not a black-box score. Results are structured for handoff — each failed check links straight to the header evidence behind it.

Challenge: Spoofed and brand-impersonating emails pass casual inspection; authentication results are buried in headers most users never read.
  • 11-step header validation: SPF, DKIM, DMARC, ARC chain
  • Impersonation checks against ~50 commonly faked brands
  • QR-code (quishing) detection inside message bodies
  • ~40-signal scoring model mapped to MITRE ATT&CK T1566
SPF/DKIM/DMARC T1566 Flask QR Detection
GitHub ↗
PROJECT 03 / 05
EVTX FORENSICS

Event Analyzer
Web

Fast, analyst-first Windows EVTX viewer with built-in IOC extraction, Sigma rule matching, and MITRE ATT&CK mapping. Upload one or many .evtx files, filter by event ID/channel/provider, bookmark events, take notes, and export results. Drag in a full event dump and pivot instantly — jump to matched Sigma hits, correlate extracted IOCs, and build the case notes as you go. One search box covers events, IOCs, and notes, so the pivot from timeline to evidence stays under a minute. Results export cleanly for case notes and reports.

Challenge: Windows EVTX files are cumbersome to parse manually; analysts need a fast viewer with built-in intel enrichment in one interface.
  • Parallel multi-file .evtx ingestion and indexing
  • IOC extraction with hash, domain, and IP correlation
  • Sigma rule matching with MITRE ATT&CK technique mapping
  • Bookmarks, analyst notes, and exportable case output
EVTX Parsing Sigma Rules IOC Extraction Vue 3
GitHub ↗
PROJECT 04 / 05
DFIR REPORTS

Investigation
Portfolio

Real phishing, malware, and network attack investigations with detailed reports — IOC extraction and MITRE ATT&CK mapping. Full case log in Field Reports below. Seven published write-ups so far. Each report follows the same SOC cadence — scope, timeline, evidence, IoCs, detection gaps, and recommendations — so every conclusion is reviewable end to end. The library keeps growing as each new lab case gets written up in the same format. Every report is written to be reviewed end to end, with the full artifact trail attached.

Challenge: Lab skills are invisible without documented, reviewable investigations written the way a real SOC writes them.
  • Seven published investigation write-ups and counting
  • IOC extraction with hash, domain, and IP correlation
  • MITRE ATT&CK technique mapping per case
  • Detection-gap analysis and remediation guidance
IOC Extraction ATT&CK Mapping Wazuh Suricata
GitHub ↗
PROJECT 05 / 05
THREAT INTELLIGENCE

Threat
Sentry

Real-time threat intelligence aggregator pulling from multiple open-source feeds — URLhaus, ThreatFox, AlienVault OTX, and AbuseIPDB. Correlates indicators, surfaces attacker infrastructure, and maps IoCs to MITRE ATT&CK techniques. Indicators are normalized, deduplicated, and confidence-scored, so campaign infrastructure stands out from feed noise instead of drowning in it. Related indicators are grouped into campaigns, turning the raw feed into attacker storylines. The dashboard surfaces the infrastructure behind each campaign.

Challenge: Threat intel feeds generate massive noise. Analysts need correlation, dedup, and ATT&CK mapping — not a raw feed reader.
  • Live ingestion from URLhaus, ThreatFox, OTX, and AbuseIPDB
  • Cross-feed normalization and deduplication
  • Confidence scoring with attacker-infrastructure grouping
  • MITRE ATT&CK technique mapping on correlated indicators
Threat Intelligence IOC Correlation ATT&CK Mapping Flask
GitHub ↗
01 / 05 PROJECTS

SECTION 07 / FIELD REPORTS

Case files

Real investigations and CTF walkthroughs — from phishing campaigns to RAT analysis, documented step by step.

CASE-2026-001

Job Phishing Campaign — Impersonating Naukri.com

Social engineering · Advance-fee job scam

Read report →
CASE-2026-002

Storm-2949: How One Cloud Account Led to a Full Enterprise Breach

Cloud compromise

Read report →
CASE-2026-003

AI Helpdesk Compromise Leading to Workstation Breach (HackTheBox)

Artifact triage

Read report →
CASE-2026-004

Yellow Cockatoo RAT “Jupyter’s Bro” (CyberDefenders)

RAT analysis

Read report →
CASE-2026-005

PoisonedCredentials (CyberDefenders)

LLMNR/NBT-NS poisoning · NTLMv2 credential capture

Read report →
CASE-2026-006

Stealc Infostealer (Oski Lab) (CyberDefenders)

Sandbox/C2 analysis

Read report →
CASE-2026-007

WebStrike (CyberDefenders)

Web attack investigation

Read report →
All write-ups on Medium

SECTION 07A / LIVE ACTIVITY

Pulled live

Fetched client-side from the GitHub API and Medium RSS at page load — not hand-written cards. If a network fetch fails, the block below says so instead of faking it.

GitHub — recent activity

FETCHING github.com/EclipseManic …

Medium — latest write-ups

FETCHING eclipsemanic.medium.com …

SECTION 08 / CREDENTIALS

Verified credentials

Featured credentials first — scroll the posters to explore each one, then open its verification link. The full list below is grouped by issuer, and every entry links to a public page the issuer controls.

01Professional Certifications

02Analyst & Offensive Pathways

03E&ICT Academy, IIT Kanpur

Joint initiative of MeitY and IIT Kanpur.

04AI Security University

7.5 training hours across four completion certificates.

05Microsoft Learn

06Job Simulations

Forage employer simulations — practical task simulations, not graded certificates. Verification code XD9YMde4GgFmgr8yu.

Education

Bachelor of Computer Application — Cloud & Cybersecurity

IIMT University, Meerut

2024 — 2027
IN PROGRESS

Focus areas

Cloud Security SOC Operations DFIR Threat Detection

Verification standard

Every credential on this page links to a public issuer verification page. A credential is only listed when the issuer, the issue date, and the skills it supports can all be checked independently — nothing here rests on my own description of it.

  • Professional certificates verify on Coursera and the issuing academy’s own credential service.
  • NPTEL certificates verify by roll number on the NPTEL archive.
  • AI Security University credentials verify on a public, per-certificate page.
  • Job simulations are listed separately and labelled as simulations, not certificates.

THE COMPLETE PROFILE · RESUME

The complete profile

Education · practice · projects · skills
Tailor the profile Security Analyst · SOC Operations

MANICHAND GUPTAMEERUT, UP · OPEN TO RELOCATION

Professional Summary

Security analyst covering the full defense workflow — detection engineering, alert triage, endpoint investigation, and incident reporting. Built and operated a seven-node hybrid-cloud SOC lab (Wazuh, Suricata, Sysmon, Auditd, TheHive, Velociraptor) across VMware and Azure, where a 20-stage, 74-minute attack campaign was caught by 148 custom Wazuh rules and 11 Suricata signatures — 16 of 20 stages detected — and 9,567 raw events were compressed into 46 alerts and three closed cases. Independently investigates fraud and malware campaigns and builds open-source DFIR tools. Ranked World #4,017 / India #229 on CyberDefenders and World #836 of 174,154 (under 0.5%) on KC7.

Relevant Experience

Personal SOC & Incident Response LabSelf-directed security operations practice
  • Built a seven-node hybrid SOC across VMware Workstation and Microsoft Azure, joining a dual-homed Suricata gateway and an isolated Active Directory victim subnet to cloud Wazuh, TheHive, and Velociraptor over a Zero-NAT WireGuard tunnel.
  • Authored 148 custom Wazuh detection rules and 87 false-positive suppressions; built a Python SOAR bridge (SQLite WAL buffer) that reduced 9,567 raw alerts to 46 actionable alerts and three closed cases.
  • Ran a 20-stage, 74-minute red-team campaign and executed containment: severed reverse shells, quarantined webshells, revoked SSH keys, and invalidated forged Kerberos tickets via a double krbtgt reset (KVNO 3).
  • Investigated hosts with Velociraptor; recorded evidence, indicators, response actions, and MITRE ATT&CK mappings in TheHive 5.7.5.

Selected Projects

Malware Analyzer – Web ↗Scans PE, ELF, Mach-O, APK, Office, PDF, scripts, and archives; 2,778 YARA rules from 40 sources; VirusTotal, Hybrid Analysis, Tria.ge, and FileScan.io; 75 automated tests.
Phishing Mail Detector ↗SPF, DKIM, DMARC, and ARC checks; about 50 impersonated brands; QR scam detection; roughly 40 risk signals; VirusTotal and urlscan.io lookups; ATT&CK T1566.
Investigation Portfolio ↗Seven cases, including a live internship-fraud campaign analyzed through company and domain research plus PDF malware reverse engineering, and a multi-vector job scam; five additional CyberDefenders and HackTheBox reports.
ThreatSentry AI ↗PyQt5 threat-hunting dashboard featured on Kitploit: Shodan asset discovery, NVD CVE correlation across 14+ product types, and an ensemble risk model with scheduled retraining.

Education

Bachelor of Computer Application – Cloud & CybersecurityIIMT University, Meerut · 7.5 CGPA through 4th semester

Open to SOC / DFIR internships

Manichand Gupta

Manichand Gupta

Security Analyst

I investigate what others miss — every alert triaged, every log correlated, every threat mapped to how it actually happened.

World #836Verify ↗KC7KC7 · of 174,154 players
Investigations56Hands-on labs CyberDefenders ↗HackTheBox ↗
CyberDefenders#4,017World rank · India #229
Threat Intel2,778YARA rules
ATT&CK20Attack stages
Research7Write-ups published

Background

Analyst behind the glass

Security analyst based in Meerut, UP, finishing a BCA in Cloud & Cybersecurity (2024–2027) at IIMT University. 36 CyberDefenders and 20 HackTheBox Sherlock investigations, a seven-node SOC lab, seven published case files, and open-source detection and analysis tooling.

Jul 2026 — Oct 2026

Self-Built SOC Detection & Monitoring Lab

Independent

  • Monitored Wazuh and Suricata alerts across a seven-node hybrid lab spanning VMware and Azure
  • Triaged a 9,567-alert feed down to 46 alerts across three incident cases using 148 rules
  • Contained a live intrusion — severed reverse shells, quarantined webshells, reset krbtgt
  • Documented IOCs, root causes and ATT&CK mappings, then hardened the root-cause flaws

2025 — Present

Independent Investigations & Security Tooling

Self-directed

  • 36 CyberDefenders and 20 HackTheBox Sherlock investigations across web, AD, memory, disk and log forensics
  • Seven ATT&CK-mapped case files published, including Stealc infostealer and a live job-phishing campaign
  • RuleForge (8 SIEM dialects), IOC platform (33 sources), malware analyzer (2,778 YARA rules)
  • KC7 World #836 of 174,154 · CyberDefenders World #4,017 / India #229

Education

BCA — Cloud & Cybersecurity IIMT University, Meerut 2024 — 2027 · in progress

Inside the lab

A fake company, built to be attacked

Seven machines split between my computer and Azure: a firewall, a Windows domain with real accounts, a website to break into, and the tools that catch and record an attack. A private tunnel joins the two halves. I built it to run a real intrusion, catch it, and investigate it end to end.

THE ENVIRONMENT

Four pillars, mapped to the sensors

DETECTION SCORECARD

An honest score: 14 / 2 / 2 / 2

14Caught directly70.0%
2Only from another machine10.0%
2Nothing was logging it10.0%
2Happened off the network10.0%
16 / 20Covered in practice · 80%No rule failed

9,567 raw log events → 46 alerts → 3 cases

ALERT PIPELINE

The alert pipeline I rebuilt

My first pipeline used Shuffle and collapsed under a flood of alerts. The replacement saves each one locally in under 0.2 ms, wakes every 5 s to group them, and still lands at 4.6 s typical with zero alerts lost.

< 0.2 msto save one alert
5 show often it wakes
4.6 salert to case, typical
0alerts lost
18.9 msto store 500 alerts
17 / 17self-tests passing

THE CAMPAIGN

74 min · 20 steps · 3 cases

CAS-001-WEBCOMP01–07

Breaking in through the website, then becoming root

Nmap scanned for ports. Hydra guessed a password. SQL injection dumped the user table. A PHP web shell went up, then a Netcat reverse shell. A sudo mistake let it become root, and an SSH key was planted to come back.

Resolved · Closed

CAS-002-LATMOVE08–14

Moving inward, stealing credentials, climbing the domain

Internal scan, then Responder poisoned Windows name lookups to capture hashes. Hashcat cracked them offline. BloodHound mapped the domain, Kerberoasting took a service ticket, and DCSync copied the password database.

Resolved · Closed

CAS-003-DOMIMP15–20

Taking the domain, then causing damage

Pass-the-Hash reused a stolen hash. A forged Golden Ticket let the attacker sign in as any account without a password, and a second Domain Admin was created. Files were copied out over SCP, the wallpaper replaced, and fake ransomware forced a reboot.

Resolved · Closed

RESPONSE

Stop the bleeding, then fix the cause

7 containment actions

  1. Blocked the attacker’s address at the gateway firewall
  2. Killed the reverse shell and quarantined shell.php
  3. Deleted the rogue root SSH key and closed its sessions
  4. Disabled backdoor and removed it from Domain Admins
  5. Reset krbtgt twice to KVNO 3
  6. Quarantined the staged files and restored the wallpaper
  7. Changed the DVWA / MariaDB / app passwords

7 root causes erased

  1. Removed the sudo rule that ran python3 as root without a password
  2. Turned off root SSH login and password login
  3. Blocked scripts from running in the uploads folder
  4. Demoted the service account, deleted its SPN, forced stronger encryption
  5. Moved privileged accounts into Protected Users
  6. Disabled LLMNR and NetBIOS (old Windows name lookup) on every machine
  7. Split the flat network into firewall zones that allow only the needed traffic

EVIDENCE

The evidence register

  • NetworkAddresses and ports seen in traffic — attacker/pivot/victim IPs; C2 4444; SMB 445; Kerberos 88; LDAP 389; LLMNR/NetBIOS
  • HostFiles left on disk — web shell, named pipe, hash dumps, hacked.bmp, ransom note, dummy .enc files
  • RegistryWindows settings changed — Wallpaper tampering on the DC and workstation
  • IdentityAccounts involved — admin, dvwa, www-data, root, win10, HANK_SHERMAN, Administrator, backdoor, krbtgt
  • HashesPasswords and keys stolen — MD5, NetNTLMv2, Kerberoast TGS, krbtgt NTLM, a Golden Ticket, an ED25519 key
  • ToolingWhat the attacker ran — Nmap, sqlmap, Netcat, Responder, Impacket, Hashcat, reg.exe

THE DIAGRAMS

Eight figures from the lab

Each figure was built from the lab’s own config, rules, logs and case notes. Tap any one to open it full size.

Every figure above comes from the lab itself and sits next to the section it explains. Full write-up at soc-lab.eclipsemanic.me.

Selected work

Projects I Built

Credentials

Certifications

20 credentials in total — every one links to a public issuer verification page. Open the desktop page for the full list grouped by issuer.

Field reports

Case files

  1. CASE-2026-001 Job Phishing Campaign — Impersonating Naukri.com Social engineeringAdvance-fee job scam
  2. CASE-2026-002 Storm-2949: How One Cloud Account Led to a Full Enterprise Breach Cloud compromise
  3. CASE-2026-003 AI Helpdesk Compromise Leading to Workstation Breach Artifact triageHackTheBox
  4. CASE-2026-004 Yellow Cockatoo RAT “Jupyter’s Bro” RAT analysisCyberDefenders
  5. CASE-2026-005 PoisonedCredentials LLMNR/NBT-NS poisoningCyberDefenders
  6. CASE-2026-006 Stealc Infostealer — Oski Lab Sandbox / C2 analysisCyberDefenders
  7. CASE-2026-007 WebStrike Web attack investigationCyberDefenders
All write-ups on Medium ↗

Contact

Let’s investigate together

Open to SOC / DFIR internships and junior analyst roles. If you have alerts, I have eyes.

manichand.gupta.contact@gmail.com

Meerut, UP · © 2026 Manichand Gupta