incidents

deep technical investigations and analysis

INC-2025-040

Memory Forensic Investigation: Credential Theft

High

severity

High

type

Forensics

status

Mitigated

hash

N/A

incident summary

credential theft from corporate laptop via lsass memory dump. attacker gained initial access via phishing email with weaponized excel document. lateral movement observed over 5 days before detection.

forensic timeline

  • 2025-04-10 09:23 - malicious excel document opened (user interaction)
  • 2025-04-10 09:24 - powershell process spawned from excel
  • 2025-04-10 09:25 - cobalt strike beacon established
  • 2025-04-10 14:12 - lsass.exe accessed by suspicious process
  • 2025-04-10 14:13 - memory dump created in temp folder
  • 2025-04-11 02:00 - credentials exfiltrated (23mb upload)
  • 2025-04-12 to 2025-04-14 - lateral movement across 3 hosts
  • 2025-04-15 11:45 - edr alert triggered, investigation began

artifacts recovered

  • - memory dump file: 20250410_1413_lsass.dmp (768mb)
  • - prefetch files showing malicious process execution
  • - registry run keys modified for persistence
  • - amcache showing application execution history
  • - mft entries showing file creation/deletion patterns

key findings

  • - 44 unique credentials extracted from lsass
  • - credential theft targeted service accounts, not users
  • - attacker used mimikatz with custom obfuscation
  • - no disk encryption detected on stolen credentials

remediation actions

  • - all 44 service accounts credentials rotated
  • - affected hosts reimaged and hardened
  • - edr policies updated to block lsass access patterns
  • - user training on phishing awareness conducted
INC-2025-041

Ransomware Variant Reverse Engineering

High

severity

High

type

Malware

status

Confirmed

hash

e8b3f2a1...

incident summary

rust-based ransomware with custom encryption and anti-debugging features. analysis reveals c2 infrastructure and detection signatures.

malware characteristics

  • - rust-based encryptor with custom cha20 implementation
  • - double-extortion: encrypts data and exfiltrates before notification
  • - uses windows api for process hollowing injection
  • - encrypted c2 over https with custom protocol
  • - anti-debugging: checks for debugger presence via ntqueryinformationprocess

encryption workflow

  • 1. enumerate network shares via netshareenum
  • 2. generate 256-bit key per victim (ecdh on victim id)
  • 3. encrypt files with custom cha20-chacha variant
  • 4. create ransom note with payment instructions
  • 5. send victim id + encryption proof to c2

c2 infrastructure

  • primary c2: 185.xxx.xxx.44 (domain parked, now offline)
  • backup c2: onion service active (tracked)
  • traffic pattern: keep-alive every 30 seconds, encrypted payload

detection signatures

  • yara rule: win_ransom_rust_variant
  • strings:
  • $rust_string = "core::str::slice_error_fail"
  • $custom_encrypt = {e8 3c 4a 7b 9f 21 d4}
  • $c2_pattern = /https?:\\\\//\\\\[a-z0-9]{32}\\\\.onion/
  • condition: 2 of them
INC-2025-042

Heap Overflow Analysis in libpng 1.6.40

Critical

severity

Critical

type

Memory Corruption

status

Analyzing

cve

CVE-2025-16847

incident summary

heap-based buffer overflow in png_handle_iTXt() function when processing malformed iTXt chunks. attacker-controlled data can overwrite adjacent heap metadata, leading to arbitrary code execution when png_free() is called.

technical details

  • offset calculation error at line 2347:
  • text_len = png_get_uint_16(buf);
  • lang_len = png_get_uint_16(buf + 2);
  • key_len = png_get_uint_16(buf + 4);
  • // unchecked allocation
  • text = png_malloc(png_ptr, text_len); // vulnerable
  • // bounds check fails when lang_len + key_len > text_len

exploitation requirements

  • - target must process attacker-supplied png file
  • - heap grooming required for reliable control
  • - aslr and heap mitigations present on modern systems
  • - confirmed exploitable on ubuntu 22.04, fedora 38

mitigation

upgrade to libpng 1.6.41 or later. apply patch from upstream commit 8a3d2c1. validate all png inputs through sanitization proxy before processing.