Improve results scanning for Linux malware (chainguard-dev#608)

* Improve coverage for timb-machine

* address more Ubuntu high false-positive

* further rule tuning

* further rule tuning

* Renamed rules

* rule tuning

pkg/action/testdata/scan_archive

@@ -1,6 +1,6 @@
 # testdata/apko_nested.tar.gz ∴ /apko_0.13.2_linux_arm64/apko: medium
 c2/addr/ip: medium
-c2/server_address: medium
+c2/addr/server: medium
 collect/archives/zip: medium
 credential/keychain: medium
 credential/password: low
@@ -25,14 +25,15 @@ discover/network/interface_list: medium
 discover/network/mac_address: medium
 discover/network/netstat: medium
 discover/processes/pgrep: medium
-discover/system/cpu_info: low
-discover/system/hostname_get: low
+discover/system/cpu: low
+discover/system/dmesg: low
+discover/system/hostname: low
 discover/system/platform: low
 discover/user/HOME: low
 discover/user/USER: low
 discover/user/name_get: medium
 evasion/bypass_security/linux/se: medium
-evasion/hidden_paths/hidden: medium
+evasion/file/prefix: medium
 evasion/hide_artifacts/pivot_root: medium
 exec/plugin: low
 exec/program: medium

rules/anti-static/packer/elf.yara → rules/anti-static/elf/content.yara

@@ -1,4 +1,4 @@
-import "math"
+import "elf"
 
 rule obfuscated_elf: high linux {
   meta:
@@ -27,21 +27,5 @@ rule obfuscated_elf: high linux {
     $debuglink         = ".gnu_debuglink" fullword
 
   condition:
-    uint32(0) == 1179403647 and none of them
-}
-
-rule high_entropy_header: high {
-  meta:
-    description                                                       = "high entropy ELF header (>7)"
-    hash_2023_UPX_0c25                                                = "0c25a05bdddc144fbf1ffa29372481b50ec6464592fdfb7dec95d9e1c6101d0d"
-    hash_2023_UPX_5a59                                                = "5a5960ccd31bba5d47d46599e4f10e455b74f45dad6bc291ae448cef8d1b0a59"
-    hash_2023_FontOnLake_38B09D690FAFE81E964CBD45EC7CF20DCB296B4D_elf = "f155fafa36d1094433045633741df98bbbc1153997b3577c3fa337cc525713c0"
-
-  strings:
-    $not_pyinst = "pyi-bootloader-ignore-signals"
-    $not_go     = "syscall_linux.go"
-    $not_go2    = "vdso_linux.go"
-
-  condition:
-    uint32(0) == 1179403647 and math.entropy(1200, 4096) > 7 and none of ($not*)
+    filesize > 512 and elf.type == elf.ET_EXEC and uint32(0) == 1179403647 and none of them
 }

rules/anti-static/elf/entropy.yara

@@ -0,0 +1,30 @@
+import "math"
+
+private rule normal_elf {
+  condition:
+    filesize < 64MB and uint32(0) == 1179403647
+}
+
+private rule small_elf {
+  condition:
+    filesize < 400KB and uint32(0) == 1179403647
+}
+
+rule normal_elf_high_entropy_7: medium {
+  meta:
+    description = "higher entropy ELF binary (>7.1)"
+
+  condition:
+    normal_elf and math.entropy(1, filesize) >= 7.1
+}
+
+rule normal_elf_high_entropy_7_4: high {
+  meta:
+    description = "high entropy ELF binary (>7.4)"
+
+  strings:
+    $not_whirlpool = "libgcrypt-grub/cipher/whirlpool.c"
+
+  condition:
+    normal_elf and math.entropy(1, filesize) >= 7.4 and none of ($not*)
+}

rules/anti-static/elf/header.yara

@@ -1,4 +1,5 @@
 import "elf"
+import "math"
 
 rule single_load_rwe: critical {
   meta:
@@ -38,3 +39,20 @@ rule fake_dynamic_symbols: critical {
   condition:
     elf.type == elf.ET_EXEC and elf.entry_point < filesize and elf.number_of_sections > 0 and elf.dynamic_section_entries > 0 and for any i in (0..elf.dynamic_section_entries): (elf.dynamic[i].type == elf.DT_SYMTAB and not (for any j in (0..elf.number_of_sections): (elf.sections[j].type == elf.SHT_DYNSYM and for any k in (0..elf.number_of_segments): ((elf.segments[k].virtual_address <= elf.dynamic[i].val) and ((elf.segments[k].virtual_address + elf.segments[k].file_size) >= elf.dynamic[i].val) and (elf.segments[k].offset + (elf.dynamic[i].val - elf.segments[k].virtual_address)) == elf.sections[j].offset))))
 }
+
+rule high_entropy_header: high {
+  meta:
+    description                                                       = "high entropy ELF header (>7)"
+    hash_2023_UPX_0c25                                                = "0c25a05bdddc144fbf1ffa29372481b50ec6464592fdfb7dec95d9e1c6101d0d"
+    hash_2023_UPX_5a59                                                = "5a5960ccd31bba5d47d46599e4f10e455b74f45dad6bc291ae448cef8d1b0a59"
+    hash_2023_FontOnLake_38B09D690FAFE81E964CBD45EC7CF20DCB296B4D_elf = "f155fafa36d1094433045633741df98bbbc1153997b3577c3fa337cc525713c0"
+
+  strings:
+    $not_pyinst = "pyi-bootloader-ignore-signals"
+    $not_go     = "syscall_linux.go"
+    $not_go2    = "vdso_linux.go"
+    $not_module = ".module_license" fullword
+
+  condition:
+    uint32(0) == 1179403647 and elf.type == elf.ET_EXEC and math.entropy(1200, 4096) > 7 and none of ($not*)
+}

rules/anti-static/elf/tiny.yara

@@ -0,0 +1,12 @@
+import "elf"
+
+rule impossibly_small_elf_program: high {
+  meta:
+    description = "ELF binary is unusually small"
+
+  strings:
+    $not_hello_c = "hello.c"
+
+  condition:
+    filesize < 8192 and filesize > 900 and uint32(0) == 1179403647 and elf.type == elf.ET_EXEC and none of ($not*)
+}

rules/anti-static/macho/entropy.yara

@@ -0,0 +1,26 @@
+import "math"
+
+private rule smaller_macho {
+  condition:
+    filesize < 64MB and (uint32(0) == 4277009102 or uint32(0) == 3472551422 or uint32(0) == 4277009103 or uint32(0) == 3489328638 or uint32(0) == 3405691582 or uint32(0) == 3199925962)
+}
+
+rule high_entropy_7_2: medium {
+  meta:
+    description = "higher entropy binary (>7.2)"
+
+  condition:
+    smaller_macho and math.entropy(1, filesize) >= 7.2
+}
+
+rule high_entropy_7_9: high {
+  meta:
+    description = "high entropy binary (>7.9)"
+
+  strings:
+    // prevent bazel false positive
+    $bin_java = "bin/java"
+
+  condition:
+    smaller_macho and math.entropy(1, filesize) >= 7.9 and not $bin_java
+}

rules/anti-static/binary/tiny.yara → rules/anti-static/macho/tiny.yara

@@ -1,13 +1,3 @@
-import "elf"
-
-rule impossibly_small_elf_program: high {
-  meta:
-    description = "ELF binary is unusually small"
-
-  condition:
-    filesize < 8192 and uint32(0) == 1179403647 and elf.type == elf.ET_EXEC
-}
-
 rule impossibly_small_macho_program: medium {
   meta:
     description = "machO binary is unusually small"

rules/anti-static/obfuscation/bitwise.yara

@@ -22,14 +22,22 @@ rule excessive_bitwise_math: high {
     hash_2023_aiohttpp_0_1_setup  = "cfa4137756f7e8243e7c7edc7cb0b431a2f4c9fa401f2570f1b960dbc86ca7c6"
 
   strings:
-    $x           = /\-{0,1}[\da-z]{1,8} \<\< \-{0,1}\d{1,8}/
-    $not_Sodium  = "Sodium_Core"
-    $not_SHA512  = "SHA512"
-    $not_SHA256  = "SHA256"
-    $not_MD4     = "MD4"
-    $not_algbase = "algbase" fullword
-    $not_jslint  = "jslint bitwise"
-    $not_include = "#define "
+    $x                  = /\-{0,1}[\da-z]{1,8} \<\< \-{0,1}\d{1,8}/
+    $not_Sodium         = "Sodium_Core"
+    $not_SHA512         = "SHA512"
+    $not_SHA256         = "SHA256"
+    $not_MD4            = "MD4"
+    $not_algbase        = "algbase" fullword
+    $not_jslint         = "jslint bitwise"
+    $not_include        = "#define "
+    $not_bitwise        = "bitwise" fullword
+    $not_bitmasks       = "bitmasks" fullword
+    $not_ckbcomp        = "ckbcomp" fullword
+    $not_bit_test       = "bits_test" fullword
+    $not_testing        = "*testing.T"
+    $not_effective_bits = "effective bits"
+    $not_bit_offsets    = "bit offsets"
+    $not_uuid           = "uuid" fullword
 
   condition:
     filesize < 192KB and #x > 64 and none of ($not*)

rules/anti-static/obfuscation/hex.yara

@@ -51,13 +51,14 @@ rule hex_parse_base64_high: high {
     hash_1985_package_index      = "8d4daa082c46bfdef3d85a6b5e29a53ae4f45197028452de38b729d76d3714d1"
 
   strings:
-    $lang_node      = /Buffer\.from\(\w{0,16}, {0,2}'hex'\)/
-    $lang_python    = /\.unhexlify\(/
-    $b_base64       = "base64"
-    $b_b64decode    = "b64decode"
-    $not_sha256     = "sha256" fullword
-    $not_sha512     = "sha512" fullword
-    $not_algorithms = "algorithms" fullword
+    $lang_node         = /Buffer\.from\(\w{0,16}, {0,2}'hex'\)/
+    $lang_python       = /\.unhexlify\(/
+    $b_base64          = "base64"
+    $b_b64decode       = "b64decode"
+    $not_sha256        = "sha256" fullword
+    $not_sha512        = "sha512" fullword
+    $not_algorithms    = "algorithms" fullword
+    $not_python_base64 = "return binascii.unhexlify(s)"
 
   condition:
     filesize < 32KB and any of ($lang*) and any of ($b*) and none of ($not*)

rules/anti-static/obfuscation/python.yara

@@ -220,8 +220,11 @@ rule fernet_base64: high {
     $o6         = "exec("
     $o7         = "eval("
 
+    $not_utils         = "from cryptography import utils"
+    $not_fernet_itself = "class Fernet"
+
   condition:
-    filesize < 2MB and any of ($fernet*) and any of ($bdecode*) and any of ($o*)
+    filesize < 2MB and any of ($fernet*) and any of ($bdecode*) and any of ($o*) and none of ($not*)
 }
 
 rule python_long_hex: medium {
@@ -261,8 +264,10 @@ rule python_hex_decimal: high {
 
     $trash = /\\x{0,1}\d{1,3}\\/
 
+    $not_testing_t = "*testing.T" fullword
+
   condition:
-    filesize < 1MB and any of ($f*) and #trash in (filesize - 1024..filesize) > 100
+    filesize < 1MB and any of ($f*) and #trash in (filesize - 1024..filesize) > 100 and none of ($not*)
 }
 
 rule dumb_int_compares: high {

rules/anti-static/packer/cx_freeze.yara

@@ -3,8 +3,9 @@ rule cxFreeze_Python_executable: high {
     hash_2023_MacStealer_weed = "6a4f8b65a568a779801b72bce215036bea298e2c08ec54906bb3ebbe5c16c712"
 
   strings:
-    $cxfreeze = "cx_Freeze"
+    $cxfreeze      = "cx_Freeze"
+    $not_importlib = "tool like cx_Freeze"
 
   condition:
-    filesize < 10485760 and $cxfreeze
+    filesize < 10485760 and $cxfreeze and none of ($not*)
 }

rules/c2/addr/ip.yara

@@ -30,15 +30,17 @@ rule elf_hardcoded_ip: high {
 
   strings:
     // stricter version of what's above: excludes 255.* and *.0.* *.1.*, and 8.* (likely Google)
-    $sus_ipv4           = /((25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2345679])\.){3}(25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])/ fullword
-    $not_version        = /((25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])\.){3}(25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])[\.\-]/
-    $not_incr           = "10.11.12.13"
-    $not_169            = "169.254.169.254"
-    $not_spyder         = "/search/spider"
-    $not_ruby           = "210.251.121.214"
-    $not_1_2_3_4        = "1.2.3.4"
-    $not_root_servers_h = "128.63.2.53"
-    $not_root_servers_i = "192.36.148.17"
+    $sus_ipv4              = /((25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2345679])\.){3}(25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])/ fullword
+    $not_version           = /((25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])\.){3}(25[0-4]|2[0-4][0-9]|1[0-9][0-9]|[1-9][0-9]|[2-9])[\.\-]/
+    $not_incr              = "10.11.12.13"
+    $not_169               = "169.254.169.254"
+    $not_spyder            = "/search/spider"
+    $not_ruby              = "210.251.121.214"
+    $not_1_2_3_4           = "1.2.3.4"
+    $not_root_servers_h    = "128.63.2.53"
+    $not_root_servers_i    = "192.36.148.17"
+    $not_123456789         = "123.45.67.89"
+    $not_libebt_among_init = "libebt_among_init"
 
   condition:
     filesize < 12MB and uint32(0) == 1179403647 and 1 of ($sus_ip*) and none of ($not*)
@@ -73,14 +75,17 @@ rule hardcoded_ip_port: high {
 
   strings:
     $ipv4            = /([1-9][0-9]{1,2}\.){3}[1-9][0-9]{1,2}:\d{2,5}/ fullword
-    $not_ssdp        = "239.255.255.250:1900"
+    $not_ssdp        = "239.255.255.250:"
     $not_2181        = "10.101.203.230:2181"
-    $not_meta        = "169.254.169.254:80"
-    $not_vnc         = "10.10.10.10:5900"
+    $not_meta        = "169.254.169.254:"
+    $not_vnc         = "10.10.10.10:"
     $not_azure_pgsql = "20.66.25.58:5432"
     $not_wireguard   = "127.212.121.99:999"
     $not_minio       = "172.16.34.31:9000"
     $not_test        = "def test_" fullword
+    $not_12          = "12.12.12.12:"
+    $not_21          = "21.21.21.21:"
+    $not_255         = "255.255.255.255:"
 
   condition:
     any of ($ip*) and none of ($not*)

rules/c2/addr/server.yara

@@ -1,21 +1,19 @@
-rule server_addr: medium {
+rule server_address: medium {
   meta:
-    description = "may execute a shell and communicate with a server"
+    description                          = "references a 'server address', possible C2 client"
+    hash_2024_Downloads_3105             = "31054fb826b57c362cc0f0dbc8af15b22c029c6b9abeeee9ba8d752f3ee17d7d"
+    hash_2023_Linux_Malware_Samples_450a = "450a7e35f13b57e15c8f4ce1fa23025a7c313931a394c40bd9f3325b981eb8a8"
+    hash_2023_Linux_Malware_Samples_458e = "458e3e66eff090bc5768779d5388336c8619a744f486962f5dfbf436a524ee04"
 
   strings:
-    $serverAddr    = "serverAddr"
-    $server_addr   = "server_addr"
-    $exec          = "exec"
-    $sh            = "/bin/sh" fullword
-    $sh_bash       = "/bin/bash" fullword
-    $sh_zsh        = "/bin/zsh" fullword
-    $sh_script     = "ShellScript"
-    $sh_exec       = "ExecShell"
-    $sh_cmd        = "cmd.exe"
-    $sh_powershell = "powershell.exe"
+    $s_underscores = /\w{0,32}server_addr\w{0,32}/
+    $s_mixed       = /\w{0,32}serverAddr\w{0,32}/
+    $s_url         = "serverURL" fullword
+    $s_url2        = "serverUrl" fullword
+    $s_connect     = /\w{0,32}ConnectServer\w{0,32}/
 
   condition:
-    filesize < 10MB and any of ($server*) and $exec and any of ($sh*)
+    any of ($s*)
 }
 
 rule server_addr_small: high {
@@ -34,6 +32,8 @@ rule server_addr_small: high {
     $sh_cmd        = "cmd.exe"
     $sh_powershell = "powershell.exe"
 
+    $hash_bang = "#!"
+
   condition:
-    filesize < 128KB and any of ($server*) and $exec and any of ($sh*)
+    filesize < 1MB and any of ($server*) and $exec and any of ($sh*) and not $hash_bang in (0..3)
 }