Hunt-Benito

Classification basis

The artifact includes executable code (payload withheld) that actively calls the vulnerable functions with crafted malicious inputs to trigger the described vulnerabilities, which is the definition of an exploit. The payload withheld describes it as a 'PoC' and provides instructions to run it, confirming its purpose is to exercise the vulnerability.

Requirements

• Elixir >= 1.14 and internet connection to fetch the vulnerable html_sanitize_ex 1.5.2 dependency.README.md:24
• The vulnerable version of html_sanitize_ex (1.5.2) must be installed via mix deps.get.README.md:27

Observed behavior

• Generates a benign CSS payload and a malicious CSS payload (differing by a single ':') and calls HtmlSanitizeEx.html5/1 on both, measuring execution time to demonstrate the quadratic regex backtracking (CVE-2026-68749).poc.exs:18-25poc.exs:42-52
• Generates payloads of 2,000 and 20,000 sibling HTML elements and calls HtmlSanitizeEx.basic_html/1 on both, measuring execution time to demonstrate the quadratic traversal (CVE-2026-68750).poc.exs:31-33poc.exs:54-62
• Prints the measured timings and the ratio of attack time to benign time, showing the super-linear performance degradation.poc.exs:48-51poc.exs:58-61

Observables

File

Payload withheldThe main PoC script that generates payloads and measures execution time of the vulnerable library.poc.exs:1-76

File

Payload withheldDocumentation explaining the vulnerabilities, reproduction steps, and expected results.README.md:1-93

Classification basis

The artifact does not include or interact with the actual vulnerable TECNO Hi Browser APK. It provides a standalone server and a client that reproduces the vulnerable pattern to demonstrate and detect the path-traversal class, which is characteristic of a scanner.

Requirements

• A vulnerable browser or download manager that trusts the Content-Disposition filename without sanitization.README.md:5-10
• The malicious server (evil_server.py) must be reachable by the client.evil_server.py:17-18

Observed behavior

• evil_server.py serves an HTTP response with a Content-Disposition header containing '../' sequences in the filename.evil_server.py:39-44
• naive_downloader.py fetches the response, extracts the filename, and writes the body to a path constructed by joining the download directory with the unsanitized filename, causing the file to be written outside the intended directory.naive_downloader.py:52-54naive_downloader.py:97-101
• The script reports whether the file escaped the download directory, confirming the vulnerability class.naive_downloader.py:101-104

Observables

Payload

Payload withheldThe payload written by the PoC server is a benign marker string used to demonstrate the file-write primitive.evil_server.py:26

Network Behavior

Payload withheldThe server is the delivery mechanism for the path-traversal payload; it does not establish persistence, exfiltrate data, or accept commands.evil_server.py:34-55

File Write

Payload withheldThe naive_downloader.py script writes the payload to a path that may escape the download directory, demonstrating the vulnerability. This is the intended PoC behavior.naive_downloader.py:96-104

Classification basis

The primary artifact is classified as exploit because poc_dos.py actively exercises the vulnerability by sending a malicious payload that triggers the stack overflow and crashes the target service, which is confirmed by the code's behavior of re-probing to verify the crash. The README explicitly labels it as a 'Confirmed DoS reproducer' and the code includes logic to send the oversized parameter and confirm the resulting denial of service.

Requirements

• Network access to the Tenda W6-S management interface (default http://payload withheld)poc_dos.py:46-48
• Target device must be running vulnerable firmware v1.0.0.4(510)README.md:22-23

Observed behavior

• Sends an HTTP POST request to payload withheld with an oversized 'GO' parameter (default 2000 bytes of 'A') to overflow a 64-byte stack bufferpoc_dos.py:46-53
• Detects successful denial of service by re-probing the management interface and confirming httpd is no longer respondingpoc_dos.py:92-100
• Provides a conceptual RCE skeleton that constructs a payload to overwrite the saved return address ($ra) on the stack, demonstrating control-flow hijack shapeexploit_conceptual.py:92-104

Observables

Url

Payload withheldVulnerable endpoint targeted by the PoC scripts.poc_dos.py:30exploit_conceptual.py:35

Ip Address

Payload withheldDefault LAN IP of the Tenda W6-S device, used as example target in documentation and scripts.README.md:43poc_dos.py:67

Classification basis

The artifact's primary purpose is to detect, validate, and report the integer overflow vulnerability. It does not contain code that exercises the vulnerability to achieve heap corruption or code execution. The README explicitly states that overflow_demo.c performs no out-of-bounds write and craft_gguf.py produces a non-runnable model, confirming the artifact is a scanner/demonstrator, not an exploit.

Requirements

• Attacker must provide a crafted n_seq_max, n_tokens, or embd value through a malicious model file or JNI call.README.md:12-14
• The target must be a 32-bit ABI (armeabi-v7a) for heap corruption; 64-bit ABIs result in a large allocation failure (DoS).README.md:60-63

Observed behavior

• overflow_demo.c computes the malloc size using the same arithmetic as the vulnerable code and prints the wrapped result on 32-bit vs 64-bit size_t, without performing any allocation or write.overflow_demo.c:33-47
• craft_gguf.py generates a minimal GGUF file with an attacker-controlled llama.embedding_length metadata value, demonstrating the model-file delivery vector, but produces a non-runnable model with no tensors.craft_gguf.py:12-15
• hook_new_batch.js intercepts the vulnerable JNI function to log and optionally override the multiplicands, enabling observation of the overflow path on a live device without triggering the corruption.hook_new_batch.js:53-66

Observables

Vulnerability Demonstration

Payload withheldThe file only prints allocation sizes to prove the integer overflow; it does not trigger heap corruption.overflow_demo.c:8-11

Model File Generation

Payload withheldThe script demonstrates the model-file delivery vector for the vulnerability but produces a non-runnable model with no tensors.craft_gguf.py:12-15

Dynamic Instrumentation

Payload withheldThe script logs and optionally overrides function arguments for research purposes; it does not perform any harmful action.hook_new_batch.js:2-3

Classification basis

The artifact contains executable code (poc.php) that actively performs SQL injection attacks against a live database. It uses pg_select() and pg_insert() with malicious payloads to exfiltrate data and escalate privileges, which constitutes exploitation, not just detection or scanning.

Requirements

• A vulnerable PHP build (8.2 < 8.2.33, 8.3 < 8.3.33, 8.4 < 8.4.24, 8.5 < 8.5.9) with ext/pgsql enabled.README.md:27-32
• A PostgreSQL database (9.1+) with standard_conforming_strings = on.README.md:107
• Network access to the target PostgreSQL database.poc.php:16

Observed behavior

• Connects to a PostgreSQL database using credentials from environment or defaults.poc.php:16-17
• Creates a 'users' table and inserts two rows.poc.php:31-33
• Executes a pg_select() with a crafted payload to perform data exfiltration via SQL injection, returning all rows instead of a specific match.poc.php:39-48
• Executes a pg_insert() with a crafted payload to perform privilege escalation, forcing the 'admin' column to true.poc.php:60-70
• Prints the generated SQL statements to show the E'...' backslash breakout.poc.php:44poc.php:65

Observables

Network Connection

Payload withheldThe script connects to a PostgreSQL database specified by the PGCONN environment variable or a hardcoded default. This is necessary for the PoC to demonstrate the SQL injection.poc.php:16-17

Sql Execution

Payload withheldThe script sets up a test schema to demonstrate the vulnerability. This is destructive only to the 'users' table in the connected database, which is expected behavior for a PoC.poc.php:31-33

Exploit Payload

Payload withheldThese are the documented SQL injection payloads used to demonstrate data exfiltration and privilege escalation. They are the core of the PoC and match the CVE description.poc.php:39poc.php:60

Classification basis

The artifact contains code that actively exercises the vulnerability by executing a crafted SQL payload against a live PostgreSQL database to bypass a read-only transaction guard and create a table, which constitutes exploitation.

Requirements

• Requires a running PostgreSQL instance with valid credentials.poc.py:15-21
• Requires psycopg and sqlparse Python libraries.poc.py:13

Observed behavior

• Validates the payload using a replicated pgAdmin 9.16 sqlparse-based validator, which incorrectly classifies the multi-statement payload as a single SELECT.poc.py:104-112
• Connects to PostgreSQL, executes the payload via the simple query protocol inside a BEGIN TRANSACTION READ ONLY block, and checks if the smuggled CREATE TABLE succeeded.poc.py:118-157
• Demonstrates the fix by connecting with prepare_threshold=0 and showing that the extended query protocol rejects the multi-statement payload.poc.py:164-181

Observables

Payload

Payload withheldThe documented SQL injection payload that exploits the lexer differential.poc.py:48

Network Connection

Payload withheldThe PoC connects to a PostgreSQL instance specified by the user to demonstrate the vulnerability.poc.py:121poc.py:206-210

File Operation

Payload withheldThe PoC creates and drops a table named 'pwn' on the target database to demonstrate the bypass.poc.py:126poc.py:133poc.py:157

Classification basis

The artifact contains a complete Python script (exploit.py) that actively exercises the vulnerability by sending crafted requests to perform unauthorized actions (credential theft, file write, admin login). The README explicitly labels it as a 'Proof of Concept' and 'Full exploit'.

Requirements

• Target SiYuan instance with Publish server enabled in anonymous mode (Conf.Publish.Enable=true, Conf.Publish.Auth.Enable=false).README.md:22
• Python 3.8+ with requests library.README.md:33-34

Observed behavior

• Performs an MCP initialize handshake to obtain a session ID without authentication.exploit.py:79-102
• Reads payload withheld via the MCP file tool to extract accessAuthCode, api.token, and cookieKey.exploit.py:105-140
• Writes a malicious plugin (plugin.json and index.js) to data/plugins/pwn/ that executes an arbitrary shell command on next desktop launch.exploit.py:143-185
• Uses the stolen accessAuthCode to authenticate on the admin port (6806) and obtain an administrator session token.exploit.py:188-214

Observables

Network Target

Payload withheldThe exploit targets the SiYuan Publish port as described in the CVE.exploit.py:68

Credential Exfiltration

Payload withheldThe script reads the SiYuan configuration file to extract accessAuthCode, api.token, and cookieKey, matching the documented attack chain.exploit.py:105-140

File Write

Payload withheldThe script writes a malicious plugin to the SiYuan workspace for RCE on next desktop launch, as described in the CVE.exploit.py:143-185

Admin Escalation

Payload withheldThe script uses the stolen accessAuthCode to authenticate on the admin port, consistent with the documented attack chain.exploit.py:188-214

Classification basis

The script exploit.py is explicitly designed to send a crafted Intent to execute arbitrary commands on a vulnerable Bixby instance, which constitutes an exploit. The README.md describes it as a 'PoC exploit' and details the attack steps.

Requirements

• adb installed and a Samsung device connected with USB debugging enabledexploit.py:16-18
• Bixby version < 4.0.70.8 on the target deviceexploit.py:17

Observed behavior

• Sends a broadcast Intent to Bixby's exported command receiver to execute a shell commandexploit.py:108-114
• Reads the command output from a file on the device to confirm executionexploit.py:124-127

Observables

Command Execution

Payload withheldThe exploit sends a broadcast Intent to a Bixby component to execute a user-supplied command. This is the documented CVE-2026-21055 attack vector.exploit.py:108-114

File Write

Payload withheldThe exploit writes command output to a world-writable temporary file and reads it back. This is part of the PoC demonstration.exploit.py:34exploit.py:124-128

Classification basis

The artifact contains functional code (configuration_poc.py) designed to be executed by a vulnerable application to achieve remote code execution. The README explicitly labels it as a 'Proof of concept' and provides deployment instructions, including uploading the malicious model to Hugging Face Hub and triggering it via the WebUI. This constitutes an exploit.

Requirements

• Access to a LLaMA-Factory WebUI instance (version <= 0.9.5) or a local transformers environment for sink verification.README.md:3-4README.md:50-56
• The ability to supply a malicious model path to the WebUI or to run the provided build_and_verify.py script.README.md:76-83

Observed behavior

• The PoCConfig class in configuration_poc.py executes a payload (execute_payload()) upon instantiation, which runs the 'id' command via subprocess and prints host information.poc-model/configuration_poc.py:27-38poc-model/configuration_poc.py:41-46
• The README describes the expected output of the exploit, showing the payload banner and the output of the 'id' command.README.md:60-72

Observables

Payload Behavior

Payload withheldThe payload in configuration_poc.py is a harmless demonstration that prints system information and runs the 'id' command, consistent with the educational purpose stated in the README.poc-model/configuration_poc.py:27-38

Exploit Mechanism

Payload withheldThe PoC relies on the documented vulnerability where LLaMA-Factory hardcodes trust_remote_code=True, allowing the transformers library to execute code from a malicious model repository. This is the exact mechanism described in CVE-2026-58116.README.md:6-11poc-model/configuration_poc.py:1-16

Classification basis

The primary artifact, poc.py, sends crafted requests and inspects responses to determine if a vulnerability exists. It does not perform any post-exploitation actions, exfiltrate data, or establish persistence; it only reports whether the target is vulnerable. This is consistent with a scanner.

Requirements

• A running Traefik instance with a specific configuration (public router with StripPrefix middleware and a protected router with basicAuth) must be accessible.README.md:38-40

Observed behavior

• Sends unauthenticated HTTP GET requests to paths like /api../admin and /api%2e%2e/admin.poc.py:27-28
• Checks if the response status is 200 and the body contains a secret string, indicating a successful bypass.poc.py:56-60
• Prints a confirmation message if any protected path was reached without credentials.poc.py:76-78

Observables

Network Target

Payload withheldThe PoC script sends HTTP requests to this default local address to demonstrate the CVE. The target is user-configurable via command-line argument.poc.py:18

Vulnerability Demonstration

Payload withheldThe script sends crafted requests (e.g., /api../admin) to a Traefik instance to show that protected paths can be reached without credentials. This is the stated purpose of the PoC.poc.py:27-32

Classification basis

The script is named exploit.py, described as a 'Proof of Concept', and its primary function is to exercise the vulnerability by simulating an OAuth callback that results in the attacker obtaining a session for the victim's account. It includes code that directly demonstrates the account takeover (lines 105-112) and is intended to be executed to show the vulnerability.

Requirements

• Victim's email addressexploit.py:50
• Attacker controls an account on any accepted OAuth provider with the victim's emailexploit.py:74-79

Observed behavior

• Registers a local victim user with admin roleexploit.py:49-54
• Simulates an attacker OAuth callback with the victim's email but a different sub and email_verified=falseexploit.py:88-93
• Checks if the returned session user_id matches the victim's user_id, indicating account takeoverexploit.py:105-112
• Demonstrates fixed handlers that block the attack using (strategy, sub) matching and email verification policiesexploit.py:120-300

Observables

Vulnerability Demonstration

Payload withheldThe script simulates vulnerable and fixed authentication handlers to illustrate the security flaw and its remediation.exploit.py:37-117exploit.py:120-175

Educational Content

Payload withheldThe code is clearly documented for educational purposes, with no hidden or obfuscated logic.exploit.py:1-20README.md:1-94

Classification basis

The Python script constructs and sends malicious payloads to a vulnerable endpoint to execute arbitrary commands on the target device. The README explicitly describes it as a 'PoC' that 'exploits three unauthenticated command injection vulnerabilities', and the code contains functions named 'exploit_cve_*' that deliver command injection payloads.

Requirements

• Network access to the router's LAN IPREADME.md:35
• Python 3.6+README.md:34
• For CVE-2026-11452, a NAS user must exist on the targetpoc.py:83-85

Observed behavior

• Sends a JSON-RPC-like POST request to /cgi-bin/glc with method 'eject_disk_do1' and a crafted dev_name containing a command substitution payload to exploit a buffer size mismatch (CVE-2026-11450)poc.py:43-55
• Sends a JSON-RPC-like POST request to /cgi-bin/glc with method 'set_proto_config' and a crafted media_dir containing a single-quote escape and injected command (CVE-2026-11451)poc.py:57-68
• Sends a JSON-RPC-like POST request to /cgi-bin/glc with method 'set_user_pwd' and a crafted password containing a command substitution payload inside double-quoted printf (CVE-2026-11452)poc.py:70-100
• Executes an attacker-supplied shell command on the target and writes output to a remote filepoc.py:45poc.py:58poc.py:90

Observables

Network Target

Payload withheldAll HTTP requests are sent to the target specified by the user; no hardcoded or secondary exfiltration endpoint exists.poc.py:28-30poc.py:102-104

Command Execution

Payload withheldThe script constructs command injection payloads using the user-provided command and output path, consistent with the stated PoC purpose.poc.py:43-45poc.py:57-58poc.py:90-92

Classification basis

The artifact contains a complete, executable Python script (exploit.py) that actively crafts and injects malicious network frames designed to trigger a heap buffer overflow. The README explicitly labels it as a 'Proof of concept exploit' and provides usage instructions for running it against a target. The code's primary purpose is to exercise the vulnerability, not merely detect it.

Requirements

• Python 3.7+ with scapy library installedREADME.md:22-23exploit.py:29-37
• Linux with a Wi-Fi adapter supporting monitor and injection modeREADME.md:24
• Attacker must be an associated client on the target Wi-Fi network (adjacent network access)README.md:11
• Target AP BSSID and attacker MAC address must be knownexploit.py:112-114

Observed behavior

• Constructs an 802.11 QoS Data frame (type=2, subtype=8) with a malformed Information Element payloadexploit.py:52-59exploit.py:61-80
• Injects the crafted frames via a specified monitor-mode interface using scapy's sendp functionexploit.py:150-151
• Supports multiple IE overflow variants: Vendor Specific (221), Chained IEs, HT Capabilities (45), Extended Capabilities (127)exploit.py:61-80
• Allows configurable overflow size, number of injected frames, and inter-frame delayexploit.py:115-118

Observables

Exploit Behavior

Payload withheldThis is the normal, advertised behavior of the PoC and does not constitute a backdoor.exploit.py:43-86exploit.py:142-155

Network Activity

Payload withheldThe network activity is limited to the exploit's stated purpose of injecting malformed frames to the target AP.exploit.py:150-152

Classification basis

The artifact is a Python script that actively constructs and sends a malicious network packet designed to overflow a buffer on a target system. The README and script header explicitly describe it as a 'Proof of concept exploit' and 'exploit' that 'constructs a valid SNMPv3 message with an oversized msgAuthenticationParameters' to trigger a stack-based buffer overflow. The code includes network transmission logic (send_packet function) and is intended to cause a crash or execute a payload, which is the definition of an exploit.

Requirements

• Target must be running lwIP <= 2.2.1 with LWIP_SNMP_V3 enabled and listening on the specified UDP port (default 161).README.md:12
• Attacker must have network connectivity to the target's SNMP port.exploit.py:155-159

Observed behavior

• Constructs an SNMPv3 message with a USM security parameters sequence containing an oversized octet string for msgAuthenticationParameters.exploit.py:111-129
• Sends the crafted SNMPv3 packet to the target over UDP.exploit.py:155-159
• Supports sending multiple packets with configurable delay and loading a custom payload from a file.exploit.py:191-201

Observables

Network Behavior

Payload withheldThis is the expected behavior of a PoC exploit for the described vulnerability.exploit.py:155-165exploit.py:225-235

Payload Handling

Payload withheldThis allows a user to supply custom shellcode or data for the overflow, which is standard for exploit PoCs and does not constitute a backdoor.exploit.py:187-190exploit.py:206-214exploit.py:114-120

Classification basis

The script explicitly states it 'checks vulnerability status only' and 'does NOT exploit the vulnerability'. Its code only validates the presence of the vulnerable kernel interface (AF_ALG, authencesn template) and prints a status; it contains no code to perform the page-cache write or escalate privileges.

Requirements

• Python 3.10+ for os.splice supportREADME.md:25
• Linux kernel with AF_ALG supportREADME.md:26
• Unprivileged user accountREADME.md:27

Observed behavior

• Creates an AF_ALG socket and attempts to bind to the 'authencesn(hmac(sha256),cbc(aes))' AEAD templatecve_2026_31431_check.py:62-69
• Sets a key on the socket and attempts to accept a request socketcve_2026_31431_check.py:76-95
• Checks for the existence of the target file /usr/bin/sucve_2026_31431_check.py:99-101
• Prints a vulnerability status message and a conceptual description of the attack flow without executing any write primitivecve_2026_31431_check.py:105-132

Observables

Url

Payload withheldExternal link to the real PoC, referenced in README and script comments.README.md:21cve_2026_31431_check.py:10

Url

Payload withheldExternal article link in README and script output.README.md:11cve_2026_31431_check.py:162

Classification basis

The Python script generates a URL containing a crafted XSS payload designed to execute arbitrary JavaScript in the victim's browser, steal access tokens, and exfiltrate data. This is code intended to exercise a vulnerability, not merely detect it.

Requirements

• Victim must visit the crafted malicious URL while having an active STIG Manager session in another browser tab for the SharedWorker token theft mode.README.md:9
• Attacker must control or know the STIG Manager callback URL and optionally the API and exfiltration hosts.cve-2026-41200-poc.py:72-86

Observed behavior

• Constructs a URL with an 'error_description' query parameter containing a JavaScript payload that, when rendered by the vulnerable application, creates a SharedWorker to steal an access token and exfiltrate it along with collection data to an attacker-controlled host.cve-2026-41200-poc.py:29-50
• Provides a simple mode that injects a script tag to exfiltrate the victim's cookies.cve-2026-41200-poc.py:53-55
• Outputs the malicious URL in plain, curl, or HTML phishing link formats.cve-2026-41200-poc.py:120-129

Observables

Exfiltration Endpoint

Payload withheldThe sharedworker payload sends stolen tokens and collection data to an attacker-specified exfiltration host.cve-2026-41200-poc.py:38-42

Exfiltration Endpoint

Payload withheldThe sharedworker payload fallback sends document cookies to an attacker-specified host.cve-2026-41200-poc.py:48

Exfiltration Endpoint

Payload withheldThe simple payload sends document cookies to an attacker-specified host.cve-2026-41200-poc.py:54

Classification basis

The Python script's primary purpose is to generate a payload that triggers a stack-based buffer overflow, which is the definition of exploit code. The README explicitly states it 'Generates a raw RP-DATA message... that would trigger a stack-based buffer overflow'.

Requirements

• Requires a method to inject the generated RP-DATA message into the target baseband, such as a fake BTS with SDR, an SMS gateway with raw PDU mode, or direct memory injection via JTAG/UART.README.md:25-28poc_cve_2025_54328.py:107-110

Observed behavior

• Constructs an RP-DATA message with a 200-byte 'A' pattern in the TP-UD field, exceeding the expected buffer size to overwrite the stack.poc_cve_2025_54328.py:54-57
• Saves the crafted malicious message to a binary file for later injection.poc_cve_2025_54328.py:102-105

Observables

File Write

Payload withheldThe script writes the generated malicious RP-DATA message to a local file for later injection. This is expected behavior for a PoC that generates a payload.poc_cve_2025_54328.py:102-104

Command Line Argument

Payload withheldThe script accepts an optional payload withheld as a command-line argument, which is used to construct the destination address in the RP-DATA message. This is a standard parameter for a PoC of this type.poc_cve_2025_54328.py:73-76