All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.zaproxy.zap.spider.URLCanonicalizer Maven / Gradle / Ivy

Go to download

The Zed Attack Proxy (ZAP) is an easy to use integrated penetration testing tool for finding vulnerabilities in web applications. It is designed to be used by people with a wide range of security experience and as such is ideal for developers and functional testers who are new to penetration testing. ZAP provides automated scanners as well as a set of tools that allow you to find security vulnerabilities manually.

There is a newer version: 2.15.0
Show newest version
/*
 * Zed Attack Proxy (ZAP) and its related class files.
 *
 * ZAP is an HTTP/HTTPS proxy for assessing web application security.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 *
 * ZAP: Based on work by Yasser Ganjisaffar 
 * from project http://code.google.com/p/crawler4j/
 */
package org.zaproxy.zap.spider;

import java.net.URI;
import java.net.URL;
import java.util.HashSet;
import java.util.Set;
import java.util.SortedSet;
import java.util.TreeSet;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import org.apache.commons.httpclient.URIException;
import org.apache.commons.httpclient.util.URIUtil;
import org.apache.log4j.Logger;
import org.zaproxy.zap.spider.SpiderParam.HandleParametersOption;

/**
 * The URLCanonicalizer is used for the process of converting an URL into a canonical (normalized)
 * form. See URL Normalization for a
 * reference.
 *
 * 

Note: some parts of the code are adapted from: stackoverflow * *

Added support for OData URLs */ public final class URLCanonicalizer { /** The Constant log. */ private static final Logger log = Logger.getLogger(URLCanonicalizer.class); private static final String HTTP_SCHEME = "http"; private static final int HTTP_DEFAULT_PORT = 80; private static final String HTTPS_SCHEME = "https"; private static final int HTTPS_DEFAULT_PORT = 443; /** * The Constant IRRELEVANT_PARAMETERS defining the parameter names which are ignored in the URL. */ private static final Set IRRELEVANT_PARAMETERS = new HashSet<>(3); static { IRRELEVANT_PARAMETERS.add("jsessionid"); IRRELEVANT_PARAMETERS.add("phpsessid"); IRRELEVANT_PARAMETERS.add("aspsessionid"); } /** * OData support Extract the ID of a resource including the surrounding quote First group is the * resource_name Second group is the ID (quote will be taken as part of the value) */ private static final Pattern patternResourceIdentifierUnquoted = Pattern.compile("/([\\w%]*)\\(([\\w']*)\\)"); /** OData support Detect a section containing a composite IDs */ private static final Pattern patternResourceMultipleIdentifier = Pattern.compile("/[\\w%]*\\((.*)\\)"); /** OData support Extract the detail of the multiples IDs */ private static final Pattern patternResourceMultipleIdentifierDetail = Pattern.compile("([\\w%]*)=([\\w']*)"); /** Private constructor to avoid initialization of object. */ private URLCanonicalizer() {} /** * Gets the canonical url. * * @param url the url * @return the canonical url */ public static String getCanonicalURL(String url) { return getCanonicalURL(url, null); } /** * Gets the canonical url, starting from a relative or absolute url found in a given context * (baseURL). * * @param url the url string defining the reference * @param baseURL the context in which this url was found * @return the canonical url */ public static String getCanonicalURL(String url, String baseURL) { try { /* Build the absolute URL, from the url and the baseURL */ String resolvedURL = URLResolver.resolveUrl(baseURL == null ? "" : baseURL, url); log.debug("Resolved URL: " + resolvedURL); URI canonicalURI; try { canonicalURI = new URI(resolvedURL); } catch (Exception e) { canonicalURI = new URI(URIUtil.encodeQuery(resolvedURL)); } /* Some checking. */ if (canonicalURI.getScheme() == null) { log.warn( "Protocol could not be reliably evaluated from uri: " + canonicalURI + " and base url: " + baseURL); return null; } if (canonicalURI.getRawAuthority() == null) { log.debug( "Ignoring URI with no authority (host[\":\"port]): " + canonicalURI + " (on base " + baseURL + ")"); return null; } if (canonicalURI.getHost() == null) { log.warn( "Host could not be reliably evaluated from: " + canonicalURI + " (on base " + baseURL + ")"); return null; } /* * Normalize: no empty segments (i.e., "//"), no segments equal to ".", and no segments equal to * ".." that are preceded by a segment not equal to "..". */ String path = canonicalURI.normalize().getRawPath(); /* Convert '//' -> '/' */ int idx = path.indexOf("//"); while (idx >= 0) { path = path.replace("//", "/"); idx = path.indexOf("//"); } /* Drop starting '/../' */ while (path.startsWith("/../")) { path = path.substring(3); } /* Trim */ path = path.trim(); /* Process parameters and sort them. */ final SortedSet params = createSortedParameters(canonicalURI.getRawQuery()); final String queryString; String canonicalParams = canonicalize(params); queryString = (canonicalParams.isEmpty() ? "" : "?" + canonicalParams); /* Add starting slash if needed */ if (path.length() == 0) { path = "/" + path; } /* Drop default port: example.com:80 -> example.com */ int port = canonicalURI.getPort(); if (isDefaultPort(canonicalURI.getScheme(), port)) { port = -1; } /* Lowercasing protocol and host */ String protocol = canonicalURI.getScheme().toLowerCase(); String host = canonicalURI.getHost().toLowerCase(); String pathAndQueryString = normalizePath(path) + queryString; URL result = new URL(protocol, host, port, pathAndQueryString); return result.toExternalForm(); } catch (Exception ex) { log.warn( "Error while Processing URL [" + url + "] in the spidering process (on base " + baseURL + "): " + ex.getMessage()); return null; } } /** * Tells whether or not the given port is the default for the given scheme. * *

Note: Only HTTP and HTTPS schemes are taken into account. * * @param scheme the scheme * @param port the port * @return {@code true} if given the port is the default port for the given scheme, {@code * false} otherwise. */ private static boolean isDefaultPort(String scheme, int port) { return HTTP_SCHEME.equalsIgnoreCase(scheme) && port == HTTP_DEFAULT_PORT || HTTPS_SCHEME.equalsIgnoreCase(scheme) && port == HTTPS_DEFAULT_PORT; } /** * Builds a String representation of the URI with cleaned parameters, that can be used when * checking if an URI was already visited. The URI provided as a parameter should be already * cleaned and canonicalized, so it should be build with a result from {@link * #getCanonicalURL(String)}. * *

When building the URI representation, the same format should be used for all the cases, as * it may affect the number of times the pages are visited and reported if the option * HandleParametersOption is changed while the spider is running. * * @param uri the uri * @param handleParameters the handle parameters option * @param handleODataParametersVisited Should we handle specific OData parameters * @return the string representation of the URI * @throws URIException the URI exception */ public static String buildCleanedParametersURIRepresentation( org.apache.commons.httpclient.URI uri, SpiderParam.HandleParametersOption handleParameters, boolean handleODataParametersVisited) throws URIException { // If the option is set to use all the information, just use the default string // representation if (handleParameters.equals(HandleParametersOption.USE_ALL)) { return uri.toString(); } // If the option is set to ignore parameters completely, ignore the query completely if (handleParameters.equals(HandleParametersOption.IGNORE_COMPLETELY)) { return createBaseUriWithCleanedPath( uri, handleParameters, handleODataParametersVisited); } // If the option is set to ignore the value, we get the parameters and we only add their // name to the // query if (handleParameters.equals(HandleParametersOption.IGNORE_VALUE)) { StringBuilder retVal = new StringBuilder( createBaseUriWithCleanedPath( uri, handleParameters, handleODataParametersVisited)); String cleanedQuery = getCleanedQuery(uri.getEscapedQuery()); // Add the parameters' names to the uri representation. if (cleanedQuery.length() > 0) { retVal.append('?').append(cleanedQuery); } return retVal.toString(); } // Should not be reached return uri.toString(); } private static String createBaseUriWithCleanedPath( org.apache.commons.httpclient.URI uri, HandleParametersOption handleParameters, boolean handleODataParametersVisited) throws URIException { StringBuilder uriBuilder = new StringBuilder(createBaseUri(uri)); uriBuilder.append( getCleanedPath( uri.getEscapedPath(), handleParameters, handleODataParametersVisited)); return uriBuilder.toString(); } private static String createBaseUri(org.apache.commons.httpclient.URI uri) throws URIException { StringBuilder baseUriBuilder = new StringBuilder(); baseUriBuilder.append(uri.getScheme()).append("://").append(uri.getHost()); if (uri.getPort() != -1) { baseUriBuilder.append(':').append(uri.getPort()); } return baseUriBuilder.toString(); } private static String getCleanedPath( String escapedPath, HandleParametersOption handleParameters, boolean handleODataParametersVisited) { if (escapedPath == null) { return ""; } String cleanedPath; if (handleODataParametersVisited) { cleanedPath = cleanODataPath(escapedPath, handleParameters); } else { cleanedPath = escapedPath; } return cleanedPath; } private static String getCleanedQuery(String escapedQuery) { // Get the parameters' names SortedSet params = createSortedParameters(escapedQuery); Set parameterNames = new HashSet<>(); StringBuilder cleanedQueryBuilder = new StringBuilder(); if (params != null && !params.isEmpty()) { for (QueryParameter parameter : params) { String name = parameter.getName(); if (parameterNames.contains(name)) { continue; } parameterNames.add(name); // Ignore irrelevant parameters if (IRRELEVANT_PARAMETERS.contains(name) || name.startsWith("utm_")) { continue; } if (cleanedQueryBuilder.length() > 0) { cleanedQueryBuilder.append('&'); } cleanedQueryBuilder.append(name); } } return cleanedQueryBuilder.toString(); } /** * Clean the path in the case of an OData Uri containing a resource identifier (simple or * multiple) * * @param path The path to clean * @param handleParameters tThe cleaning mode * @return A cleaned path */ private static String cleanODataPath(String path, HandleParametersOption handleParameters) { String cleanedPath = path; if (HandleParametersOption.USE_ALL.equals(handleParameters)) { cleanedPath = path; } else { // check for single ID (unnamed) Matcher matcher = patternResourceIdentifierUnquoted.matcher(path); if (matcher.find()) { String resourceName = matcher.group(1); String resourceID = matcher.group(2); String subString = resourceName + "(" + resourceID + ")"; int begin = path.indexOf(subString); int end = begin + subString.length(); String beforeSubstring = path.substring(0, begin); String afterSubstring = path.substring(end); if (HandleParametersOption.IGNORE_COMPLETELY.equals(handleParameters) || HandleParametersOption.IGNORE_VALUE.equals(handleParameters)) { StringBuilder sb = new StringBuilder(beforeSubstring); sb.append(resourceName).append("()").append(afterSubstring); cleanedPath = sb.toString(); } } else { matcher = patternResourceMultipleIdentifier.matcher(path); if (matcher.find()) { // We've found a composite identifier. i.e: /Resource(field1=a,field2=3) String multipleIdentifierSection = matcher.group(1); int begin = path.indexOf(multipleIdentifierSection); int end = begin + multipleIdentifierSection.length(); String beforeSubstring = path.substring(0, begin); String afterSubstring = path.substring(end); if (HandleParametersOption.IGNORE_COMPLETELY.equals(handleParameters)) { cleanedPath = beforeSubstring + afterSubstring; } else { StringBuilder sb = new StringBuilder(beforeSubstring); matcher = patternResourceMultipleIdentifierDetail.matcher( multipleIdentifierSection); int i = 1; while (matcher.find()) { if (i > 1) { sb.append(','); } String paramName = matcher.group(1); sb.append(paramName); i++; } sb.append(afterSubstring); cleanedPath = sb.toString(); } } } } return cleanedPath; } /** * Creates a sorted set with all the parameters from the given {@code query}, ordered * lexicographically by name and value. * * @param queryString the query string * @return a sorted set with all parameters, or {@code null} if the query string is {@code null} * or empty. */ private static SortedSet createSortedParameters(final String queryString) { if (queryString == null || queryString.isEmpty()) { return null; } final String[] pairs = queryString.split("&"); final SortedSet params = new TreeSet<>(); for (final String pair : pairs) { if (pair.length() == 0) { continue; } String[] tokens = pair.split("=", 2); switch (tokens.length) { case 1: if (pair.charAt(0) == '=') { params.add(new QueryParameter("", tokens[0])); } else { params.add(new QueryParameter(tokens[0], "")); } break; case 2: params.add(new QueryParameter(tokens[0], tokens[1])); break; } } return params; } /** * Canonicalize the query string. * * @param sortedParameters Parameter name-value pairs in lexicographical order. * @return Canonical form of query string. */ private static String canonicalize(final SortedSet sortedParameters) { if (sortedParameters == null || sortedParameters.isEmpty()) { return ""; } final StringBuilder sb = new StringBuilder(100); for (QueryParameter parameter : sortedParameters) { final String name = parameter.getName().toLowerCase(); // Ignore irrelevant parameters if (IRRELEVANT_PARAMETERS.contains(name) || name.startsWith("utm_")) { continue; } if (sb.length() > 0) { sb.append('&'); } sb.append(parameter.getName()); if (!parameter.getValue().isEmpty()) { sb.append('='); sb.append(parameter.getValue()); } } return sb.toString(); } /** * Normalize path. * * @param path the path * @return the string */ private static String normalizePath(final String path) { return path.replace("%7E", "~").replace(" ", "%20"); } /** * A query parameter, with non-{@code null} name and value. * *

The query parameters are ordered by name and value. */ private static class QueryParameter implements Comparable { private final String name; private final String value; public QueryParameter(String name, String value) { if (name == null) { throw new IllegalArgumentException("Parameter name must not be null."); } if (value == null) { throw new IllegalArgumentException("Parameter value must not be null."); } this.name = name; this.value = value; } public String getName() { return name; } public String getValue() { return value; } @Override public int compareTo(QueryParameter other) { if (other == null) { return 1; } int result = name.compareTo(other.name); if (result != 0) { return result; } return value.compareTo(other.value); } @Override public int hashCode() { final int prime = 31; int result = 1; result = prime * result + name.hashCode(); result = prime * result + value.hashCode(); return result; } @Override public boolean equals(Object obj) { if (this == obj) { return true; } if (obj == null) { return false; } if (getClass() != obj.getClass()) { return false; } QueryParameter other = (QueryParameter) obj; if (!name.equals(other.name)) { return false; } if (!value.equals(other.value)) { return false; } return true; } } }





© 2015 - 2024 Weber Informatics LLC | Privacy Policy