wallet: wipe seed from memory where appropriate

3 files changed, 86 insertions, 75 deletions

diff --git a/src/mnemonics/electrum-words.cpp b/src/mnemonics/electrum-words.cpp
index 19a9c26bb..290f2cb93 100644
--- a/src/mnemonics/electrum-words.cpp
+++ b/src/mnemonics/electrum-words.cpp
@@ -43,6 +43,8 @@
 #include <vector>
 #include <unordered_map>
 #include <boost/algorithm/string.hpp>
+#include "wipeable_string.h"
+#include "misc_language.h"
 #include "crypto/crypto.h"  // for declaration of crypto::secret_key
 #include <fstream>
 #include "mnemonics/electrum-words.h"
@@ -80,9 +82,9 @@ namespace crypto
 
 namespace
 {
-  uint32_t create_checksum_index(const std::vector<std::string> &word_list,
+  uint32_t create_checksum_index(const std::vector<epee::wipeable_string> &word_list,
     uint32_t unique_prefix_length);
-  bool checksum_test(std::vector<std::string> seed, uint32_t unique_prefix_length);
+  bool checksum_test(std::vector<epee::wipeable_string> seed, uint32_t unique_prefix_length);
 
   /*!
    * \brief Finds the word list that contains the seed words and puts the indices
@@ -93,7 +95,7 @@ namespace
    * \param  language        Language instance pointer to write to after it is found.
    * \return                 true if all the words were present in some language false if not.
    */
-  bool find_seed_language(const std::vector<std::string> &seed,
+  bool find_seed_language(const std::vector<epee::wipeable_string> &seed,
     bool has_checksum, std::vector<uint32_t> &matched_indices, Language::Base **language)
   {
     // If there's a new language added, add an instance of it here.
@@ -114,17 +116,19 @@ namespace
     });
     Language::Base *fallback = NULL;
 
+    std::vector<epee::wipeable_string>::const_iterator it2;
+    matched_indices.reserve(seed.size());
+
     // Iterate through all the languages and find a match
     for (std::vector<Language::Base*>::iterator it1 = language_instances.begin();
       it1 != language_instances.end(); it1++)
     {
-      const std::unordered_map<std::string, uint32_t> &word_map = (*it1)->get_word_map();
-      const std::unordered_map<std::string, uint32_t> &trimmed_word_map = (*it1)->get_trimmed_word_map();
+      const std::unordered_map<epee::wipeable_string, uint32_t> &word_map = (*it1)->get_word_map();
+      const std::unordered_map<epee::wipeable_string, uint32_t> &trimmed_word_map = (*it1)->get_trimmed_word_map();
       // To iterate through seed words
-      std::vector<std::string>::const_iterator it2;
       bool full_match = true;
 
-      std::string trimmed_word;
+      epee::wipeable_string trimmed_word;
       // Iterate through all the words and see if they're all present
       for (it2 = seed.begin(); it2 != seed.end(); it2++)
       {
@@ -167,6 +171,7 @@ namespace
         return true;
       }
       // Some didn't match. Clear the index array.
+      memwipe(matched_indices.data(), matched_indices.size() * sizeof(matched_indices[0]));
       matched_indices.clear();
     }
 
@@ -181,6 +186,7 @@ namespace
     }
 
     MINFO("No match found");
+    memwipe(matched_indices.data(), matched_indices.size() * sizeof(matched_indices[0]));
     return false;
   }
 
@@ -190,12 +196,12 @@ namespace
    * \param unique_prefix_length  the prefix length of each word to use for checksum
    * \return                      Checksum index
    */
-  uint32_t create_checksum_index(const std::vector<std::string> &word_list,
+  uint32_t create_checksum_index(const std::vector<epee::wipeable_string> &word_list,
     uint32_t unique_prefix_length)
   {
-    std::string trimmed_words = "";
+    epee::wipeable_string trimmed_words = "";
 
-    for (std::vector<std::string>::const_iterator it = word_list.begin(); it != word_list.end(); it++)
+    for (std::vector<epee::wipeable_string>::const_iterator it = word_list.begin(); it != word_list.end(); it++)
     {
       if (it->length() > unique_prefix_length)
       {
@@ -217,22 +223,22 @@ namespace
    * \param unique_prefix_length  the prefix length of each word to use for checksum
    * \return                      True if the test passed false if not.
    */
-  bool checksum_test(std::vector<std::string> seed, uint32_t unique_prefix_length)
+  bool checksum_test(std::vector<epee::wipeable_string> seed, uint32_t unique_prefix_length)
   {
     if (seed.empty())
       return false;
     // The last word is the checksum.
-    std::string last_word = seed.back();
+    epee::wipeable_string last_word = seed.back();
     seed.pop_back();
 
-    std::string checksum = seed[create_checksum_index(seed, unique_prefix_length)];
+    epee::wipeable_string checksum = seed[create_checksum_index(seed, unique_prefix_length)];
 
-    std::string trimmed_checksum = checksum.length() > unique_prefix_length ? Language::utf8prefix(checksum, unique_prefix_length) :
+    epee::wipeable_string trimmed_checksum = checksum.length() > unique_prefix_length ? Language::utf8prefix(checksum, unique_prefix_length) :
       checksum;
-    std::string trimmed_last_word = last_word.length() > unique_prefix_length ? Language::utf8prefix(last_word, unique_prefix_length) :
+    epee::wipeable_string trimmed_last_word = last_word.length() > unique_prefix_length ? Language::utf8prefix(last_word, unique_prefix_length) :
       last_word;
     bool ret = trimmed_checksum == trimmed_last_word;
-    MINFO("Checksum is %s" << (ret ? "valid" : "invalid"));
+    MINFO("Checksum is " << (ret ? "valid" : "invalid"));
     return ret;
   }
 }
@@ -260,13 +266,12 @@ namespace crypto
      * \param  language_name   Language of the seed as found gets written here.
      * \return                 false if not a multiple of 3 words, or if word is not in the words list
      */
-    bool words_to_bytes(std::string words, std::string& dst, size_t len, bool duplicate,
+    bool words_to_bytes(const epee::wipeable_string &words, epee::wipeable_string& dst, size_t len, bool duplicate,
       std::string &language_name)
     {
-      std::vector<std::string> seed;
+      std::vector<epee::wipeable_string> seed;
 
-      boost::algorithm::trim(words);
-      boost::split(seed, words, boost::is_any_of(" "), boost::token_compress_on);
+      words.split(seed);
 
       if (len % 4)
       {
@@ -291,6 +296,7 @@ namespace crypto
       }
 
       std::vector<uint32_t> matched_indices;
+      auto wiper = epee::misc_utils::create_scope_leave_handler([&](){memwipe(matched_indices.data(), matched_indices.size() * sizeof(matched_indices[0]));});
       Language::Base *language;
       if (!find_seed_language(seed, has_checksum, matched_indices, &language))
       {
@@ -313,33 +319,33 @@ namespace crypto
 
       for (unsigned int i=0; i < seed.size() / 3; i++)
       {
-        uint32_t val;
-        uint32_t w1, w2, w3;
-        w1 = matched_indices[i*3];
-        w2 = matched_indices[i*3 + 1];
-        w3 = matched_indices[i*3 + 2];
+        uint32_t w[4];
+        w[1] = matched_indices[i*3];
+        w[2] = matched_indices[i*3 + 1];
+        w[3] = matched_indices[i*3 + 2];
 
-        val = w1 + word_list_length * (((word_list_length - w1) + w2) % word_list_length) +
-          word_list_length * word_list_length * (((word_list_length - w2) + w3) % word_list_length);
+        w[0]= w[1] + word_list_length * (((word_list_length - w[1]) + w[2]) % word_list_length) +
+          word_list_length * word_list_length * (((word_list_length - w[2]) + w[3]) % word_list_length);
 
-        if (!(val % word_list_length == w1))
+        if (!(w[0]% word_list_length == w[1]))
         {
+          memwipe(w, sizeof(w));
           MERROR("Invalid seed: mumble mumble");
           return false;
         }
 
-        dst.append((const char*)&val, 4);  // copy 4 bytes to position
+        dst.append((const char*)&w[0], 4);  // copy 4 bytes to position
+        memwipe(w, sizeof(w));
       }
 
       if (len > 0 && duplicate)
       {
         const size_t expected = len * 3 / 32;
-        std::string wlist_copy = words;
         if (seed.size() == expected/2)
         {
-          dst.append(dst);                    // if electrum 12-word seed, duplicate
-          wlist_copy += ' ';
-          wlist_copy += words;
+          dst += ' ';                    // if electrum 12-word seed, duplicate
+          dst += dst;                    // if electrum 12-word seed, duplicate
+          dst.pop_back();                // trailing space
         }
       }
 
@@ -353,10 +359,10 @@ namespace crypto
      * \param  language_name   Language of the seed as found gets written here.
      * \return                 false if not a multiple of 3 words, or if word is not in the words list
      */
-    bool words_to_bytes(std::string words, crypto::secret_key& dst,
+    bool words_to_bytes(const epee::wipeable_string &words, crypto::secret_key& dst,
       std::string &language_name)
     {
-      std::string s;
+      epee::wipeable_string s;
       if (!words_to_bytes(words, s, sizeof(dst), true, language_name))
       {
         MERROR("Invalid seed: failed to convert words to bytes");
@@ -378,7 +384,7 @@ namespace crypto
      * \param  language_name Seed language name
      * \return               true if successful false if not. Unsuccessful if wrong key size.
      */
-    bool bytes_to_words(const char *src, size_t len, std::string& words,
+    bool bytes_to_words(const char *src, size_t len, epee::wipeable_string& words,
       const std::string &language_name)
     {
 
@@ -397,39 +403,38 @@ namespace crypto
       }
       const std::vector<std::string> &word_list = language->get_word_list();
       // To store the words for random access to add the checksum word later.
-      std::vector<std::string> words_store;
+      std::vector<epee::wipeable_string> words_store;
 
       uint32_t word_list_length = word_list.size();
       // 4 bytes -> 3 words.  8 digits base 16 -> 3 digits base 1626
-      for (unsigned int i=0; i < len/4; i++, words += ' ')
+      for (unsigned int i=0; i < len/4; i++, words.push_back(' '))
       {
-        uint32_t w1, w2, w3;
-        
-        uint32_t val;
+        uint32_t w[4];
 
-        memcpy(&val, src + (i * 4), 4);
+        memcpy(&w[0], src + (i * 4), 4);
 
-        w1 = val % word_list_length;
-        w2 = ((val / word_list_length) + w1) % word_list_length;
-        w3 = (((val / word_list_length) / word_list_length) + w2) % word_list_length;
+        w[1] = w[0] % word_list_length;
+        w[2] = ((w[0] / word_list_length) + w[1]) % word_list_length;
+        w[3] = (((w[0] / word_list_length) / word_list_length) + w[2]) % word_list_length;
 
-        words += word_list[w1];
+        words += word_list[w[1]];
         words += ' ';
-        words += word_list[w2];
+        words += word_list[w[2]];
         words += ' ';
-        words += word_list[w3];
+        words += word_list[w[3]];
+
+        words_store.push_back(word_list[w[1]]);
+        words_store.push_back(word_list[w[2]]);
+        words_store.push_back(word_list[w[3]]);
 
-        words_store.push_back(word_list[w1]);
-        words_store.push_back(word_list[w2]);
-        words_store.push_back(word_list[w3]);
+        memwipe(w, sizeof(w));
       }
 
-      words.pop_back();
-      words += (' ' + words_store[create_checksum_index(words_store, language->get_unique_prefix_length())]);
+      words += words_store[create_checksum_index(words_store, language->get_unique_prefix_length())];
       return true;
     }
 
-    bool bytes_to_words(const crypto::secret_key& src, std::string& words,
+    bool bytes_to_words(const crypto::secret_key& src, epee::wipeable_string& words,
       const std::string &language_name)
     {
       return bytes_to_words(src.data, sizeof(src), words, language_name);
@@ -473,11 +478,10 @@ namespace crypto
      * \param  seed The seed to check (a space delimited concatenated word list)
      * \return      true if the seed passed is a old style seed false if not.
      */
-    bool get_is_old_style_seed(std::string seed)
+    bool get_is_old_style_seed(const epee::wipeable_string &seed)
     {
-      std::vector<std::string> word_list;
-      boost::algorithm::trim(seed);
-      boost::split(word_list, seed, boost::is_any_of(" "), boost::token_compress_on);
+      std::vector<epee::wipeable_string> word_list;
+      seed.split(word_list);
       return word_list.size() != (seed_length + 1);
     }
 
diff --git a/src/mnemonics/electrum-words.h b/src/mnemonics/electrum-words.h
index 856edb92a..5401b9779 100644
--- a/src/mnemonics/electrum-words.h
+++ b/src/mnemonics/electrum-words.h
@@ -44,6 +44,8 @@
 #include <map>
 #include "crypto/crypto.h"  // for declaration of crypto::secret_key
 
+namespace epee {  class wipeable_string; }
+
 /*!
  * \namespace crypto
  * 
@@ -70,7 +72,7 @@ namespace crypto
      * \param  language_name   Language of the seed as found gets written here.
      * \return                 false if not a multiple of 3 words, or if word is not in the words list
      */
-    bool words_to_bytes(std::string words, std::string& dst, size_t len, bool duplicate,
+    bool words_to_bytes(const epee::wipeable_string &words, epee::wipeable_string& dst, size_t len, bool duplicate,
       std::string &language_name);
     /*!
      * \brief Converts seed words to bytes (secret key).
@@ -79,7 +81,7 @@ namespace crypto
      * \param  language_name   Language of the seed as found gets written here.
      * \return                 false if not a multiple of 3 words, or if word is not in the words list
      */
-    bool words_to_bytes(std::string words, crypto::secret_key& dst,
+    bool words_to_bytes(const epee::wipeable_string &words, crypto::secret_key& dst,
       std::string &language_name);
 
     /*!
@@ -90,7 +92,7 @@ namespace crypto
      * \param  language_name Seed language name
      * \return               true if successful false if not. Unsuccessful if wrong key size.
      */
-    bool bytes_to_words(const char *src, size_t len, std::string& words,
+    bool bytes_to_words(const char *src, size_t len, epee::wipeable_string& words,
       const std::string &language_name);
 
     /*!
@@ -100,7 +102,7 @@ namespace crypto
      * \param  language_name Seed language name
      * \return               true if successful false if not. Unsuccessful if wrong key size.
      */
-    bool bytes_to_words(const crypto::secret_key& src, std::string& words,
+    bool bytes_to_words(const crypto::secret_key& src, epee::wipeable_string& words,
       const std::string &language_name);
 
     /*!
@@ -115,7 +117,7 @@ namespace crypto
      * \param  seed The seed to check (a space delimited concatenated word list)
      * \return      true if the seed passed is a old style seed false if not.
      */
-    bool get_is_old_style_seed(std::string seed);
+    bool get_is_old_style_seed(const epee::wipeable_string &seed);
 
     /*!
      * \brief Returns the name of a language in English
diff --git a/src/mnemonics/language_base.h b/src/mnemonics/language_base.h
index 2b0c37c6b..cf518ab2a 100644
--- a/src/mnemonics/language_base.h
+++ b/src/mnemonics/language_base.h
@@ -53,15 +53,20 @@ namespace Language
    * \param  count           How many characters to return.

    * \return                 A string consisting of the first count characters in s.

    */

-  inline std::string utf8prefix(const std::string &s, size_t count)

+  template<typename T>

+  inline T utf8prefix(const T &s, size_t count)

   {

-    std::string prefix = "";

-    const char *ptr = s.c_str();

-    while (count-- && *ptr)

+    T prefix = "";

+    size_t avail = s.size();

+    const char *ptr = s.data();

+    while (count-- && avail--)

     {

       prefix += *ptr++;

-      while (((*ptr) & 0xc0) == 0x80)

+      while (avail && ((*ptr) & 0xc0) == 0x80)

+      {

         prefix += *ptr++;

+        --avail;

+      }

     }

     return prefix;

   }

@@ -79,8 +84,8 @@ namespace Language
       ALLOW_DUPLICATE_PREFIXES = 1<<1,

     };

     const std::vector<std::string> word_list; /*!< A pointer to the array of words */

-    std::unordered_map<std::string, uint32_t> word_map; /*!< hash table to find word's index */

-    std::unordered_map<std::string, uint32_t> trimmed_word_map; /*!< hash table to find word's trimmed index */

+    std::unordered_map<epee::wipeable_string, uint32_t> word_map; /*!< hash table to find word's index */

+    std::unordered_map<epee::wipeable_string, uint32_t> trimmed_word_map; /*!< hash table to find word's trimmed index */

     std::string language_name; /*!< Name of language */

     std::string english_language_name; /*!< Name of language */

     uint32_t unique_prefix_length; /*!< Number of unique starting characters to trim the wordlist to when matching */

@@ -103,7 +108,7 @@ namespace Language
           else

             throw std::runtime_error("Too short word in " + language_name + " word list: " + *it);

         }

-        std::string trimmed;

+        epee::wipeable_string trimmed;

         if (it->length() > unique_prefix_length)

         {

           trimmed = utf8prefix(*it, unique_prefix_length);

@@ -115,9 +120,9 @@ namespace Language
         if (trimmed_word_map.find(trimmed) != trimmed_word_map.end())

         {

           if (flags & ALLOW_DUPLICATE_PREFIXES)

-            MWARNING("Duplicate prefix in " << language_name << " word list: " << trimmed);

+            MWARNING("Duplicate prefix in " << language_name << " word list: " << std::string(trimmed.data(), trimmed.size()));

           else

-            throw std::runtime_error("Duplicate prefix in " + language_name + " word list: " + trimmed);

+            throw std::runtime_error("Duplicate prefix in " + language_name + " word list: " + std::string(trimmed.data(), trimmed.size()));

         }

         trimmed_word_map[trimmed] = ii;

       }

@@ -145,7 +150,7 @@ namespace Language
      * \brief Returns a pointer to the word map.

      * \return A pointer to the word map.

      */

-    const std::unordered_map<std::string, uint32_t>& get_word_map() const

+    const std::unordered_map<epee::wipeable_string, uint32_t>& get_word_map() const

     {

       return word_map;

     }

@@ -153,7 +158,7 @@ namespace Language
      * \brief Returns a pointer to the trimmed word map.

      * \return A pointer to the trimmed word map.

      */

-    const std::unordered_map<std::string, uint32_t>& get_trimmed_word_map() const

+    const std::unordered_map<epee::wipeable_string, uint32_t>& get_trimmed_word_map() const

     {

       return trimmed_word_map;

     }