Merge #9037: net: Add test-before-evict discipline to addrman

e68172e Add test-before-evict discipline to addrman (Ethan Heilman)

Pull request description:

  This change implement countermeasures 3 (test-before-evict) suggested in our paper: ["Eclipse Attacks on Bitcoin’s Peer-to-Peer Network"](http://cs-people.bu.edu/heilman/eclipse/).
  # Design:

  A collision occurs when an address, addr1, is being moved to the tried table from the new table, but maps to a position in the tried table which already contains an address (addr2). The current behavior is that addr1 would evict addr2 from the tried table.

  This change ensures that during a collision, addr1 is not inserted into tried but instead inserted into a buffer (setTriedCollisions). The to-be-evicted address, addr2, is then tested by [a feeler connection](bitcoin/bitcoin#8282). If addr2 is found to be online, we remove addr1 from the buffer and addr2 is not evicted, on the other hand if addr2 is found be offline it is replaced by addr1.

  An additional small advantage of this change is that, as no more than ten addresses can be in the test buffer at once, and addresses are only cleared one at a time from the test buffer (at 2 minute intervals), thus an attacker is forced to wait at least two minutes to insert a new address into tried after filling up the test buffer. This rate limits an attacker attempting to launch an eclipse attack.
  # Risk mitigation:
  - To prevent this functionality from being used as a DoS vector, we limit the number of addresses which are to be tested to ten. If we have more than ten addresses to test, we drop new addresses being added to tried if they would evict an address. Since the feeler thread only creates one new connection every 2 minutes the additional network overhead is limited.
  - An address in tried gains immunity from tests for 4 hours after it has been tested or successfully connected to.
  # Tests:

  This change includes additional addrman unittests which test this behavior.

  I ran an instance of this change with a much smaller tried table (2 buckets of 64 addresses) so that collisions were much more likely and observed evictions.

  ```
  2016-10-27 07:20:26 Swapping 208.12.64.252:8333 for 68.62.95.247:8333 in tried table
  2016-10-27 07:20:26 Moving 208.12.64.252:8333 to tried
  ```

  I documented tests we ran against similar earlier versions of this change in #6355.
  # Security Benefit

  This is was originally posted in PR #8282 see [this comment for full details](bitcoin/bitcoin#8282 (comment)).

  To determine the security benefit of these larger numbers of IPs in the tried table I modeled the attack presented in [Eclipse Attacks on Bitcoin’s Peer-to-Peer Network](https://eprint.iacr.org/2015/263).

  ![attackergraph40000-10-1000short-line](https://cloud.githubusercontent.com/assets/274814/17366828/372af458-595b-11e6-81e5-2c9f97282305.png)

  **Default node:** 595 attacker IPs for ~50% attack success.
  **Default node + test-before-evict:** 620 attacker IPs for ~50% attack success.
  **Feeler node:** 5540 attacker IPs for ~50% attack success.
  **Feeler node + test-before-evict:** 8600 attacker IPs for ~50% attack success.

  The node running feeler connections has 10 times as many online IP addresses in its tried table making an attack 10 times harder (i.e. requiring the an attacker require 10 times as many IP addresses in different /16s). Adding test-before-evict increases resistance of the node by an additional 3000 attacker IP addresses.

  Below I graph the attack over even greater attacker resources (i.e. more attacker controled IP addresses). Note that test-before-evict maintains some security far longer even against an attacker with 50,000 IPs. If this node had a larger tried table test-before-evict could greatly boost a nodes resistance to eclipse attacks.

  ![attacker graph long view](https://cloud.githubusercontent.com/assets/274814/17367108/96f46d64-595c-11e6-91cd-edba160598e7.png)

Tree-SHA512: fdad4d26aadeaad9bcdc71929b3eb4e1f855b3ee3541fbfbe25dca8d7d0a1667815402db0cb4319db6bd3fcd32d67b5bbc0e12045c4252d62d6239b7d77c4395

Author: laanwj

src/addrman.cpp

@@ -187,7 +187,7 @@ void CAddrMan::MakeTried(CAddrInfo& info, int nId)
     info.fInTried = true;
 }
 
-void CAddrMan::Good_(const CService& addr, int64_t nTime)
+void CAddrMan::Good_(const CService& addr, bool test_before_evict, int64_t nTime)
 {
     int nId;
 
@@ -233,10 +233,22 @@ void CAddrMan::Good_(const CService& addr, int64_t nTime)
     if (nUBucket == -1)
         return;
 
-    LogPrint(BCLog::ADDRMAN, "Moving %s to tried\n", addr.ToString());
+    // which tried bucket to move the entry to
+    int tried_bucket = info.GetTriedBucket(nKey);
+    int tried_bucket_pos = info.GetBucketPosition(nKey, false, tried_bucket);
+
+    // Will moving this address into tried evict another entry?
+    if (test_before_evict && (vvTried[tried_bucket][tried_bucket_pos] != -1)) {
+        LogPrint(BCLog::ADDRMAN, "addrman", "Collision inserting element into tried table, moving %s to m_tried_collisions=%d\n", addr.ToString(), m_tried_collisions.size());
+        if (m_tried_collisions.size() < ADDRMAN_SET_TRIED_COLLISION_SIZE) {
+            m_tried_collisions.insert(nId);
+        }
+    } else {
+        LogPrint(BCLog::ADDRMAN, "Moving %s to tried\n", addr.ToString());
 
-    // move nId to the tried tables
-    MakeTried(info, nId);
+        // move nId to the tried tables
+        MakeTried(info, nId);
+    }
 }
 
 bool CAddrMan::Add_(const CAddress& addr, const CNetAddr& source, int64_t nTimePenalty)
@@ -521,3 +533,82 @@ void CAddrMan::SetServices_(const CService& addr, ServiceFlags nServices)
 int CAddrMan::RandomInt(int nMax){
     return GetRandInt(nMax);
 }
+
+void CAddrMan::ResolveCollisions_()
+{
+    for (std::set<int>::iterator it = m_tried_collisions.begin(); it != m_tried_collisions.end();) {
+        int id_new = *it;
+
+        bool erase_collision = false;
+
+        // If id_new not found in mapInfo remove it from m_tried_collisions
+        if (mapInfo.count(id_new) != 1) {
+            erase_collision = true;
+        } else {
+            CAddrInfo& info_new = mapInfo[id_new];
+
+            // Which tried bucket to move the entry to.
+            int tried_bucket = info_new.GetTriedBucket(nKey);
+            int tried_bucket_pos = info_new.GetBucketPosition(nKey, false, tried_bucket);
+            if (!info_new.IsValid()) { // id_new may no longer map to a valid address
+                erase_collision = true;
+            } else if (vvTried[tried_bucket][tried_bucket_pos] != -1) { // The position in the tried bucket is not empty
+
+                // Get the to-be-evicted address that is being tested
+                int id_old = vvTried[tried_bucket][tried_bucket_pos];
+                CAddrInfo& info_old = mapInfo[id_old];
+
+                // Has successfully connected in last X hours
+                if (GetAdjustedTime() - info_old.nLastSuccess < ADDRMAN_REPLACEMENT_HOURS*(60*60)) {
+                    erase_collision = true;
+                } else if (GetAdjustedTime() - info_old.nLastTry < ADDRMAN_REPLACEMENT_HOURS*(60*60)) { // attempted to connect and failed in last X hours
+
+                    // Give address at least 60 seconds to successfully connect
+                    if (GetAdjustedTime() - info_old.nLastTry > 60) {
+                        LogPrint(BCLog::ADDRMAN, "addrman", "Swapping %s for %s in tried table\n", info_new.ToString(), info_old.ToString());
+
+                        // Replaces an existing address already in the tried table with the new address
+                        Good_(info_new, false, GetAdjustedTime());
+                        erase_collision = true;
+                    }
+                }
+            } else { // Collision is not actually a collision anymore
+                Good_(info_new, false, GetAdjustedTime());
+                erase_collision = true;
+            }
+        }
+
+        if (erase_collision) {
+            m_tried_collisions.erase(it++);
+        } else {
+            it++;
+        }
+    }
+}
+
+CAddrInfo CAddrMan::SelectTriedCollision_()
+{
+    if (m_tried_collisions.size() == 0) return CAddrInfo();
+
+    std::set<int>::iterator it = m_tried_collisions.begin();
+
+    // Selects a random element from m_tried_collisions
+    std::advance(it, GetRandInt(m_tried_collisions.size()));
+    int id_new = *it;
+
+    // If id_new not found in mapInfo remove it from m_tried_collisions
+    if (mapInfo.count(id_new) != 1) {
+        m_tried_collisions.erase(it);
+        return CAddrInfo();
+    }
+
+    CAddrInfo& newInfo = mapInfo[id_new];
+
+    // which tried bucket to move the entry to
+    int tried_bucket = newInfo.GetTriedBucket(nKey);
+    int tried_bucket_pos = newInfo.GetBucketPosition(nKey, false, tried_bucket);
+
+    int id_old = vvTried[tried_bucket][tried_bucket_pos];
+
+    return mapInfo[id_old];
+}

src/addrman.h

@@ -165,6 +165,9 @@ class CAddrInfo : public CAddress
 //! ... in at least this many days
 #define ADDRMAN_MIN_FAIL_DAYS 7
 
+//! how recent a successful connection should be before we allow an address to be evicted from tried
+#define ADDRMAN_REPLACEMENT_HOURS 4
+
 //! the maximum percentage of nodes to return in a getaddr call
 #define ADDRMAN_GETADDR_MAX_PCT 23
 
@@ -176,6 +179,9 @@ class CAddrInfo : public CAddress
 #define ADDRMAN_NEW_BUCKET_COUNT (1 << ADDRMAN_NEW_BUCKET_COUNT_LOG2)
 #define ADDRMAN_BUCKET_SIZE (1 << ADDRMAN_BUCKET_SIZE_LOG2)
 
+//! the maximum number of tried addr collisions to store
+#define ADDRMAN_SET_TRIED_COLLISION_SIZE 10
+
 /** 
  * Stochastical (IP) address manager 
  */
@@ -212,6 +218,9 @@ class CAddrMan
     //! last time Good was called (memory only)
     int64_t nLastGood;
 
+    //! Holds addrs inserted into tried table that collide with existing entries. Test-before-evict discpline used to resolve these collisions.
+    std::set<int> m_tried_collisions;
+
 protected:
     //! secret key to randomize bucket select with
     uint256 nKey;
@@ -239,7 +248,7 @@ class CAddrMan
     void ClearNew(int nUBucket, int nUBucketPos);
 
     //! Mark an entry "good", possibly moving it from "new" to "tried".
-    void Good_(const CService &addr, int64_t nTime);
+    void Good_(const CService &addr, bool test_before_evict, int64_t time);
 
     //! Add an entry to the "new" table.
     bool Add_(const CAddress &addr, const CNetAddr& source, int64_t nTimePenalty);
@@ -250,6 +259,12 @@ class CAddrMan
     //! Select an address to connect to, if newOnly is set to true, only the new table is selected from.
     CAddrInfo Select_(bool newOnly);
 
+    //! See if any to-be-evicted tried table entries have been tested and if so resolve the collisions.
+    void ResolveCollisions_();
+
+    //! Return a random to-be-evicted tried table address.
+    CAddrInfo SelectTriedCollision_();
+
     //! Wraps GetRandInt to allow tests to override RandomInt and make it determinismistic.
     virtual int RandomInt(int nMax);
 
@@ -537,11 +552,11 @@ class CAddrMan
     }
 
     //! Mark an entry as accessible.
-    void Good(const CService &addr, int64_t nTime = GetAdjustedTime())
+    void Good(const CService &addr, bool test_before_evict = true, int64_t nTime = GetAdjustedTime())
     {
         LOCK(cs);
         Check();
-        Good_(addr, nTime);
+        Good_(addr, test_before_evict, nTime);
         Check();
     }
 
@@ -554,6 +569,28 @@ class CAddrMan
         Check();
     }
 
+    //! See if any to-be-evicted tried table entries have been tested and if so resolve the collisions.
+    void ResolveCollisions()
+    {
+        LOCK(cs);
+        Check();
+        ResolveCollisions_();
+        Check();
+    }
+
+    //! Randomly select an address in tried that another address is attempting to evict.
+    CAddrInfo SelectTriedCollision()
+    {
+        CAddrInfo ret;
+        {
+            LOCK(cs);
+            Check();
+            ret = SelectTriedCollision_();
+            Check();
+        }
+        return ret;
+    }
+
     /**
      * Choose an address to connect to.
      */

src/net.cpp

@@ -1828,11 +1828,18 @@ void CConnman::ThreadOpenConnections(const std::vector<std::string> connect)
             }
         }
 
+        addrman.ResolveCollisions();
+
         int64_t nANow = GetAdjustedTime();
         int nTries = 0;
         while (!interruptNet)
         {
-            CAddrInfo addr = addrman.Select(fFeeler);
+            CAddrInfo addr = addrman.SelectTriedCollision();
+
+            // SelectTriedCollision returns an invalid address if it is empty.
+            if (!fFeeler || !addr.IsValid()) {
+                addr = addrman.Select(fFeeler);
+            }
 
             // if we selected an invalid address, restart
             if (!addr.IsValid() || setConnected.count(addr.GetGroup()) || IsLocal(addr))