X-Git-Url: http://cvs.zerfleddert.de/cgi-bin/gitweb.cgi/proxmark3-svn/blobdiff_plain/61dda55df6b4246dcaca5d67c557c74ba9f3838e..388d8618c7ef6ca8fed92df09fc41e9b6014f693:/client/cmdhfmfhard.c

diff --git a/client/cmdhfmfhard.c b/client/cmdhfmfhard.c
index eac783ff..3fed7c95 100644
--- a/client/cmdhfmfhard.c
+++ b/client/cmdhfmfhard.c
@@ -1,6 +1,7 @@
 //-----------------------------------------------------------------------------
 // Copyright (C) 2015 piwi
-//
+// fiddled with 2016 Azcid (hardnested bitsliced Bruteforce imp)
+// fiddled with 2016 Matrix ( sub testing of nonces while collecting )
 // This code is licensed to you under the terms of the GNU GPL, version 2 or,
 // at your option, any later version. See the LICENSE.txt file for the text of
 // the license.
@@ -13,26 +14,16 @@
 //   Mifare Classic Cards" in Proceedings of the 22nd ACM SIGSAC Conference on 
 //   Computer and Communications Security, 2015
 //-----------------------------------------------------------------------------
-
-#include <stdio.h>
-#include <stdlib.h> 
-#include <string.h>
-#include <pthread.h>
-#include <locale.h>
-#include <math.h>
-#include "proxmark3.h"
-#include "cmdmain.h"
-#include "ui.h"
-#include "util.h"
-#include "nonce2key/crapto1.h"
-#include "parity.h"
-
-// uint32_t test_state_odd = 0;
-// uint32_t test_state_even = 0;
+#include "cmdhfmfhard.h"
+#include "cmdhw.h"
 
 #define CONFIDENCE_THRESHOLD	0.95		// Collect nonces until we are certain enough that the following brute force is successfull
-#define GOOD_BYTES_REQUIRED		30
+#define GOOD_BYTES_REQUIRED	13		// default 28, could be smaller == faster
+#define NONCES_THRESHOLD	5000		// every N nonces check if we can crack the key
+#define CRACKING_THRESHOLD	36.0f //38.50f		// as 2^38.5
+#define MAX_BUCKETS		128
 
+#define END_OF_LIST_MARKER		0xFFFFFFFF
 
 static const float p_K[257] = {		// the probability that a random nonce has a Sum Property == K 
 	0.0290, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 
@@ -68,7 +59,6 @@ static const float p_K[257] = {		// the probability that a random nonce has a Su
 	0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
 	0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
 	0.0290 };
-
 		
 typedef struct noncelistentry {
 	uint32_t nonce_enc;
@@ -84,10 +74,12 @@ typedef struct noncelist {
 	float Sum8_prob;
 	bool updated;
 	noncelistentry_t *first;
-	float score1, score2;
+	float score1;
+	uint_fast8_t score2;
 } noncelist_t;
 
-
+static size_t nonces_to_bruteforce = 0;
+static noncelistentry_t *brute_force_nonces[256];
 static uint32_t cuid = 0;
 static noncelist_t nonces[256];
 static uint8_t best_first_bytes[256];
@@ -123,9 +115,28 @@ typedef struct {
 
 static partial_indexed_statelist_t partial_statelist[17];
 static partial_indexed_statelist_t statelist_bitflip;
-
 static statelist_t *candidates = NULL;
 
+bool field_off = false;
+
+uint64_t foundkey = 0;
+size_t keys_found = 0;
+size_t bucket_count = 0;
+statelist_t* buckets[MAX_BUCKETS];
+static uint64_t total_states_tested = 0;
+size_t thread_count = 4;
+
+// these bitsliced states will hold identical states in all slices
+bitslice_t bitsliced_rollback_byte[ROLLBACK_SIZE];
+
+// arrays of bitsliced states with identical values in all slices
+bitslice_t bitsliced_encrypted_nonces[NONCE_TESTS][STATE_SIZE];
+bitslice_t bitsliced_encrypted_parity_bits[NONCE_TESTS][ROLLBACK_SIZE];
+
+#define EXACT_COUNT
+
+static bool generate_candidates(uint16_t, uint16_t);
+static bool brute_force(void);
 
 static int add_nonce(uint32_t nonce_enc, uint8_t par_enc) 
 {
@@ -154,14 +165,17 @@ static int add_nonce(uint32_t nonce_enc, uint8_t par_enc)
 		} else {					// add new entry at end of existing list.
 			p2 = p2->next = malloc(sizeof(noncelistentry_t));
 		}
-	} else if ((p1->nonce_enc & 0x00ff0000) != (nonce_enc & 0x00ff0000)) {				// found distinct 2nd byte. Need to insert.
+		if (p2 == NULL) return 0;							// memory allocation failed
+	}
+	else if ((p1->nonce_enc & 0x00ff0000) != (nonce_enc & 0x00ff0000)) {			// found distinct 2nd byte. Need to insert.
 		if (p2 == NULL) {			// need to insert at start of list
 			p2 = nonces[first_byte].first = malloc(sizeof(noncelistentry_t));
 		} else {
 			p2 = p2->next = malloc(sizeof(noncelistentry_t));
 		}
-	} else {											// we have seen this 2nd byte before. Nothing to add or insert. 
-		return (0);
+		if (p2 == NULL) return 0;							// memory allocation failed
+	} else {
+		return 0;									// we have seen this 2nd byte before. Nothing to add or insert.
 	}
 
 	// add or insert new data
@@ -169,14 +183,18 @@ static int add_nonce(uint32_t nonce_enc, uint8_t par_enc)
 	p2->nonce_enc = nonce_enc;
 	p2->par_enc = par_enc;
 
+    if(nonces_to_bruteforce < 256){
+        brute_force_nonces[nonces_to_bruteforce] = p2;
+        nonces_to_bruteforce++;
+    }
+
 	nonces[first_byte].num++;
 	nonces[first_byte].Sum += evenparity32((nonce_enc & 0x00ff0000) | (par_enc & 0x04));
 	nonces[first_byte].updated = true;   // indicates that we need to recalculate the Sum(a8) probability for this first byte
 
-	return (1);				// new nonce added
+	return 1;				// new nonce added
 }
 
-
 static void init_nonce_memory(void)
 {
 	for (uint16_t i = 0; i < 256; i++) {
@@ -192,7 +210,6 @@ static void init_nonce_memory(void)
 	num_good_first_bytes = 0;
 }
 
-
 static void free_nonce_list(noncelistentry_t *p)
 {
 	if (p == NULL) {
@@ -203,7 +220,6 @@ static void free_nonce_list(noncelistentry_t *p)
 	}
 }
 
-
 static void free_nonces_memory(void)
 {
 	for (uint16_t i = 0; i < 256; i++) {
@@ -211,7 +227,6 @@ static void free_nonces_memory(void)
 	}
 }
 
-
 static uint16_t PartialSumProperty(uint32_t state, odd_even_t odd_even)
 { 
 	uint16_t sum = 0;
@@ -235,7 +250,6 @@ static uint16_t PartialSumProperty(uint32_t state, odd_even_t odd_even)
 	return sum;
 }
 
-
 // static uint16_t SumProperty(struct Crypto1State *s)
 // {
 	// uint16_t sum_odd = PartialSumProperty(s->odd, ODD_STATE);
@@ -243,7 +257,6 @@ static uint16_t PartialSumProperty(uint32_t state, odd_even_t odd_even)
 	// return (sum_odd*(16-sum_even) + (16-sum_odd)*sum_even);
 // }
 
-
 static double p_hypergeometric(uint16_t N, uint16_t K, uint16_t n, uint16_t k) 
 {
 	// for efficient computation we are using the recursive definition
@@ -281,8 +294,7 @@ static double p_hypergeometric(uint16_t N, uint16_t K, uint16_t n, uint16_t k)
 		}
 	}
 }
-	
-	
+
 static float sum_probability(uint16_t K, uint16_t n, uint16_t k)
 {
 	const uint16_t N = 256;
@@ -290,19 +302,19 @@ static float sum_probability(uint16_t K, uint16_t n, uint16_t k)
 	if (k > K || p_K[K] == 0.0) return 0.0;
 
 	double p_T_is_k_when_S_is_K = p_hypergeometric(N, K, n, k);
+	if (p_T_is_k_when_S_is_K == 0.0) return 0.0;
+
 	double p_S_is_K = p_K[K];
-	double p_T_is_k = 0;
+	double p_T_is_k = 0.0;
 	for (uint16_t i = 0; i <= 256; i++) {
 		if (p_K[i] != 0.0) {
 			p_T_is_k += p_K[i] * p_hypergeometric(N, i, n, k);
 		}
 	}
+	if (p_T_is_k == 0.0) return 0.0;
 	return(p_T_is_k_when_S_is_K * p_S_is_K / p_T_is_k);
 }
 
-		
-
-	
 static inline uint_fast8_t common_bits(uint_fast8_t bytes_diff) 
 {
 	static const uint_fast8_t common_bits_LUT[256] = {
@@ -327,7 +339,6 @@ static inline uint_fast8_t common_bits(uint_fast8_t bytes_diff)
 	return common_bits_LUT[bytes_diff];
 }
 
-
 static void Tests()
 {
 	// printf("Tests: Partial Statelist sizes\n");
@@ -445,32 +456,31 @@ static void Tests()
 	// crypto1_destroy(pcs);
 
 	
-	
 	// printf("\nTests: number of states with BitFlipProperty: %d, (= %1.3f%% of total states)\n", statelist_bitflip.len[0], 100.0 * statelist_bitflip.len[0] / (1<<20));
 
-	printf("\nTests: Actual BitFlipProperties odd/even:\n");
-	for (uint16_t i = 0; i < 256; i++) {
-		printf("[%02x]:%c  ", i, nonces[i].BitFlip[ODD_STATE]?'o':nonces[i].BitFlip[EVEN_STATE]?'e':' ');
-		if (i % 8 == 7) {
-			printf("\n");
-		}
-	}
+	// printf("\nTests: Actual BitFlipProperties odd/even:\n");
+	// for (uint16_t i = 0; i < 256; i++) {
+		// printf("[%02x]:%c  ", i, nonces[i].BitFlip[ODD_STATE]?'o':nonces[i].BitFlip[EVEN_STATE]?'e':' ');
+		// if (i % 8 == 7) {
+			// printf("\n");
+		// }
+	// }
 	
-	printf("\nTests: Sorted First Bytes:\n");
-	for (uint16_t i = 0; i < 256; i++) {
-		uint8_t best_byte = best_first_bytes[i];
-		printf("#%03d Byte: %02x, n = %3d, k = %3d, Sum(a8): %3d, Confidence: %5.1f%%, Bitflip: %c\n", 
-		//printf("#%03d Byte: %02x, n = %3d, k = %3d, Sum(a8): %3d, Confidence: %5.1f%%, Bitflip: %c, score1: %1.5f, score2: %1.0f\n", 
-			i, best_byte, 
-			nonces[best_byte].num,
-			nonces[best_byte].Sum,
-			nonces[best_byte].Sum8_guess,
-			nonces[best_byte].Sum8_prob * 100,
-			nonces[best_byte].BitFlip[ODD_STATE]?'o':nonces[best_byte].BitFlip[EVEN_STATE]?'e':' '
-			//nonces[best_byte].score1,
-			//nonces[best_byte].score2
-			);
-	}
+	// printf("\nTests: Sorted First Bytes:\n");
+	// for (uint16_t i = 0; i < 256; i++) {
+		// uint8_t best_byte = best_first_bytes[i];
+		// printf("#%03d Byte: %02x, n = %3d, k = %3d, Sum(a8): %3d, Confidence: %5.1f%%, Bitflip: %c\n", 
+		// //printf("#%03d Byte: %02x, n = %3d, k = %3d, Sum(a8): %3d, Confidence: %5.1f%%, Bitflip: %c, score1: %1.5f, score2: %1.0f\n", 
+			// i, best_byte, 
+			// nonces[best_byte].num,
+			// nonces[best_byte].Sum,
+			// nonces[best_byte].Sum8_guess,
+			// nonces[best_byte].Sum8_prob * 100,
+			// nonces[best_byte].BitFlip[ODD_STATE]?'o':nonces[best_byte].BitFlip[EVEN_STATE]?'e':' '
+			// //nonces[best_byte].score1,
+			// //nonces[best_byte].score2
+			// );
+	// }
 	
 	// printf("\nTests: parity performance\n");
 	// time_t time1p = clock();
@@ -490,8 +500,7 @@ static void Tests()
 
 }
 
-
-static void sort_best_first_bytes(void)
+static uint16_t sort_best_first_bytes(void)
 {
 	// sort based on probability for correct guess	
 	for (uint16_t i = 0; i < 256; i++ ) {
@@ -506,8 +515,8 @@ static void sort_best_first_bytes(void)
 				best_first_bytes[k] = best_first_bytes[k-1];
 			}
 		}
-			best_first_bytes[j] = i;
-		}
+		best_first_bytes[j] = i;
+	}
 
 	// determine how many are above the CONFIDENCE_THRESHOLD
 	uint16_t num_good_nonces = 0;
@@ -517,6 +526,8 @@ static void sort_best_first_bytes(void)
 		}
 	}
 	
+	if (num_good_nonces == 0) return 0;
+
 	uint16_t best_first_byte = 0;
 
 	// select the best possible first byte based on number of common bits with all {b'}
@@ -539,25 +550,28 @@ static void sort_best_first_bytes(void)
 	for (uint16_t i = 0; i < num_good_nonces; i++ ) {
 		uint16_t sum8 = nonces[best_first_bytes[i]].Sum8_guess;
 		float bitflip_prob = 1.0;
-		if (nonces[best_first_bytes[i]].BitFlip[ODD_STATE] || nonces[best_first_bytes[i]].BitFlip[EVEN_STATE]) {
+		
+		if (nonces[best_first_bytes[i]].BitFlip[ODD_STATE] || nonces[best_first_bytes[i]].BitFlip[EVEN_STATE])
 			bitflip_prob = 0.09375;
-		}
+		
 		nonces[best_first_bytes[i]].score1 = p_K[sum8] * bitflip_prob;
-		if (p_K[sum8] * bitflip_prob <= min_p_K) {
+		
+		if (p_K[sum8] * bitflip_prob <= min_p_K)
 			min_p_K = p_K[sum8] * bitflip_prob;
-		}
+		
 	}
 
 
 	// use number of commmon bits as a tie breaker
-	uint16_t max_common_bits = 0;
+	uint_fast8_t max_common_bits = 0;
 	for (uint16_t i = 0; i < num_good_nonces; i++) {
+
 		float bitflip_prob = 1.0;
-		if (nonces[best_first_bytes[i]].BitFlip[ODD_STATE] || nonces[best_first_bytes[i]].BitFlip[EVEN_STATE]) {
+		if (nonces[best_first_bytes[i]].BitFlip[ODD_STATE] || nonces[best_first_bytes[i]].BitFlip[EVEN_STATE])
 			bitflip_prob = 0.09375;
-		}
+		
 		if (p_K[nonces[best_first_bytes[i]].Sum8_guess] * bitflip_prob == min_p_K) {
-			uint16_t sum_common_bits = 0;
+			uint_fast8_t sum_common_bits = 0;
 			for (uint16_t j = 0; j < num_good_nonces; j++) {
 				sum_common_bits += common_bits(best_first_bytes[i] ^ best_first_bytes[j]);
 			}
@@ -570,16 +584,17 @@ static void sort_best_first_bytes(void)
 	}	
 
 	// swap best possible first byte to the pole position
-	uint16_t temp = best_first_bytes[0];
-	best_first_bytes[0] = best_first_bytes[best_first_byte];
-	best_first_bytes[best_first_byte] = temp;
+	if (best_first_byte != 0) {
+		uint16_t temp = best_first_bytes[0];
+		best_first_bytes[0] = best_first_bytes[best_first_byte];
+		best_first_bytes[best_first_byte] = temp;
+	}
 	
+	return num_good_nonces;
 }
 
-
 static uint16_t estimate_second_byte_sum(void) 
-{
-	
+{	
 	for (uint16_t first_byte = 0; first_byte < 256; first_byte++) {
 		float Sum8_prob = 0.0;
 		uint16_t Sum8 = 0;
@@ -596,28 +611,17 @@ static uint16_t estimate_second_byte_sum(void)
 			nonces[first_byte].updated = false;
 		}
 	}
-	
-	sort_best_first_bytes();
-
-	uint16_t num_good_nonces = 0;
-	for (uint16_t i = 0; i < 256; i++) {
-		if (nonces[best_first_bytes[i]].Sum8_prob >= CONFIDENCE_THRESHOLD) {
-			++num_good_nonces;
-		}
-	}
-	
-	return num_good_nonces;
+	return sort_best_first_bytes();
 }	
 
-
 static int read_nonce_file(void)
 {
 	FILE *fnonces = NULL;
-	uint8_t trgBlockNo;
-	uint8_t trgKeyType;
+	uint8_t trgBlockNo = 0;
+	uint8_t trgKeyType = 0;
 	uint8_t read_buf[9];
-	uint32_t nt_enc1, nt_enc2;
-	uint8_t par_enc;
+	uint32_t nt_enc1 = 0, nt_enc2 = 0;
+	uint8_t par_enc = 0;
 	int total_num_nonces = 0;
 	
 	if ((fnonces = fopen("nonces.bin","rb")) == NULL) { 
@@ -626,6 +630,7 @@ static int read_nonce_file(void)
 	}
 
 	PrintAndLog("Reading nonces from file nonces.bin...");
+	memset (read_buf, 0, sizeof (read_buf));
 	size_t bytes_read = fread(read_buf, 1, 6, fnonces);
 	if ( bytes_read == 0) {
 		PrintAndLog("File reading error.");
@@ -635,8 +640,10 @@ static int read_nonce_file(void)
 	cuid = bytes_to_num(read_buf, 4);
 	trgBlockNo = bytes_to_num(read_buf+4, 1);
 	trgKeyType = bytes_to_num(read_buf+5, 1);
-
-	while (fread(read_buf, 1, 9, fnonces) == 9) {
+	size_t ret = 0;
+	do {
+		memset (read_buf, 0, sizeof (read_buf));
+		if ((ret = fread(read_buf, 1, 9, fnonces)) == 9) {
 		nt_enc1 = bytes_to_num(read_buf, 4);
 		nt_enc2 = bytes_to_num(read_buf+4, 4);
 		par_enc = bytes_to_num(read_buf+8, 1);
@@ -646,17 +653,16 @@ static int read_nonce_file(void)
 		add_nonce(nt_enc2, par_enc & 0x0f);
 		total_num_nonces += 2;
 	}
+	} while (ret == 9);
+
 	fclose(fnonces);
 	PrintAndLog("Read %d nonces from file. cuid=%08x, Block=%d, Keytype=%c", total_num_nonces, cuid, trgBlockNo, trgKeyType==0?'A':'B');
-
 	return 0;
 }
 
-
 static void Check_for_FilterFlipProperties(void)
 {
 	printf("Checking for Filter Flip Properties...\n");
-
 	uint16_t num_bitflips = 0;
 	
 	for (uint16_t i = 0; i < 256; i++) {
@@ -665,6 +671,8 @@ static void Check_for_FilterFlipProperties(void)
 	}
 	
 	for (uint16_t i = 0; i < 256; i++) {
+		if (!nonces[i].first || !nonces[i^0x80].first || !nonces[i^0x40].first) continue;
+
 		uint8_t parity1 = (nonces[i].first->par_enc) >> 3;				// parity of first byte
 		uint8_t parity2_odd = (nonces[i^0x80].first->par_enc) >> 3;  	// XOR 0x80 = last bit flipped
 		uint8_t parity2_even = (nonces[i^0x40].first->par_enc) >> 3;	// XOR 0x40 = second last bit flipped
@@ -678,17 +686,13 @@ static void Check_for_FilterFlipProperties(void)
 		}
 	}
 	
-	if (write_stats) {
+	if (write_stats)
 		fprintf(fstats, "%d;", num_bitflips);
-	}
 }
 
-
 static void simulate_MFplus_RNG(uint32_t test_cuid, uint64_t test_key, uint32_t *nt_enc, uint8_t *par_enc)
 {
 	struct Crypto1State sim_cs = {0, 0};
-//	sim_cs.odd = sim_cs.even = 0;
-
 	// init cryptostate with key:
 	for(int8_t i = 47; i > 0; i -= 2) {
 		sim_cs.odd  = sim_cs.odd  << 1 | BIT(test_key, (i - 1) ^ 7);
@@ -708,7 +712,6 @@ static void simulate_MFplus_RNG(uint32_t test_cuid, uint64_t test_key, uint32_t
 	
 }
 
-
 static void simulate_acquire_nonces()
 {
 	clock_t time1 = clock();
@@ -741,8 +744,8 @@ static void simulate_acquire_nonces()
 			num_good_first_bytes = estimate_second_byte_sum();
 			if (total_num_nonces > next_fivehundred) {
 				next_fivehundred = (total_num_nonces/500+1) * 500;
-				printf("Acquired %5d nonces (%5d with distinct bytes 0 and 1). Number of bytes with probability for correctly guessed Sum(a8) > %1.1f%%: %d\n",
-					total_num_nonces, 
+				printf("Acquired %5d nonces (%5d with distinct bytes 0,1). Bytes with probability for correctly guessed Sum(a8) > %1.1f%%: %d\n",
+					total_num_nonces,
 					total_added_nonces,
 					CONFIDENCE_THRESHOLD * 100.0,
 					num_good_first_bytes);
@@ -762,12 +765,10 @@ static void simulate_acquire_nonces()
 		
 }
 
-
 static int acquire_nonces(uint8_t blockNo, uint8_t keyType, uint8_t *key, uint8_t trgBlockNo, uint8_t trgKeyType, bool nonce_file_write, bool slow)
 {
 	clock_t time1 = clock();
 	bool initialize = true;
-	bool field_off = false;
 	bool finished = false;
 	bool filter_flip_checked = false;
 	uint32_t flags = 0;
@@ -775,126 +776,134 @@ static int acquire_nonces(uint8_t blockNo, uint8_t keyType, uint8_t *key, uint8_
 	uint32_t total_num_nonces = 0;
 	uint32_t next_fivehundred = 500;
 	uint32_t total_added_nonces = 0;
+	uint32_t idx = 1;
+	uint32_t timeout  = 0;
 	FILE *fnonces = NULL;
+	field_off = false;
 	UsbCommand resp;
-
+	UsbCommand c = {CMD_MIFARE_ACQUIRE_ENCRYPTED_NONCES, {0,0,0} };
+	memcpy(c.d.asBytes, key, 6);	
+	c.arg[0] = blockNo + (keyType * 0x100);
+	c.arg[1] = trgBlockNo + (trgKeyType * 0x100);
+		
 	printf("Acquiring nonces...\n");
-	
-	clearCommandBuffer();
-
 	do {
+	
 		flags = 0;
 		flags |= initialize ? 0x0001 : 0;
 		flags |= slow ? 0x0002 : 0;
 		flags |= field_off ? 0x0004 : 0;
-		UsbCommand c = {CMD_MIFARE_ACQUIRE_ENCRYPTED_NONCES, {blockNo + keyType * 0x100, trgBlockNo + trgKeyType * 0x100, flags}};
-		memcpy(c.d.asBytes, key, 6);
+		c.arg[2] = flags;
 
+		clearCommandBuffer();
 		SendCommand(&c);
 		
-		if (field_off) finished = true;
-		
-		if (initialize) {
-			if (!WaitForResponseTimeout(CMD_ACK, &resp, 3000)) return 1;
-			if (resp.arg[0]) return resp.arg[0];  // error during nested_hard
+		if (field_off) break;
+
+		while(!WaitForResponseTimeout(CMD_ACK, &resp, 2000)) {
+			timeout++;
+			printf(".");
+			if (timeout > 3) {
+				PrintAndLog("\nNo response from Proxmark. Aborting...");
+				if (fnonces) fclose(fnonces);
+				return 1;
+			}
+		}		
 
+		if (resp.arg[0]) {
+			if (fnonces) fclose(fnonces);
+			return resp.arg[0];  // error during nested_hard
+		}
+			
+		if (initialize) {
+			// global var CUID
 			cuid = resp.arg[1];
-			// PrintAndLog("Acquiring nonces for CUID 0x%08x", cuid); 
 			if (nonce_file_write && fnonces == NULL) {
 				if ((fnonces = fopen("nonces.bin","wb")) == NULL) { 
 					PrintAndLog("Could not create file nonces.bin");
 					return 3;
 				}
 				PrintAndLog("Writing acquired nonces to binary file nonces.bin");
+				memset (write_buf, 0, sizeof (write_buf));
 				num_to_bytes(cuid, 4, write_buf);
 				fwrite(write_buf, 1, 4, fnonces);
 				fwrite(&trgBlockNo, 1, 1, fnonces);
 				fwrite(&trgKeyType, 1, 1, fnonces);
+				fflush(fnonces);
 			}
+			initialize = false;			
 		}
-
-		if (!initialize) {
-			uint32_t nt_enc1, nt_enc2;
-			uint8_t par_enc;
-			uint16_t num_acquired_nonces = resp.arg[2];
-			uint8_t *bufp = resp.d.asBytes;
-			for (uint16_t i = 0; i < num_acquired_nonces; i+=2) {
-				nt_enc1 = bytes_to_num(bufp, 4);
-				nt_enc2 = bytes_to_num(bufp+4, 4);
-				par_enc = bytes_to_num(bufp+8, 1);
-				
-				//printf("Encrypted nonce: %08x, encrypted_parity: %02x\n", nt_enc1, par_enc >> 4);
-				total_added_nonces += add_nonce(nt_enc1, par_enc >> 4);
-				//printf("Encrypted nonce: %08x, encrypted_parity: %02x\n", nt_enc2, par_enc & 0x0f);
-				total_added_nonces += add_nonce(nt_enc2, par_enc & 0x0f);
-				
-
-				if (nonce_file_write) {
-					fwrite(bufp, 1, 9, fnonces);
-				}
-				
-				bufp += 9;
+		
+		uint32_t nt_enc1, nt_enc2;
+		uint8_t par_enc;
+		uint16_t num_acquired_nonces = resp.arg[2];
+		uint8_t *bufp = resp.d.asBytes;
+		for (uint16_t i = 0; i < num_acquired_nonces; i += 2) {
+			nt_enc1 = bytes_to_num(bufp, 4);
+			nt_enc2 = bytes_to_num(bufp+4, 4);
+			par_enc = bytes_to_num(bufp+8, 1);
+			
+			total_added_nonces += add_nonce(nt_enc1, par_enc >> 4);
+			total_added_nonces += add_nonce(nt_enc2, par_enc & 0x0f);
+			
+			if (nonce_file_write && fnonces) {
+				fwrite(bufp, 1, 9, fnonces);
+				fflush(fnonces);
 			}
-
-			total_num_nonces += num_acquired_nonces;
+			bufp += 9;
 		}
-		
-		if (first_byte_num == 256 ) {
-			// printf("first_byte_num = %d, first_byte_Sum = %d\n", first_byte_num, first_byte_Sum);
+		total_num_nonces += num_acquired_nonces;
+
+		if (first_byte_num == 256) {
+
 			if (!filter_flip_checked) {
 				Check_for_FilterFlipProperties();
 				filter_flip_checked = true;
 			}
+
 			num_good_first_bytes = estimate_second_byte_sum();
+
 			if (total_num_nonces > next_fivehundred) {
 				next_fivehundred = (total_num_nonces/500+1) * 500;
-				printf("Acquired %5d nonces (%5d with distinct bytes 0 and 1). Number of bytes with probability for correctly guessed Sum(a8) > %1.1f%%: %d\n",
-					total_num_nonces, 
+				printf("Acquired %5d nonces (%5d/%5d with distinct bytes 0,1). Bytes with probability for correctly guessed Sum(a8) > %1.1f%%: %d\n",
+					total_num_nonces,
 					total_added_nonces,
+					NONCES_THRESHOLD * idx,
 					CONFIDENCE_THRESHOLD * 100.0,
-					num_good_first_bytes);
+					num_good_first_bytes
+					);				
 			}
-			if (num_good_first_bytes >= GOOD_BYTES_REQUIRED) {
-				field_off = true;	// switch off field with next SendCommand and then finish
-			}
-		}
-
-		if (!initialize) {
-			if (!WaitForResponseTimeout(CMD_ACK, &resp, 3000)) {
-				fclose(fnonces);
-				return 1;
-			}
-			if (resp.arg[0]) {
-				fclose(fnonces);
-				return resp.arg[0];  // error during nested_hard
+			
+			if (total_added_nonces >= (NONCES_THRESHOLD * idx)) {
+				if (num_good_first_bytes > 0) {
+					if (generate_candidates(first_byte_Sum, nonces[best_first_bytes[0]].Sum8_guess) || known_target_key != -1) {
+						field_off = brute_force(); // switch off field with next SendCommand and then finish
+					}
+				}
+				idx++;
 			}
 		}
-
-		initialize = false;
-
 	} while (!finished);
 
-	
-	if (nonce_file_write) {
+	if (nonce_file_write && fnonces)
 		fclose(fnonces);
-	}
 	
 	time1 = clock() - time1;
 	if ( time1 > 0 ) {
-	PrintAndLog("Acquired a total of %d nonces in %1.1f seconds (%0.0f nonces/minute)", 
-		total_num_nonces, 
-		((float)time1)/CLOCKS_PER_SEC, 
-		total_num_nonces * 60.0 * CLOCKS_PER_SEC/(float)time1
+		PrintAndLog("Acquired a total of %d nonces in %1.1f seconds (%0.0f nonces/minute)", 
+			total_num_nonces, 
+			((float)time1)/CLOCKS_PER_SEC, 
+			total_num_nonces * 60.0 * CLOCKS_PER_SEC/(float)time1
 		);
 	}
 	return 0;
 }
 
-
 static int init_partial_statelists(void)
 {
 	const uint32_t sizes_odd[17] = { 126757, 0, 18387, 0, 74241, 0, 181737, 0, 248801, 0, 182033, 0, 73421, 0, 17607, 0, 125601 };
-	const uint32_t sizes_even[17] = { 125723, 0, 17867, 0, 74305, 0, 178707, 0, 248801, 0, 185063, 0, 73356, 0, 18127, 0, 126634 };
+//	const uint32_t sizes_even[17] = { 125723, 0, 17867, 0, 74305, 0, 178707, 0, 248801, 0, 185063, 0, 73356, 0, 18127, 0, 126634 };
+	const uint32_t sizes_even[17] = { 125723, 0, 17867, 0, 74305, 0, 178707, 0, 248801, 0, 185063, 0, 73357, 0, 18127, 0, 126635 };
 	
 	printf("Allocating memory for partial statelists...\n");
 	for (odd_even_t odd_even = EVEN_STATE; odd_even <= ODD_STATE; odd_even++) {
@@ -934,13 +943,12 @@ static int init_partial_statelists(void)
 		for (uint16_t i = 0; i <= 16; i += 2) {
 			uint32_t *p = partial_statelist[i].states[odd_even];
 			p += partial_statelist[i].len[odd_even];
-			*p = 0xffffffff;
+			*p = END_OF_LIST_MARKER;
 		}
 	}
 	
 	return 0;
 }	
-		
 
 static void init_BitFlip_statelist(void)
 {
@@ -961,10 +969,9 @@ static void init_BitFlip_statelist(void)
 	}
 	// set len and add End Of List marker
 	statelist_bitflip.len[0] = p - statelist_bitflip.states[0];
-	*p = 0xffffffff;
-	statelist_bitflip.states[0] = realloc(statelist_bitflip.states[0], sizeof(uint32_t) * (statelist_bitflip.len[0] + 1));
+	*p = END_OF_LIST_MARKER;
+	//statelist_bitflip.states[0] = realloc(statelist_bitflip.states[0], sizeof(uint32_t) * (statelist_bitflip.len[0] + 1));
 }
-
 		
 static inline uint32_t *find_first_state(uint32_t state, uint32_t mask, partial_indexed_statelist_t *sl, odd_even_t odd_even)
 {
@@ -972,12 +979,11 @@ static inline uint32_t *find_first_state(uint32_t state, uint32_t mask, partial_
 
 	if (p == NULL) return NULL;
 	while (*p < (state & mask)) p++;
-	if (*p == 0xffffffff) return NULL;					// reached end of list, no match
+	if (*p == END_OF_LIST_MARKER) return NULL;					// reached end of list, no match
 	if ((*p & mask) == (state & mask)) return p;		// found a match.
 	return NULL;										// no match
 } 
 
-
 static inline bool /*__attribute__((always_inline))*/ invariant_holds(uint_fast8_t byte_diff, uint_fast32_t state1, uint_fast32_t state2, uint_fast8_t bit, uint_fast8_t state_bit)
 {
 	uint_fast8_t j_1_bit_mask = 0x01 << (bit-1);
@@ -989,7 +995,6 @@ static inline bool /*__attribute__((always_inline))*/ invariant_holds(uint_fast8
 	return !all_diff;
 }
 
-
 static inline bool /*__attribute__((always_inline))*/ invalid_state(uint_fast8_t byte_diff, uint_fast32_t state1, uint_fast32_t state2, uint_fast8_t bit, uint_fast8_t state_bit)
 {
 	uint_fast8_t j_bit_mask = 0x01 << bit;
@@ -1000,7 +1005,6 @@ static inline bool /*__attribute__((always_inline))*/ invalid_state(uint_fast8_t
 	return all_diff;
 }
 
-
 static inline bool remaining_bits_match(uint_fast8_t num_common_bits, uint_fast8_t byte_diff, uint_fast32_t state1, uint_fast32_t state2, odd_even_t odd_even)
 {
 	if (odd_even) {
@@ -1031,7 +1035,6 @@ static inline bool remaining_bits_match(uint_fast8_t num_common_bits, uint_fast8
 	return true;					// valid state
 }
 
-
 static bool all_other_first_bytes_match(uint32_t state, odd_even_t odd_even) 
 {
 	for (uint16_t i = 1; i < num_good_first_bytes; i++) {
@@ -1054,7 +1057,7 @@ static bool all_other_first_bytes_match(uint32_t state, odd_even_t odd_even)
 					uint16_t part_sum_a8 = (odd_even == ODD_STATE) ? r : s;
 					uint32_t *p = find_first_state(state, mask, &partial_statelist[part_sum_a8], odd_even);
 					if (p != NULL) {
-						while ((state & mask) == (*p & mask) && (*p != 0xffffffff)) {
+						while ((state & mask) == (*p & mask) && (*p != END_OF_LIST_MARKER)) {
 							if (remaining_bits_match(j, bytes_diff, state, (state&0x00fffff0) | *p, odd_even)) {
 								found_match = true;
 								// if ((odd_even == ODD_STATE && state == test_state_odd)
@@ -1095,7 +1098,6 @@ static bool all_other_first_bytes_match(uint32_t state, odd_even_t odd_even)
 	return true;
 }
 
-
 static bool all_bit_flips_match(uint32_t state, odd_even_t odd_even) 
 {
 	for (uint16_t i = 0; i < 256; i++) {
@@ -1113,7 +1115,7 @@ static bool all_bit_flips_match(uint32_t state, odd_even_t odd_even)
 			bool found_match = false;
 			uint32_t *p = find_first_state(state, mask, &statelist_bitflip, 0);
 			if (p != NULL) {
-				while ((state & mask) == (*p & mask) && (*p != 0xffffffff)) {
+				while ((state & mask) == (*p & mask) && (*p != END_OF_LIST_MARKER)) {
 					if (remaining_bits_match(j, bytes_diff, state, (state&0x00fffff0) | *p, odd_even)) {
 						found_match = true;
 						// if ((odd_even == ODD_STATE && state == test_state_odd)
@@ -1152,13 +1154,11 @@ static bool all_bit_flips_match(uint32_t state, odd_even_t odd_even)
 	return true;
 }
 
-
 static struct sl_cache_entry {
 	uint32_t *sl;
 	uint32_t len;
 	} sl_cache[17][17][2];
 
-
 static void init_statelist_cache(void)
 {
 	for (uint16_t i = 0; i < 17; i+=2) {
@@ -1171,7 +1171,6 @@ static void init_statelist_cache(void)
 	}		
 }
 
-
 static int add_matching_states(statelist_t *candidates, uint16_t part_sum_a0, uint16_t part_sum_a8, odd_even_t odd_even)
 {
 	uint32_t worstcase_size = 1<<20;
@@ -1189,11 +1188,11 @@ static int add_matching_states(statelist_t *candidates, uint16_t part_sum_a0, ui
 		return 4;
 	}
 	uint32_t *add_p = candidates->states[odd_even]; 
-	for (uint32_t *p1 = partial_statelist[part_sum_a0].states[odd_even]; *p1 != 0xffffffff; p1++) {
+	for (uint32_t *p1 = partial_statelist[part_sum_a0].states[odd_even]; *p1 != END_OF_LIST_MARKER; p1++) {
 		uint32_t search_mask = 0x000ffff0;
 		uint32_t *p2 = find_first_state((*p1 << 4), search_mask, &partial_statelist[part_sum_a8], odd_even);
-		if (p2 != NULL) {
-			while (((*p1 << 4) & search_mask) == (*p2 & search_mask) && *p2 != 0xffffffff) {
+		if (p1 != NULL && p2 != NULL) {
+			while (((*p1 << 4) & search_mask) == (*p2 & search_mask) && *p2 != END_OF_LIST_MARKER) {
 				if ((nonces[best_first_bytes[0]].BitFlip[odd_even] && find_first_state((*p1 << 4) | *p2, 0x000fffff, &statelist_bitflip, 0))
 					|| !nonces[best_first_bytes[0]].BitFlip[odd_even]) {
 				if (all_other_first_bytes_match((*p1 << 4) | *p2, odd_even)) {
@@ -1208,7 +1207,7 @@ static int add_matching_states(statelist_t *candidates, uint16_t part_sum_a0, ui
 	}
 
 	// set end of list marker and len
-	*add_p = 0xffffffff; 
+	*add_p = END_OF_LIST_MARKER; 
 	candidates->len[odd_even] = add_p - candidates->states[odd_even];
 
 	candidates->states[odd_even] = realloc(candidates->states[odd_even], sizeof(uint32_t) * (candidates->len[odd_even] + 1));
@@ -1219,7 +1218,6 @@ static int add_matching_states(statelist_t *candidates, uint16_t part_sum_a0, ui
 	return 0;
 }
 
-
 static statelist_t *add_more_candidates(statelist_t *current_candidates)
 {
 	statelist_t *new_candidates = NULL;
@@ -1231,6 +1229,8 @@ static statelist_t *add_more_candidates(statelist_t *current_candidates)
 	} else {
 		new_candidates = current_candidates->next = (statelist_t *)malloc(sizeof(statelist_t));
 	}
+	if (!new_candidates) return NULL;
+
 	new_candidates->next = NULL;
 	new_candidates->len[ODD_STATE] = 0;
 	new_candidates->len[EVEN_STATE] = 0;
@@ -1239,8 +1239,7 @@ static statelist_t *add_more_candidates(statelist_t *current_candidates)
 	return new_candidates;
 }
 
-
-static void TestIfKeyExists(uint64_t key)
+static bool TestIfKeyExists(uint64_t key)
 {
 	struct Crypto1State *pcs;
 	pcs = crypto1_create(key);
@@ -1249,49 +1248,53 @@ static void TestIfKeyExists(uint64_t key)
 	uint32_t state_odd = pcs->odd & 0x00ffffff;
 	uint32_t state_even = pcs->even & 0x00ffffff;
 	//printf("Tests: searching for key %llx after first byte 0x%02x (state_odd = 0x%06x, state_even = 0x%06x) ...\n", key, best_first_bytes[0], state_odd, state_even);
-	
+	printf("Validating key search space\n");
 	uint64_t count = 0;
 	for (statelist_t *p = candidates; p != NULL; p = p->next) {
 		bool found_odd = false;
 		bool found_even = false;
 		uint32_t *p_odd = p->states[ODD_STATE];
 		uint32_t *p_even = p->states[EVEN_STATE];
-		while (*p_odd != 0xffffffff) {
+		while (*p_odd != END_OF_LIST_MARKER) {
 			if ((*p_odd & 0x00ffffff) == state_odd) {
 				found_odd = true;
 				break;
 			}
 			p_odd++;
 		}
-		while (*p_even != 0xffffffff) {
-			if ((*p_even & 0x00ffffff) == state_even) {
+		while (*p_even != END_OF_LIST_MARKER) {
+			if ((*p_even & 0x00ffffff) == state_even)
 				found_even = true;
-			}
+
 			p_even++;
 		}
 		count += (p_odd - p->states[ODD_STATE]) * (p_even - p->states[EVEN_STATE]);
 		if (found_odd && found_even) {
-			PrintAndLog("Key Found after testing %lld (2^%1.1f) out of %lld (2^%1.1f) keys. A brute force would have taken approx %lld minutes.", 
-				count, log(count)/log(2), 
-				maximum_states, log(maximum_states)/log(2),
-				(count>>23)/60);
-			if (write_stats) {
-				fprintf(fstats, "1\n");
+			if (known_target_key != -1) {
+				PrintAndLog("Key Found after testing %llu (2^%1.1f) out of %lld (2^%1.1f) keys.", 
+					count,
+					log(count)/log(2), 
+					maximum_states,
+					log(maximum_states)/log(2)
+					);
+				if (write_stats)
+					fprintf(fstats, "1\n");			
 			}
 			crypto1_destroy(pcs);
-			return;
+			return true;
 		}
 	}
 
-	printf("Key NOT found!\n");
-	if (write_stats) {
-		fprintf(fstats, "0\n");
+	if (known_target_key != -1) {
+		printf("Key NOT found!\n");
+		if (write_stats)
+			fprintf(fstats, "0\n");
 	}
 	crypto1_destroy(pcs);
+	return false;
 }
 
-	
-static void generate_candidates(uint16_t sum_a0, uint16_t sum_a8)
+static bool generate_candidates(uint16_t sum_a0, uint16_t sum_a8)
 {
 	printf("Generating crypto1 state candidates... \n");
 	
@@ -1305,30 +1308,34 @@ static void generate_candidates(uint16_t sum_a0, uint16_t sum_a8)
 			}
 		}
 	}
-	printf("Number of possible keys with Sum(a0) = %d: %"PRIu64" (2^%1.1f)\n", sum_a0, maximum_states, log(maximum_states)/log(2.0));
+
+	if (maximum_states == 0) return false; // prevent keyspace reduction error (2^-inf)
+
+	printf("Number of possible keys with Sum(a0) = %d: %"PRIu64" (2^%1.1f)\n", sum_a0, maximum_states, log(maximum_states)/log(2));
 	
 	init_statelist_cache();
 	
 	for (uint16_t p = 0; p <= 16; p += 2) {
 		for (uint16_t q = 0; q <= 16; q += 2) {
 			if (p*(16-q) + (16-p)*q == sum_a0) {
-				printf("Reducing Partial Statelists (p,q) = (%d,%d) with lengths %d, %d\n", 
-						p, q, partial_statelist[p].len[ODD_STATE], partial_statelist[q].len[EVEN_STATE]);
+				// printf("Reducing Partial Statelists (p,q) = (%d,%d) with lengths %d, %d\n", 
+						// p, q, partial_statelist[p].len[ODD_STATE], partial_statelist[q].len[EVEN_STATE]);
 				for (uint16_t r = 0; r <= 16; r += 2) {
 					for (uint16_t s = 0; s <= 16; s += 2) {
 						if (r*(16-s) + (16-r)*s == sum_a8) {
 							current_candidates = add_more_candidates(current_candidates);
+							if (current_candidates != NULL) {
 							// check for the smallest partial statelist. Try this first - it might give 0 candidates 
 							// and eliminate the need to calculate the other part
 							if (MIN(partial_statelist[p].len[ODD_STATE], partial_statelist[r].len[ODD_STATE]) 
 									< MIN(partial_statelist[q].len[EVEN_STATE], partial_statelist[s].len[EVEN_STATE])) { 
-							add_matching_states(current_candidates, p, r, ODD_STATE);
+								add_matching_states(current_candidates, p, r, ODD_STATE);
 								if(current_candidates->len[ODD_STATE]) {
-							add_matching_states(current_candidates, q, s, EVEN_STATE);
+									add_matching_states(current_candidates, q, s, EVEN_STATE);
 								} else {
 									current_candidates->len[EVEN_STATE] = 0;
 									uint32_t *p = current_candidates->states[EVEN_STATE] = malloc(sizeof(uint32_t));
-									*p = 0xffffffff;
+									*p = END_OF_LIST_MARKER;
 								}
 							} else {
 								add_matching_states(current_candidates, q, s, EVEN_STATE);
@@ -1337,35 +1344,38 @@ static void generate_candidates(uint16_t sum_a0, uint16_t sum_a8)
 								} else {
 									current_candidates->len[ODD_STATE] = 0;
 									uint32_t *p = current_candidates->states[ODD_STATE] = malloc(sizeof(uint32_t));
-									*p = 0xffffffff;
+									*p = END_OF_LIST_MARKER;
 								}
 							}
-							printf("Odd  state candidates: %6d (2^%0.1f)\n", current_candidates->len[ODD_STATE], log(current_candidates->len[ODD_STATE])/log(2)); 
-							printf("Even state candidates: %6d (2^%0.1f)\n", current_candidates->len[EVEN_STATE], log(current_candidates->len[EVEN_STATE])/log(2)); 
+							//printf("Odd  state candidates: %6d (2^%0.1f)\n", current_candidates->len[ODD_STATE], log(current_candidates->len[ODD_STATE])/log(2)); 
+							//printf("Even state candidates: %6d (2^%0.1f)\n", current_candidates->len[EVEN_STATE], log(current_candidates->len[EVEN_STATE])/log(2)); 
 						}
 					}
 				}
 			}
 		}
 	}					
+	}					
 
-	
 	maximum_states = 0;
-	for (statelist_t *sl = candidates; sl != NULL; sl = sl->next) {
+	unsigned int n = 0;
+	for (statelist_t *sl = candidates; sl != NULL && n < MAX_BUCKETS; sl = sl->next, n++) {
 		maximum_states += (uint64_t)sl->len[ODD_STATE] * sl->len[EVEN_STATE];
 	}
-	printf("Number of remaining possible keys: %"PRIu64" (2^%1.1f)\n", maximum_states, log(maximum_states)/log(2.0));
+
+	if (maximum_states == 0) return false; // prevent keyspace reduction error (2^-inf)
+
+	float kcalc = log(maximum_states)/log(2);
+	printf("Number of remaining possible keys: %"PRIu64" (2^%1.1f)\n", maximum_states, kcalc);
 	if (write_stats) {
-		if (maximum_states != 0) {
-			fprintf(fstats, "%1.1f;", log(maximum_states)/log(2.0));
-		} else {
-			fprintf(fstats, "%1.1f;", 0.0);
-		}
+		fprintf(fstats, "%1.1f;", (kcalc != 0) ? kcalc : 0.0);
 	}
-}
+	if (kcalc < CRACKING_THRESHOLD) return true;
 
+	return false;
+}
 
-static void	free_candidates_memory(statelist_t *sl)
+static void free_candidates_memory(statelist_t *sl)
 {
 	if (sl == NULL) {
 		return;
@@ -1375,7 +1385,6 @@ static void	free_candidates_memory(statelist_t *sl)
 	}
 }
 
-
 static void free_statelist_cache(void)
 {
 	for (uint16_t i = 0; i < 17; i+=2) {
@@ -1387,25 +1396,347 @@ static void free_statelist_cache(void)
 	}		
 }
 
+static const uint64_t crack_states_bitsliced(statelist_t *p){
+    // the idea to roll back the half-states before combining them was suggested/explained to me by bla
+    // first we pre-bitslice all the even state bits and roll them back, then bitslice the odd bits and combine the two in the inner loop
+    uint64_t key = -1;
+	uint8_t bSize = sizeof(bitslice_t);
+
+#ifdef EXACT_COUNT
+    size_t bucket_states_tested = 0;
+    size_t bucket_size[p->len[EVEN_STATE]/MAX_BITSLICES];
+#else
+    const size_t bucket_states_tested = (p->len[EVEN_STATE])*(p->len[ODD_STATE]);
+#endif
+
+    bitslice_t *bitsliced_even_states[p->len[EVEN_STATE]/MAX_BITSLICES];
+    size_t bitsliced_blocks = 0;
+    uint32_t const * restrict even_end = p->states[EVEN_STATE]+p->len[EVEN_STATE];
+	
+    // bitslice all the even states
+    for(uint32_t * restrict p_even = p->states[EVEN_STATE]; p_even < even_end; p_even += MAX_BITSLICES){
+
+#ifdef __WIN32
+	#ifdef __MINGW32__
+		bitslice_t * restrict lstate_p = __mingw_aligned_malloc((STATE_SIZE+ROLLBACK_SIZE) * bSize, bSize);
+	#else		
+		bitslice_t * restrict lstate_p = _aligned_malloc((STATE_SIZE+ROLLBACK_SIZE) * bSize, bSize);
+	#endif
+#else
+	#ifdef __APPLE__
+		bitslice_t * restrict lstate_p = malloc((STATE_SIZE+ROLLBACK_SIZE) * bSize);
+	#else
+		bitslice_t * restrict lstate_p = memalign(bSize, (STATE_SIZE+ROLLBACK_SIZE) * bSize);
+	#endif
+#endif
+
+		if ( !lstate_p )	{
+			__sync_fetch_and_add(&total_states_tested, bucket_states_tested);
+			return key;
+		}
+				
+		memset(lstate_p+1, 0x0, (STATE_SIZE-1)*sizeof(bitslice_t)); // zero even bits
+		
+		// bitslice even half-states
+        const size_t max_slices = (even_end-p_even) < MAX_BITSLICES ? even_end-p_even : MAX_BITSLICES;
+#ifdef EXACT_COUNT
+        bucket_size[bitsliced_blocks] = max_slices;
+#endif
+        for(size_t slice_idx = 0; slice_idx < max_slices; ++slice_idx){
+            uint32_t e = *(p_even+slice_idx);
+            for(size_t bit_idx = 1; bit_idx < STATE_SIZE; bit_idx+=2, e >>= 1){
+                // set even bits
+                if(e&1){
+                    lstate_p[bit_idx].bytes64[slice_idx>>6] |= 1ull << (slice_idx&63);
+                }
+            }
+        }
+        // compute the rollback bits
+        for(size_t rollback = 0; rollback < ROLLBACK_SIZE; ++rollback){
+            // inlined crypto1_bs_lfsr_rollback
+            const bitslice_value_t feedout = lstate_p[0].value;
+            ++lstate_p;
+            const bitslice_value_t ks_bits = crypto1_bs_f20(lstate_p);
+            const bitslice_value_t feedback = (feedout ^ ks_bits     ^ lstate_p[47- 5].value ^ lstate_p[47- 9].value ^
+                                               lstate_p[47-10].value ^ lstate_p[47-12].value ^ lstate_p[47-14].value ^
+                                               lstate_p[47-15].value ^ lstate_p[47-17].value ^ lstate_p[47-19].value ^
+                                               lstate_p[47-24].value ^ lstate_p[47-25].value ^ lstate_p[47-27].value ^
+                                               lstate_p[47-29].value ^ lstate_p[47-35].value ^ lstate_p[47-39].value ^
+                                               lstate_p[47-41].value ^ lstate_p[47-42].value ^ lstate_p[47-43].value);
+            lstate_p[47].value = feedback ^ bitsliced_rollback_byte[rollback].value;
+        }
+        bitsliced_even_states[bitsliced_blocks++] = lstate_p;
+    }
+
+    // bitslice every odd state to every block of even half-states with half-finished rollback
+    for(uint32_t const * restrict p_odd = p->states[ODD_STATE]; p_odd < p->states[ODD_STATE]+p->len[ODD_STATE]; ++p_odd){
+        // early abort
+        if(keys_found){
+            goto out;
+        }
+
+        // set the odd bits and compute rollback
+        uint64_t o = (uint64_t) *p_odd;
+        lfsr_rollback_byte((struct Crypto1State*) &o, 0, 1);
+        // pre-compute part of the odd feedback bits (minus rollback)
+        bool odd_feedback_bit = parity(o&0x9ce5c);
+
+        crypto1_bs_rewind_a0();
+        // set odd bits
+        for(size_t state_idx = 0; state_idx < STATE_SIZE-ROLLBACK_SIZE; o >>= 1, state_idx+=2){
+            state_p[state_idx] = (o & 1) ? bs_ones : bs_zeroes;
+        }
+        const bitslice_value_t odd_feedback = odd_feedback_bit ? bs_ones.value : bs_zeroes.value;
+
+        for(size_t block_idx = 0; block_idx < bitsliced_blocks; ++block_idx){
+            const bitslice_t * const restrict bitsliced_even_state = bitsliced_even_states[block_idx];
+            size_t state_idx;
+            // set even bits
+            for(state_idx = 0; state_idx < STATE_SIZE-ROLLBACK_SIZE; state_idx+=2){
+                state_p[1+state_idx] = bitsliced_even_state[1+state_idx];
+            }
+            // set rollback bits
+            uint64_t lo = o;
+            for(; state_idx < STATE_SIZE; lo >>= 1, state_idx+=2){
+                // set the odd bits and take in the odd rollback bits from the even states
+                if(lo & 1){
+                    state_p[state_idx].value = ~bitsliced_even_state[state_idx].value;
+                } else {
+                    state_p[state_idx] = bitsliced_even_state[state_idx];
+                }
+
+                // set the even bits and take in the even rollback bits from the odd states
+                if((lo >> 32) & 1){
+                    state_p[1+state_idx].value = ~bitsliced_even_state[1+state_idx].value;
+                } else {
+                    state_p[1+state_idx] = bitsliced_even_state[1+state_idx];
+                }
+            }
+
+#ifdef EXACT_COUNT
+            bucket_states_tested += (bucket_size[block_idx] > MAX_BITSLICES) ? MAX_BITSLICES : bucket_size[block_idx];
+#endif
+            // pre-compute first keystream and feedback bit vectors
+            const bitslice_value_t ksb = crypto1_bs_f20(state_p);
+            const bitslice_value_t fbb = (odd_feedback         ^ state_p[47- 0].value ^ state_p[47- 5].value ^ // take in the even and rollback bits
+                                          state_p[47-10].value ^ state_p[47-12].value ^ state_p[47-14].value ^
+                                          state_p[47-24].value ^ state_p[47-42].value);
+
+            // vector to contain test results (1 = passed, 0 = failed)
+            bitslice_t results = bs_ones;
+
+            for(size_t tests = 0; tests < NONCE_TESTS; ++tests){
+                size_t parity_bit_idx = 0;
+                bitslice_value_t fb_bits = fbb;
+                bitslice_value_t ks_bits = ksb;
+                state_p = &states[KEYSTREAM_SIZE-1];
+                bitslice_value_t parity_bit_vector = bs_zeroes.value;
+
+                // highest bit is transmitted/received first
+                for(int32_t ks_idx = KEYSTREAM_SIZE-1; ks_idx >= 0; --ks_idx, --state_p){
+                    // decrypt nonce bits
+                    const bitslice_value_t encrypted_nonce_bit_vector = bitsliced_encrypted_nonces[tests][ks_idx].value;
+                    const bitslice_value_t decrypted_nonce_bit_vector = (encrypted_nonce_bit_vector ^ ks_bits);
+
+                    // compute real parity bits on the fly
+                    parity_bit_vector ^= decrypted_nonce_bit_vector;
+
+                    // update state
+                    state_p[0].value = (fb_bits ^ decrypted_nonce_bit_vector);
+
+                    // compute next keystream bit
+                    ks_bits = crypto1_bs_f20(state_p);
+
+                    // for each byte:
+                    if((ks_idx&7) == 0){
+                        // get encrypted parity bits
+                        const bitslice_value_t encrypted_parity_bit_vector = bitsliced_encrypted_parity_bits[tests][parity_bit_idx++].value;
+
+                        // decrypt parity bits
+                        const bitslice_value_t decrypted_parity_bit_vector = (encrypted_parity_bit_vector ^ ks_bits);
+
+                        // compare actual parity bits with decrypted parity bits and take count in results vector
+                        results.value &= (parity_bit_vector ^ decrypted_parity_bit_vector);
+
+                        // make sure we still have a match in our set
+                        // if(memcmp(&results, &bs_zeroes, sizeof(bitslice_t)) == 0){
+
+                        // this is much faster on my gcc, because somehow a memcmp needlessly spills/fills all the xmm registers to/from the stack - ???
+                        // the short-circuiting also helps
+                        if(results.bytes64[0] == 0
+#if MAX_BITSLICES > 64
+                           && results.bytes64[1] == 0
+#endif
+#if MAX_BITSLICES > 128
+                           && results.bytes64[2] == 0
+                           && results.bytes64[3] == 0
+#endif
+                          ){
+                            goto stop_tests;
+                        }
+                        // this is about as fast but less portable (requires -std=gnu99)
+                        // asm goto ("ptest %1, %0\n\t"
+                        //           "jz %l2" :: "xm" (results.value), "xm" (bs_ones.value) : "cc" : stop_tests);
+                        parity_bit_vector = bs_zeroes.value;
+                    }
+                    // compute next feedback bit vector
+                    fb_bits = (state_p[47- 0].value ^ state_p[47- 5].value ^ state_p[47- 9].value ^
+                               state_p[47-10].value ^ state_p[47-12].value ^ state_p[47-14].value ^
+                               state_p[47-15].value ^ state_p[47-17].value ^ state_p[47-19].value ^
+                               state_p[47-24].value ^ state_p[47-25].value ^ state_p[47-27].value ^
+                               state_p[47-29].value ^ state_p[47-35].value ^ state_p[47-39].value ^
+                               state_p[47-41].value ^ state_p[47-42].value ^ state_p[47-43].value);
+                }
+            }
+            // all nonce tests were successful: we've found the key in this block!
+            state_t keys[MAX_BITSLICES];
+            crypto1_bs_convert_states(&states[KEYSTREAM_SIZE], keys);
+            for(size_t results_idx = 0; results_idx < MAX_BITSLICES; ++results_idx){
+                if(get_vector_bit(results_idx, results)){
+                    key = keys[results_idx].value;
+                    goto out;
+                }
+            }
+stop_tests:
+            // prepare to set new states
+            crypto1_bs_rewind_a0();
+            continue;
+        }
+    }
+
+out:
+    for(size_t block_idx = 0; block_idx < bitsliced_blocks; ++block_idx){
+		
+#ifdef __WIN32
+	#ifdef __MINGW32__
+		__mingw_aligned_free(bitsliced_even_states[block_idx]-ROLLBACK_SIZE);
+	#else
+		_aligned_free(bitsliced_even_states[block_idx]-ROLLBACK_SIZE);		
+	#endif
+#else
+		free(bitsliced_even_states[block_idx]-ROLLBACK_SIZE);
+#endif		
+		
+    }
+    __sync_fetch_and_add(&total_states_tested, bucket_states_tested);
+    return key;
+}
 
-static void brute_force(void)
-{
+static void* crack_states_thread(void* x){
+    const size_t thread_id = (size_t)x;
+    size_t current_bucket = thread_id;
+	statelist_t *bucket = NULL;
+
+    while(current_bucket < bucket_count){
+		if (keys_found) break;
+
+		if ((bucket = buckets[current_bucket])) {
+            const uint64_t key = crack_states_bitsliced(bucket);
+
+			if (keys_found) break;
+			else if(key != -1) {
+				if (TestIfKeyExists(key)) {
+                __sync_fetch_and_add(&keys_found, 1);
+				__sync_fetch_and_add(&foundkey, key);
+					printf("*");
+					fflush(stdout);
+                break;
+				}
+				printf("!");
+				fflush(stdout);
+            } else {				
+                printf(".");
+				fflush(stdout);
+            }
+        }
+        current_bucket += thread_count;
+    }
+    return NULL;
+}
+
+static bool brute_force(void) {
+	bool ret = false;
 	if (known_target_key != -1) {
 		PrintAndLog("Looking for known target key in remaining key space...");
-		TestIfKeyExists(known_target_key);
+		ret = TestIfKeyExists(known_target_key);
 	} else {
-		PrintAndLog("Brute Force phase is not implemented.");
-	}
+		if (maximum_states == 0) return false; // prevent keyspace reduction error (2^-inf)
+
+	 	PrintAndLog("Brute force phase starting.");
+
+		clock_t time1 = clock();	 	
+		keys_found = 0;
+		foundkey = 0;
+
+		crypto1_bs_init();
+		memset (bitsliced_rollback_byte, 0, sizeof (bitsliced_rollback_byte));
+		memset (bitsliced_encrypted_nonces, 0, sizeof (bitsliced_encrypted_nonces));
+		memset (bitsliced_encrypted_parity_bits, 0, sizeof (bitsliced_encrypted_parity_bits));
+
+		PrintAndLog("Using %u-bit bitslices", MAX_BITSLICES);
+		PrintAndLog("Bitslicing best_first_byte^uid[3] (rollback byte): %02X ...", best_first_bytes[0]^(cuid>>24));
+		// convert to 32 bit little-endian
+		crypto1_bs_bitslice_value32((best_first_bytes[0]<<24)^cuid, bitsliced_rollback_byte, 8);
+
+		PrintAndLog("Bitslicing nonces...");
+		for(size_t tests = 0; tests < NONCE_TESTS; tests++){
+			uint32_t test_nonce = brute_force_nonces[tests]->nonce_enc;
+			uint8_t test_parity = brute_force_nonces[tests]->par_enc;
+			// pre-xor the uid into the decrypted nonces, and also pre-xor the cuid parity into the encrypted parity bits - otherwise an exta xor is required in the decryption routine
+			crypto1_bs_bitslice_value32(cuid^test_nonce, bitsliced_encrypted_nonces[tests], 32);
+			// convert to 32 bit little-endian
+			crypto1_bs_bitslice_value32(rev32( ~(test_parity ^ ~(parity(cuid>>24 & 0xff)<<3 | parity(cuid>>16 & 0xff)<<2 | parity(cuid>>8 & 0xff)<<1 | parity(cuid&0xff)))), bitsliced_encrypted_parity_bits[tests], 4);
+		}
+		total_states_tested = 0;
+
+		// count number of states to go
+		bucket_count = 0;
+		buckets[MAX_BUCKETS-1] = NULL;
+		for (statelist_t *p = candidates; p != NULL && bucket_count < MAX_BUCKETS; p = p->next) {
+			buckets[bucket_count] = p;
+			bucket_count++;
+		}
+		if (bucket_count < MAX_BUCKETS) buckets[bucket_count] = NULL;
 
-}
+#ifndef __WIN32
+		thread_count = sysconf(_SC_NPROCESSORS_CONF);
+		if ( thread_count < 1)
+			thread_count = 1;
+#endif  /* _WIN32 */
+
+		pthread_t threads[thread_count];
+
+		// enumerate states using all hardware threads, each thread handles one bucket
+		PrintAndLog("Starting %u cracking threads to search %u buckets containing a total of %"PRIu64" states...", thread_count, bucket_count, maximum_states);
+
+		for(size_t i = 0; i < thread_count; i++){
+			pthread_create(&threads[i], NULL, crack_states_thread, (void*) i);
+		}
+		for(size_t i = 0; i < thread_count; i++){
+			pthread_join(threads[i], 0);
+		}
 
+		time1 = clock() - time1;
+		PrintAndLog("\nTime for bruteforce %0.1f seconds.",((float)time1)/CLOCKS_PER_SEC);		
+		
+		if (keys_found) {
+			PrintAndLog("\nFound key: %012"PRIx64"\n", foundkey);
+			ret = true;
+		} 
+		// reset this counter for the next call
+		nonces_to_bruteforce = 0;
+	}
+	return ret;
+}
 
-int mfnestedhard(uint8_t blockNo, uint8_t keyType, uint8_t *key, uint8_t trgBlockNo, uint8_t trgKeyType, uint8_t *trgkey, bool nonce_file_read, bool nonce_file_write, bool slow, int tests) 
+int mfnestedhard(uint8_t blockNo, uint8_t keyType, uint8_t *key, uint8_t trgBlockNo, uint8_t trgKeyType, uint8_t *trgkey, bool nonce_file_read, bool nonce_file_write, bool slow, int tests, uint64_t *found_key) 
 {
 	// initialize Random number generator
 	time_t t;
 	srand((unsigned) time(&t));
 	
+	*found_key = 0;
+	
 	if (trgkey != NULL) {
 		known_target_key = bytes_to_num(trgkey, 6);
 	} else {
@@ -1438,51 +1769,39 @@ int mfnestedhard(uint8_t blockNo, uint8_t keyType, uint8_t *key, uint8_t trgBloc
 			candidates = NULL;
 		}
 		fclose(fstats);
+		fstats = NULL;
 	} else {
 		init_nonce_memory();
-		if (nonce_file_read) {  	// use pre-acquired data from file nonces.bin
+		if (nonce_file_read) { // use pre-acquired data from file nonces.bin
 			if (read_nonce_file() != 0) {
 				return 3;
 			}
 			Check_for_FilterFlipProperties();
 			num_good_first_bytes = MIN(estimate_second_byte_sum(), GOOD_BYTES_REQUIRED);
-		} else {					// acquire nonces.
+			PrintAndLog("Number of first bytes with confidence > %2.1f%%: %d", CONFIDENCE_THRESHOLD*100.0, num_good_first_bytes);
+
+			bool cracking = generate_candidates(first_byte_Sum, nonces[best_first_bytes[0]].Sum8_guess);
+			if (cracking || known_target_key != -1) {
+				brute_force();
+			}
+
+		} else { // acquire nonces.
 			uint16_t is_OK = acquire_nonces(blockNo, keyType, key, trgBlockNo, trgKeyType, nonce_file_write, slow);
 			if (is_OK != 0) {
+				free_nonces_memory();
+				//free_statelist_cache();
+				free_candidates_memory(candidates);
+				candidates = NULL;
 				return is_OK;
 			}
 		}
 
-		Tests();
-
-		PrintAndLog("");
-		PrintAndLog("Sum(a0) = %d", first_byte_Sum);
-		// PrintAndLog("Best 10 first bytes: %02x, %02x, %02x, %02x, %02x, %02x, %02x, %02x, %02x, %02x",
-			// best_first_bytes[0],
-			// best_first_bytes[1],
-			// best_first_bytes[2],
-			// best_first_bytes[3],
-			// best_first_bytes[4],
-			// best_first_bytes[5],
-			// best_first_bytes[6],
-			// best_first_bytes[7],
-			// best_first_bytes[8],
-			// best_first_bytes[9]  );
-		PrintAndLog("Number of first bytes with confidence > %2.1f%%: %d", CONFIDENCE_THRESHOLD*100.0, num_good_first_bytes);
-
-		clock_t time1 = clock();
-		generate_candidates(first_byte_Sum, nonces[best_first_bytes[0]].Sum8_guess);
-		time1 = clock() - time1;
-		if ( time1 > 0 )
-			PrintAndLog("Time for generating key candidates list: %1.0f seconds", ((float)time1)/CLOCKS_PER_SEC);
-	
-		brute_force();
+		//Tests();
 		free_nonces_memory();
 		free_statelist_cache();
 		free_candidates_memory(candidates);
 		candidates = NULL;
-	}	
+	}
+	*found_key = foundkey;
 	return 0;
-}
-
-
+}
\ No newline at end of file